Analysis tools

This chapter covers the analysis tools: the cross-asset and macro reads (rotation, correlation, yield curves, event impact, regime analogs, Treasury auctions), the company comparison tools (relative value, relative valuation, the merger calculator, merger arbitrage, buybacks), the insider and congressional trade screens, the semiconductor supply-chain tracker, and the working tools (position sizing, Ask the Data, the report generator, risk models and saved models). Signed in, almost all of them sit under Tools in the desktop menu; Risk Models and Saved Models sit under Data › Models & Data. Many also have a free public page that anyone can open without an account. The options, futures, COT, dealer positioning, short volume, greeks and prediction-market tools have their own chapters.

The pages most readers start with, and the question each answers:

• Market rotation: which asset class, and which group inside it, has led or lagged over each window.
• US Treasury yield curve: what shape the curve is in today, and how it compares with any past date.
• Relative value: how one company compares with its industry peers on growth, margins, returns and valuation.
• Merger arbitrage: which announced deals are pending, and what spread the target's price leaves to the offer.
• Insider clusters: where several insiders traded the same stock in the same direction within a month.
• Ask the Data: a question in plain words, answered from Kresmion's own stored data.

The tools hub and the two ways in

Every tool opens at an address under /tools/. A tool is either public (anyone can open it; it was built as a free page that search engines index) or signed-in only (part of the workspace; it needs a free account). Public tools show a richer terminal layout once you are signed in, at the same address. This section explains how that works once, so the page sections below only say which applies.

Tools hub

One list of the free public tools, grouped by asset class, each with a one-line description of the question it answers.

• Header. "Market tools", with the breadcrumb Home › Tools and a context line: "N tools on the data behind the Kresmion terminal, with the source and date on the numbers that carry them. Open without a sign-in unless a tool says otherwise."
• Groups. Each group has a coloured heading with a count: Equities (insider clusters, congressional trade anomalies, buybacks, relative value, merger arbitrage board, IPOs and offerings, sector rotation, off-exchange short volume, implied probability), Macro and rates (yield curve, Treasury auctions, primary dealer positions, central banks, money markets, historical regime analogs), Futures and commodities (contango and backwardation, CFTC COT positioning, futures seasonality, futures volatility and squeeze), Cross-asset (market rotation, correlation regime), Prediction markets (prediction market screener) and Utilities (position size calculator). A group with no tools is not shown.
• Rows. The tool's name in bold and, beside it (below it on a phone), what it answers. The whole row is the link.

The hub lists only public tools. Signed-in-only tools (Relative Strength, Currency Strength, Drawdown Tracker, Equity Rel. Valuation, Merger Calculator, the supply-chain tracker, Ask the Data, Report Generator, Risk Models, Saved Models) and the greeks suite are not on it; find them in the Tools menu or with Cmd-K.

Every public tool page has the same header: a breadcrumb back to this hub (Home › Tools › tool), the title, a context line, and a closed How this is computed disclosure that holds the method, the sources and a dated one-sentence summary of the current reading. Click it to open.

Public pages and the signed-in desk

What happens when you open a public tool.

• Signed out (or a search engine): you get the public page. It has the public top bar and footer, the breadcrumb, the method disclosure and, on several tools, a frequently asked questions block. Some public pages show a bounded snapshot (for example the latest clusters only) and invite you to sign in for the rest.
• Signed in: the same address (the address bar keeps /tools/<tool>) opens the tool inside the signed-in workspace, with your menus, Cmd-K and keyboard shortcuts. For most tools this is a terminal desk: a title strip with the controls on the right, a strip of headline figures (KPIs), then panels, and a Sources link at the foot that opens /settings/data-sources at the tools family. On several tools the desk shows more than the public page (full history, extra panels, signed-in-only price layers); each page section below says what differs under Signed in.
• The switch happens on the server and is checked against your live session, not just a cookie. If the session cannot be confirmed (for example the backend is briefly unreachable) you see the public page until it can. A tool embedded in a terminal canvas panel always shows the public page.

Market Rotation, Sector Rotation, Correlation Regime, US Yield Curve, Macro Event Impact, Regime Analogs, Treasury Auctions, Relative Value (the hub and each company page), Merger Arbitrage, Buybacks (the hub and each company page), Insider Clusters (the hub and each ticker page), Congress Anomalies and the Position Calculator, plus the options, futures, COT, dealer, short-volume, greeks and prediction-market tools covered in other chapters.

Relative Strength, Currency Strength, Drawdown Tracker, Equity Rel. Valuation, Merger Calculator, the Semiconductor Supply Chain tracker, Ask the Data, Report Generator, Risk Models and Saved Models live only in the workspace. Signed out, their address shows a "Sign in to open page" card with Create a free account and Sign in (both bring you back to the page afterwards) and a link to the free public tools on the tools hub. (The old /tools/yield-curves address is not a separate page: it forwards to the yield curves desk, and the WB code still opens it.)

The /dtools/ addresses. Behind the scenes the signed-in desk of a public tool is served from an internal /dtools/<tool> address. Nothing in Kresmion links to it and you never need it: the address bar always shows /tools/.... If you paste a /dtools/... address while signed out, you are sent to the matching public page (or to the tools hub when there is none). A /dtools/<name> address that matches no tool shows the not-found page.

Cross-asset and macro reads

Rotation, correlation, rates and regime tools. Signed in, all of them are under Tools › Cross-Asset & Macro.

Market rotation

Which asset class (stocks, crypto, FX, commodities) and which group inside each class has led or lagged over one day, one week, one month, three months and the year to date, with each series' trend state and the breadth of each class.

• Window buttons (1D, 1W, 1M, 3M, YTD). Every number on the page uses the window you pick; the page opens on 1W. On the Crypto tab the 1D button reads 24H, because crypto's one-day figure is a rolling 24-hour change. Deeper windows are close-to-close returns over calendar days, so a week is the same seven days for a coin that trades every day and an ETF that trades five.
• Overview tab, class cards. One card per class. The card shows the class name, a freshness badge (LIVE green: the latest expected session is present; DELAYED amber: one session missing, usually a holiday; STALE red: older than that; hover for the reason), the class Anchor (the benchmark the class is measured against; "EW basket" is an equal-weight basket of the class's own members), the class return in the chosen window (green when up, red when down), a breadth line "N/M above 20d MA" with a thin bar (green when half or more of the members close above their 20-day average, red below half) and "as of" the last close date. "Breadth n/a" means the class reported no members for breadth.
• Group leadership over 1M. For each class, the group with the best average one-month return ("Stocks: group leads (▲ x%)").
• Cross-asset line. "Over window: class leads at x%, class trails at y%", plus a Full macro regime → link.
• Class tabs (Stocks, Crypto, FX, Commodities). Group chips at the top give each group's average return (1W when the window is 1W, otherwise 1M; hover for the member count). Below, Bars ranks every series in the class by the chosen window: name and symbol, a bar centred on zero (green right, red left, scaled to the largest move on screen), the return, and a trend chip: UP (close above both its 20-day and 50-day averages), DOWN (below both) or MIXED (between them); a dot inside the chip marks a strong trend. A series that lags the rest of its class carries an amber Stale tag and is dimmed; hover for why. Hover a name for the full name, symbol and last price.
• Quadrant view. The relative-rotation plane for the class: across, the series' price relative to the class anchor (RS, where 100 is its own 60-session average); up, the 10-session change of that ratio (momentum). The four corners are labelled LEADING (strong and gaining), WEAKENING (strong but losing momentum), LAGGING (weak and losing) and IMPROVING (weak but gaining). The anchor sits at the centre. Hover a dot for "RS x · Mom y". The quadrant does not change with the window buttons.
• FX tab, currency strength. For nine major currencies, each currency's average appreciation against the others, in percentage points against the tracked mean, derived from USD-quoted pairs (1W for the 1D and 1W windows, 1M otherwise). A positive reading is broad strength against the basket, not a statement about one pair. "Excluded (insufficient history)" lists currencies left out.
• Commodities tab note. 1W to YTD are the front-month futures price change, the same measure as every other asset; hover a commodity figure (dotted underline) for the roll-adjusted return, which includes roll yield. 1D is close to close on one contract's settlements, so a roll day does not read as a move; live quotes are indicative.
• Footer of each tab. "As of date · computed n min ago" and the freshness badge.
• Frequently asked questions. Four short answers below the board (what rotation is, how returns are calculated, what the strength meter shows, how fresh the data is).

Signed in: the desk shows the same snapshot in the terminal layout.

• Title strip. "Market rotation" with "data through day · descriptive, not a forecast", a view switch (Overview, Stocks, Crypto, FX, Commodities) and the window switch.
• KPI strip. One tile per class with its return in the window, "N/M above 20d avg" and the freshness word (live, delayed, stale). Hover a tile for the class anchor. The 1D figure is stamped with the latest quote time ("1D quote"); for commodities, and when no recent quote exists, it is the 1D close.
• Which asset class leads, in every window. Grouped bars: one group per window, one bar per class in its category colour. Click a window's group to make it the active window; hover to read "Over window, class leads ... and class trails ...".
• Breadth. Share of each class's members closing above their 20-day average, with "above/members". Click a class to open its view.
• Group leadership over one month. Every group of every class ranked by average 1M return, in two halves. Columns: Group (swatch in its class colour), Class, Members, a bar, 1M. Click a group to open its class filtered to that group.
• A class view. An "As of" note line (close date, computed time, freshness and, for commodities, the 1D rule). Ranked by window return: Symbol (or Pair for FX), Name, Group, Close vs 20d / 50d ("above 20d and 50d", "below 20d and 50d", "between 20d and 50d", "strong" when the trend is strong, "not tracked" when the series is too short), a bar and the return. Switch to Quadrant in the panel bar for the relative-rotation chart; hover a point for RS, momentum and its quadrant. Groups lists each group's average return; click one to show only that group, and All groups to clear. On FX, Currency strength shows the same nine-currency meter as bars.
• Colour. Classes keep their category colours; groups inside a class take a series colour that is a stable label, not a ranking. Red and green appear only on signed returns.
• Sources at the foot opens /settings/data-sources at the tools family.

• Pick the window first, then the class. On the public page a click on a class card opens its tab.
• The board refreshes on its own every 5 minutes; the snapshot behind it is recomputed about every 10 minutes.
• Nothing is remembered between visits: the page opens on Overview and 1W.

Read Breadth before the class return: a class can be up on a few large members while most of them sit below their 20-day average. For the 11 sectors on their own use Sector rotation; for a volatility-adjusted view of every asset use Relative strength. The one-day figure here is the latest quote against the prior close, so it can differ from the close-to-close 1D on the relative-strength board; each page states its own source and time.

Sector rotation

How the 11 US GICS sectors, through their sector ETFs, rank against each other over windows from one day to the year to date, and whether the leaders are defensive, cyclical or a mix.

• Window buttons (1D, 1W, 1M, 3M, YTD). The page opens on 1D. If the main rotation snapshot is offline the page falls back to the sector quote feed: then only 1D and 1W work, the deeper buttons are dashed and greyed ("Deep windows unavailable while the rotation snapshot is offline"), and the header says "Using the available quote feed".
• Ranked bars. One row per sector: the sector code and ETF ticker (for example TECH · XLK), a bar centred on zero (green to the right for a gain, red to the left for a loss, scaled to the largest move on screen) and the return to two decimals. Strongest at the top.
• Leadership (top 3 sectors). A fixed rule on the three best sectors in the window: "Leadership: defensive" when all three are in the defensive basket (Utilities XLU, Staples XLP, Health Care XLV); "Leadership: cyclical" when at least two are in the cyclical basket (Technology XLK, Financials XLF, Discretionary XLY, Energy XLE, Industrials XLI); otherwise "Mixed leadership" with the top three and bottom three. Real estate, materials and communication services sit in neither basket. With fewer than six sectors reporting it reads "Insufficient sector data". The baskets are fixed lists, not a reading of the market.
• Snapshot line. Above the bars, a line says what a click does ("Sign in to open a sector's companies from its bar." when you are signed out) and "Snapshot computed time UTC"; under them, "Deep windows as of date close" for any window other than 1D.

Signed in: the desk adds summary figures and a heat grid.

• Title strip. "Sector rotation", with "1W to YTD through the day close · 1D on the latest quote, time · snapshot time", and the window switch.
• KPI strip. Leader and Laggard (ticker and return), Leader minus laggard (the dispersion across the 11 sectors, in percentage points) and Cyclical minus defensive (the equal-weight cyclical basket's average minus the defensive basket's, in points).
• Sectors ranked, window. The 11 ETFs as bars with a value axis; each bar's swatch shows its style basket (cyclical, defensive, other). Hover for the full name and basket.
• Every window at once. A heat grid: one row per sector, one column per window, tinted by size within each column (1D and YTD live on different scales), the chosen window framed. Hover a tile for the two-decimal value and its date.
• Style baskets. The equal-weight average of each basket in the window.
• Who leads. The top three with their returns, and a count line: "Top 3: n cyclical, n defensive, n other. Bottom 3: ...". It states what it counts, never a risk-on or risk-off call.
• A Fallback note appears when the quote feed is in use.

• Pick a window; the bars, the leadership line and (signed in) the KPIs and framed column follow.
• The page refreshes every 5 minutes. The window is not remembered.

The one-day figure here comes from the latest live quote against the prior close, while deeper windows are closes; the relative-strength board uses close to close for 1D, so the two pages can differ on the same ETF during a session. For sectors beside crypto, FX and commodities, use Market rotation.

Relative strength

A cross-asset board that answers three questions per asset: what moved, how unusual that move is for the asset itself (measured against its own volatility), and whether it is gaining or losing ground on its class benchmark. It also builds the price ratio of any two assets. These are historical statistics, not predictions.

• Title strip. "Relative strength" with the snapshot date and computed time, a Class selector ("All classes · N" or one class with its count and its own as-of date) and three column-group toggles: Returns (1D to YTD, coloured by z), Relative (strength against the class benchmark) and Range (where the price sits in its own year). Returns and Relative are on by default.
• KPI strip (for the chosen class). Moved most, for itself and Fell most, for itself: the highest and lowest one-day z, with the z value. Gaining most on benchmark: the largest 10-session change in the ratio to the benchmark. Above 50-session mean: the share of rows closing above their 50-session average ("N of M"). Rows: how many rows are in view.
• The z legend (note line). z is the move divided by that asset's own 60-session volatility. The bands are fixed every day: under 0.5 ordinary, 0.5 to 1 notable, 1 to 2 large, over 2 rare. On the board the number is the raw move and the wash behind it is the band; green is up and red is down. The wash shows how unusual the move is, not how good it is.
• Average 1D z by class. One bar per class with "N up · N down". Each class closes on its own day (crypto trades through the weekend), so a bar's tooltip names its own close date.
• Above the 50-session mean. The share of each class closing above its 50-session average.
• Leaders and laggards. Two lists for the chosen class: Moved most today, for itself (top and bottom by 1D z) and Gaining and losing on the benchmark (top and bottom by the 10-session change of the ratio).
• Board. Always shown: # (rank by 1D z across the whole board, 1 is the strongest), Asset, Name, Class, Price 60d (60-session price line, oldest on the left) and, once a second daily snapshot exists, Rank 1d / 1w (rank change since the previous snapshot and since a week ago; "held" means unchanged). Returns group: 1D, 1W, 1M, 3M, YTD, washed by their z band; the sorted column also shows the z as a small superscript. Relative group: RS (the ratio to the class benchmark against that ratio's own 60-session mean, 100 = moving with the benchmark; green at or above 100, red below), RS mom (the 10-session change of that ratio), Trend (improving above +0.5, fading below -0.5, otherwise flat), Mom score (the mean within-class percentile of the 1M, 3M, 6M and 1Y-minus-1M returns) and RS 60d (60-session line of the ratio). Range group: 52w range (a bar and the percent of the way from the 52-week low to the high), Own year (percentile of the 1M move within the asset's own year), MA 20/50/200 (three dots: filled green when the close is above that average, hollow red when below, hollow grey when the series is too short) and Days off high (calendar days since the 52-week high). VIX is quoted in points ("pt"), not percent.
• "--" is a measurement the store does not have (too little history, fewer than 70 sessions aligned with the benchmark, and so on); hover it for the reason. It is never a zero.
• Row detail (click a row). The asset's 60-session price chart with an Open chart link, its RS line against the benchmark (100 dashed), every window (Window, Calendar days, Move, z, Sessions, How unusual: the Sessions column shows how many trading sessions the calendar window actually held for that asset) and facts (close, 52-week range, position in the range, days since the 52-week high, own-year percentile, close against the 20, 50 and 200-session means, 60-session volatility, RS, RS 10-session change, momentum score). Close shuts it.
• Pair. After you pick a second asset, the ratio of the two over time with its 20 and 50-session means, "Ratio now" and z against its own year (the current ratio minus its 252-session mean, divided by that window's standard deviation; "not tracked" with less than a year of aligned sessions), and a table of the ratio's change by window. Close pair removes it.
• Method. The rules in full: calendar-day windows, YTD from the previous year's last close, 1D close to close, adjusted closes, roll-adjusted continuous futures for commodities, VIX in points, the z construction, the benchmarks for each class (the commodity basket is an equal-weight, daily-rebalanced index of the class's own members), per-class as-of dates and ranks. "Stated by the source" lists any caveat the data job published with the snapshot.

• Class. Pick a class to restrict the KPIs, leaders and board; or click a bar in either class chart (click again to release).
• Find ( / ). Press / anywhere to focus the board search; it matches symbol or name. Esc clears it.
• Sort select in the board bar (Rank, Symbol, Class, each return window, RS, RS momentum, Mom score, 52w position, Own year, Days off high) and Asc / Desc to reverse it; missing values stay last. On a narrow screen columns drop out, but the sorted column always stays, so the sort select doubles as a column picker.
• Compare with... in the facts panel of an open row builds the pair.
• Shareable state. The class, sort, direction, search, column groups, open row and pair are all written into the address (?class=&sort=&dir=&q=&cols=&sym=&pair=), so you can bookmark or send a view. Clicking through rows does not add to your back-button history.
• The board refreshes every 10 minutes; the snapshot behind it is daily.

Sort by RS mom to see what is gaining on its benchmark regardless of the day's move, then open the row to check the RS line. The 1D column is close to close, so during a session it can differ from the quote-based 1D on Market rotation and Sector rotation. Rank changes need two daily snapshots; on the first day the board says so under the table.

Correlation regime

Which pairs of assets that usually move together (or against each other) have recently decoupled, across a fixed 45-symbol universe of equities, sectors, bonds, commodities, FX, volatility and crypto. A correlation matrix shows the state; this page shows where the state has changed.

• The rule line. One sentence at the top states the rule every count, filter and table on the page uses: a pair counts as a break when its 30-day correlation moves more than 0.40 away from its 1-year average and also flips sign, doubles or halves, and it stays a break until that gap is back under 0.30 (so a pair near the line is not counted in and out every day). The two thresholds are printed from the data, so they always match the counts.
• Baseline and current. Current is the 30-day rolling Pearson correlation of simple daily returns. Baseline is the mean of that same rolling series over the trailing 252 trading days (about a year). Both run from -1 (move opposite) to +1 (move together).
• Δ (change). Current minus baseline. Colour is not a market direction: amber (gold signed in) means the correlation strengthened, teal (cyan signed in) means it weakened. The sign is always printed. sign flip marks a pair whose correlation changed sign.
• Strongest current breaks (public). Three cards, new breaks first, then the largest: the pair with each asset's class badge, "baseline x → current y", the Δ and the reason.
• The pair table. Columns Pair, Classes, Baseline, Current, Δ and Reason. The reason is the rule that fired (for example the size of the gap, sign flip, magnitude doubled or halved with the ratio). "still in break, gap under 0.40" means the pair crossed the entry line earlier and has not yet fallen under the exit line. The line above the table counts "Showing N of M pairs · K breaks total (in above 0.40, out under 0.30)" with the snapshot date.
• Pair detail (click a row). A chart of the pair's 30-day correlation (amber line) against its baseline (dashed) over the stored history, and Historical comparable breaks: earlier dates when this pair broke by a similar size and in the same direction, each with its Δ, baseline, current and reason. An empty list means none in the backfill window yet.
• Coverage. The symbols in each class, how many, and the last date each class has data for.
• Methodology footer. The rule in full, and the note that cross-class pairs (crypto against equities, for example) are date-aligned so weekend prints do not contaminate the equity series.

Signed in: the desk reads the same data in the terminal layout.

• Title strip. "Correlation regime", the symbol count and snapshot date, two class selectors (Class A, Class B, "Any" by default) and Breaks only / All pairs.
• KPI strip. Breaks in view (or Pairs in view) with "N of M pairs, |d| > 0.40"; Sign flips; Stronger / weaker (how many pairs strengthened against weakened); Largest change with the pair.
• Biggest correlation changes. The 14 largest changes as dumbbells: a hollow dot at the baseline, a filled dot at the current value, joined by a line in gold (stronger) or cyan (weaker); the two small dots on the left are each asset's class colour. Hover for the numbers, click to pick the pair.
• Pair history panel (titled with the pair). The 30-day correlation against the dotted 252-day baseline, with the pair's classes, baseline, current and change in the read-out.
• Earlier breaks like this one. Date, Change, Baseline to current and "Why it counted" for comparable past breaks.
• Where the changes sit. A class-by-class grid counting the pairs in view for each pairing (on a phone, a list). Click a cell to set Class A and Class B to that pairing.
• Every pair. The full table (Pair, Classes, Baseline, Current, Change, Why it counts), sortable by Change, Pair, Current or Baseline, with a ticker filter. Click a row to load its history in the pair panel.
• Coverage and Method panels, then Sources.

• Class A / Class B. Restrict to pairs that involve those classes ("any" or "Any" for no restriction). Classes: crypto, equity, sector, bond, commodity, forex, vol.
• Only breaks (checkbox, public) or Breaks only / All pairs (signed in). Untick to see every pair in the universe, not only the ones in break.
• Search ticker. Type a symbol (for example SPY) to keep only pairs containing it.
• Click a row to open (public: expand inline, click again to close) or pick (signed in) its detail.
• The snapshot updates daily. Filters are not remembered between visits.

A break says two series have stopped moving the way they usually do; it says nothing about which one will move next. Start with Class A set to bond and Class B to equity to see the stock and bond relationship, or search a single ticker to see every pair it belongs to. Comparable breaks show how often this pair has decoupled before in the stored history.

Currency strength

Which of eight major currencies (USD, EUR, GBP, JPY, AUD, CAD, CHF, NZD) has risen or fallen most against the others over the last day or week, scored from every spot pair the forex matrix tracks plus the US Dollar Index.

• Title strip. "Currency strength", "8 majors ranked by average move across every tracked pair (N pairs)" and the data time. When the matrix's own stamp is more than two hours behind an open FX market (Sunday 22:00 to Friday 22:00 UTC), the line says "last update time, feed behind by nh" and a Stale note under the KPIs says every figure is from that update, not live.
• KPI strip. Strongest and Weakest (currency and its average move), Strongest minus weakest (the gap in percentage points; its sub-line says whether the strongest is up and the weakest down against the rest, or "none both up and down vs the rest") and Pairs read (spot pairs and DXY).
• All 8 majors, strongest first. One bar per currency: its average percent move. Under each label a word: strong (scaled score above 60), weak (below 40) or neutral. The scaled score rescales the averages so the largest absolute move sits at 0 or 100 and a flat currency at 50; hover a bar to see it with the number of pairs used. A currency with no tracked pair reads "no pair" and has no score (nothing is estimated).
• Every tracked pair. A heat grid, one row per pair (DXY included), with 1D and 1W side by side, the chosen window framed and sorted strongest first. Hover a tile for its two-decimal move and which currencies it counts toward. A pair raises its base currency's score and lowers its quote currency's by the same move; DXY counts toward USD.
• How the score is built. The method in two sentences, as above.

• 1D / 1W switch in the title strip. The page re-reads the matrix every 5 minutes. Nothing is remembered.

The score is only as broad as the pairs behind it: when most tracked pairs are quoted against the US dollar, USD's score rests on many pairs and the others on few. Hover a bar to see the count. The nine-currency meter on the FX tab of Market rotation is built differently (each currency against the tracked mean, over one week or one month), so the two rankings can differ. For the full currency desk see FX monitor.

Drawdown tracker

How far each tracked asset (crypto, equities, indices, sectors, commodities, FX) sits below its all-time high, how long ago that high was, and whether the current drawdown is deeper than a typical bear market for its asset class.

• Title strip. "Drawdown tracker", "Distance below the all-time high across N assets · computed time", and a Class selector.
• KPI strip (for the class in view). Median below ATH; Within 5% of ATH (how many assets sit within 5% of their high); Deeper than a typical bear (how many are past their class marker); Deepest (the asset with the largest current drawdown).
• How far below the high, by asset class. One bar per class: the median distance below the all-time high across its assets, in the class colour, with a tick at the class's typical bear-market drawdown (a fixed reference per class).
• Every asset table. Asset, Name, Class (chip), a bar headed "Below ATH, tick = typical bear" (the bar is in the class colour and turns red once the drawdown is deeper than the tick), Below ATH (current drawdown, red past the marker), ATH price (the all-time high close; hover for its date), ATH date, Days since (calendar days since the high), Max drawdown (the deepest peak-to-trough fall in the full history).
• Not shown (under the table). Assets left out because the price series the vendor supplies is not the asset itself, each with the reason.
• Cached note. On a repeat visit the page paints at once from the copy your browser last read, labelled with its computed time, while a fresh read runs and replaces it.
• A first read can take several seconds because the server reads each asset's full daily history to find its high; the page says so and counts the seconds.

• Class selector, or click a class bar to filter (click it again for every class).
• Jump to search in the table bar: it highlights and pages to the first matching symbol or name ("Jumped to X", or "No asset matches"); it does not filter. The highlight fades after 3 seconds or on a click outside the table. Esc clears it.
• Sort: Deepest first (default), Nearest the high, Longest since ATH, Deepest max drawdown, Symbol A-Z.
• The table pages instead of scrolling; the pager sits in its bar.
• Your browser keeps the last copy of the data for the fast repeat paint; the class and sort are not remembered.

Sort by Longest since ATH to find assets whose high is years old even when the current drawdown looks shallow. The typical bear marker is a fixed reference per class, not a statistic of today's market.

Yield curves

Where the US Treasury curve sits today, how it has moved against a year ago (or, signed in, any archived day), whether the 2-year/10-year spread is inverted, and how the euro-area and other government curves compare.

• Header. "US Treasury yield curve" with the GC function-code tag, a line saying the page compares the latest curve with one year ago and refreshes hourly, and a closed method note saying Kresmion plots the curve and computes the 2Y/10Y spread and the inversion flag.
• Inversion badge (top right of the chart area). Red "INVERTED 2Y/10Y · Nbps" when the 2-year yield sits above the 10-year; green "Normal slope · 2Y/10Y +Nbps" otherwise. The figure is the 10-year minus the 2-year, in basis points.
• Chart. Yield in percent per year (left axis) against maturity (bottom axis, from the shortest bill to 30 years). The amber solid line with dots is the latest available curve; the teal dashed line is the curve one year ago. The footer shows "Cached: date time UTC", the time the snapshot was taken.
• Yields table ("Treasury yields by maturity, percent per year"). One row per maturity: Latest, 1 year ago ("Unavailable" when that tenor was not published a year ago) and Change (bps), the difference in basis points, coloured green when the yield rose and red when it fell. The source line links to the US Treasury daily yield curve rates and notes the values are published observations, not intraday quotes.
• "Open the curve archive with a free account" box: the three extra views an account adds (overlay the curve from any published day back to July 1969, a monthly animator, the spread history). Create a free account and Sign in both return you to this page.

Reading it (signed in: the Yield curves desk). The page is titled "Yield curves"; its context line names the US closing day the figures describe and says it refreshes hourly. It scrolls as one page.

• KPI strip.
  • US 10Y and US 2Y: the constant-maturity yields (FRED), with the change in basis points against the same series a year earlier.
  • US 2s10s and US 3m10s: the 10-year minus the 2-year and minus the 3-month, in basis points, with the word "inverted" (below zero) or "positive slope".
  • US inverted for / US positive for: how long the 2s10s spread has kept its current sign, counted from its last crossing of zero (daily closes). "not reported" means the tracker has no count.
  • EA 10y-2y: the euro-area spread from the ECB yield curve of all euro-area central government bonds (all ratings), with its state; EA inverted for / positive for is the same count for the euro area.
• US curve. Two views on a toggle:
  • Now vs then: the latest curve (solid, with dots) against a comparison curve (dashed). The "vs" buttons pick it: 1Y ago (the year-ago curve), 1M ago, 6M ago, 2Y ago, 5Y ago, 10Y ago, or any day typed in the date box (limited to the archive's first and last day). If the day you pick was not a published day, a note gives the nearest published day used instead. The legend names each line.
  • Through time: a monthly animator. Play/Pause, ‹ and › step one month, the slider scrubs, and the range buttons (5Y, 10Y, 25Y, MAX) set how far back it starts. The readout shows the month, "i of n" and that month's 2s10s. The latest curve stays on the chart as a dashed reference.
• Change by tenor. One row per tenor: Latest, 1Y ago (or Then for an archive day), Chg bp and a bar centred on zero, scaled to the largest move. "--" means that tenor was not published on the compared day (for example, the 1-month bill before 2001).
• Spread history. Tabs 2s10s, 3m10s (US, from the daily par curve archive) and EA 2s10s (ECB). Range buttons 5Y, 10Y, 25Y, MAX. The line is the spread in basis points; shaded spans mark stretches below zero. The note under the chart counts the inverted stretches in the chosen range, names the longest with its dates, and gives the number of daily observations.
• Inversions. Tabs US and EA. Every run of the daily 10y minus 2y close below zero over the whole backfilled history, newest first: Start, End ("ongoing" if still inverted), Length in calendar days with a bar scaled to the longest run, and Deepest (the most negative spread in the run, bp). The panel header counts the runs and names the first year and source. Its description notes that every US recession since 1976 followed a 2y/10y inversion and its re-steepening; that is a record of past episodes, not a forecast.
• Euro area curves. Tabs Curve (the AAA and the all-issuer euro-area curves by tenor, from ECB yield curve data; 1M ago and 1Y ago buttons add the same curves earlier as dashed lines) and Tenors (a table with each curve's yield, 1M bp and 1Y bp change per tenor). "(pending)" on a column means that curve has not been stored yet.
• Sovereign 10Y. One row per government, its newest monthly 10-year yield. rank by buttons reorder the list: 10Y (the yield), Bund (the spread over the German 10-year, same month), UST (the spread over the US 10-year), 1M and 12M (the change in basis points). Columns: Ctry (ISO code, sometimes with a short tag for how the series is built, hover for the text), 10Y %, 1M bp, 12M bp, vs Bund, vs UST, Month, Ccy. A row whose newest month is older than the others' carries a tooltip saying so.
• Colour. Lines use series colours only. Green and red appear only on signed changes in basis points (with a small dead band, so a 1 or 2 bp wobble stays neutral).
• Sources (last line) opens the data-sources page at the yield curve entry, with the ECB attribution.

• Public page: the chart has no hover read-out (the table carries every value). The bell button next to the badge opens an alert popover on the 10Y minus 2Y spread (FRED series T10Y2Y). Signed out you can pick a condition and sign in to save it; signed in it saves to your alerts.
• Signed in: the same bell sits at the right of the title strip and saves the same 2s10s alert.
• Signed in: every pager sits in its panel's top bar; hover any column header for its unit and as-of. The compare buttons are disabled with an explanatory tooltip while the archive loads or if your session lapses ("sign in again to load it").

The KPI change is against one year ago; to see a different window, pick it in the US curve panel and read Change by tenor. To check whether an inversion is long or short by historical standards, use the Inversions table rather than the spread chart, which only counts what falls inside its range. For the euro-area country spreads over the Bund on a daily basis, see Sovereign spreads; for policy and overnight rates, Money markets.

Treasury auctions

How each recent US Treasury bill, note, bond, TIPS and FRN auction went (stop-out rate, bid-to-cover, who took the paper, the high minus median spread), each figure ranked against the same tenor's auctions over the previous five years, plus how much marketable Treasury debt matures in each time bucket.

• Header (public). "Treasury auction tracker", as-of date, and an "On this page" list (Tenor history, Maturity ladder, Method). The closed method note holds a computed sentence about the newest coupon auction and its rank.
• KPI strip.
  • Newest ranked coupon auction (signed in: Newest coupon auction): the newest note or bond auction with a percentile rank, its tenor and stop-out rate, its date and the rate type.
  • Its bid-to-cover: total tendered over total accepted, with its percentile ("Nth pctl of n prior"). A higher cover means more bids per dollar sold.
  • Its primary-dealer take (signed in: Its dealer take): the share of the competitive award left with primary dealers, with its percentile and the indirect share. A high dealer take means fewer end buyers took the paper at auction.
  • Debt maturing within 12 months: the share of unmatured marketable debt that matures within a year of the latest Monthly Statement of the Public Debt, with the dollar amount and, once enough statements are stored, its percentile.
  • Weighted average maturity: the outstanding-weighted average time to maturity, in years. This is a Kresmion calculation, not Treasury's published average.
  • Signed in only: Announced, not held, the number of announced auctions dated today or later.
• Tenor: auction by auction. Pick a tenor (grouped by class, each with its count of auctions) and a window (1Y, 2Y, 5Y). Three small charts: Bid-to-cover per auction; High minus median, bp (the stop-out rate minus the median accepted rate, with a zero line; a wider gap means the accepted bids were more spread out); and Who took the paper, stacked shares of the competitive award for Indirect (indigo), Direct (pink) and Primary dealers (amber). The panel note says if the five-year history did not load and only the recent list is shown. The tail against the when-issued yield is not shown because those dealer quotes are not public.
• The last N Treasury auctions. Class buttons: All, Bills, CMBs, Notes, Bonds, TIPS, FRNs. Columns: Auction date; Security (tenor, CUSIP, "reopening" where it applies; click to chart that tenor); High rate (bills: high discount rate; notes, bonds and TIPS: high yield; FRNs: high discount margin); High - median (bp); Bid-to-cover; Indirect, Direct, Dealers (shares of the competitive amount accepted, adding to 100%); Accepted (total, in dollars, with how many prior auctions it is ranked against); Flags. Under each ranked figure is its percentile (signed in shown as pNN). An amber-tinted cell ranks at or below the 10th or at or above the 90th percentile of its own tenor. "not ranked" (signed in "nr") means fewer than eight prior auctions of that tenor are stored, or a cash management bill. Two flags exist: low cover (bid-to-cover in the lowest tenth of prior same-tenor auctions over five years) and high dealer take (dealer share in the highest tenth); hover a flag for its definition. A flag describes one auction against its own history and says nothing about the next one.
• Marketable Treasury debt by time to maturity (signed in: Maturity ladder). One bar per maturity bucket, stacked by security class (legend above). Outstanding shows dollars; Share shows the percent of the total, with a tick at that bucket's share a year earlier. Hover a bar for its securities count and share. The note under it gives the 12-month figure and its rank, the weighted average maturity, and explains why Treasury's Total Marketable line is larger (it also counts matured debt not yet redeemed and the Federal Financing Bank line). TIPS amounts include the inflation adjustment. Signed in, a third tab, Facts, lists the statement date, securities count, totals, the 12-month share and its rank, the weighted average maturity, Treasury's Total Marketable and Federal Financing Bank lines, the year-earlier figure and how many statements are stored.
• Announced, not yet held (signed in: Announced auctions). Auctions Treasury has announced: Auction date, Security, Offering amount, Issue date and CUSIP. Signed in, a tag marks one scheduled "today" or "results pending" (held, results not stored yet), and the header counts each.
• Method and limits (public) and the Sources line: U.S. Department of the Treasury, Bureau of the Fiscal Service (Treasury Securities Auctions Data and the Monthly Statement of the Public Debt, public domain). Percentiles, shares, flags and ladder buckets are Kresmion calculations.

• Public: sort the auction table by clicking a column header; type in "Filter by tenor, CUSIP or date" to narrow it.
• Signed in: use the search box in the table bar ("Tenor, CUSIP, date"); click a row to chart its tenor (the selected tenor's rows stay highlighted); long tables page in their panel bar; hover any header for its definition.

To judge a single auction, read its percentiles rather than the raw cover: a 2.4 cover can be normal for one tenor and weak for another. A tinted cell plus a flag is the quickest way to spot an auction that sat in the tail of its own history. The ladder's 12-month share is the figure to watch for how much debt needs refinancing soon.

Macro event impact

What US assets did after Fed rate decisions, CPI releases and payroll (NFP) releases, measured from the last close before the event over 1, 5, 21 and 63 trading sessions, set against each asset's ordinary session (its unconditional baseline). Historical statistics, not predictions.

• Header. "Macro event impact", with the date of the latest event in the study. The closed method note holds a one-paragraph answer computed from the data (for example SPY's median move over Fed holds and CPI sessions) and the upstream sources: the Federal Reserve FOMC calendar, the FRED Fed funds target range, FRED release dates and the FRED 2-year Treasury yield.
• The headline matrix ("all events, 1 trading session"). Rows are seven assets: SPY, QQQ, IWM, TLT, GLD, the dollar index and the VIX (each labelled percent or points). Columns are five event types: Fed hold, Fed cut 25bp, Fed hike 25bp, CPI release and Payrolls (NFP), each with its count (n=). A cell shows the median move over the event session (green above zero, red below, with a tint that deepens with size) and hit (the share of events where the asset closed higher). A * marks fewer than ten events: read it as a short list of cases. The VIX is in index points; every other asset in percent. "no data" means the cell has no observation.
• Every horizon, per event. One collapsible block per event type (the first opens by default), titled with the count and the years covered. Per asset and horizon (1, 5, 21, 63 sessions): the median move, the interquartile range ("q1 to q3"), the hit rate against the baseline hit rate ("hit 56% vs 53%"), and vs base, the event median minus the asset's baseline median.
• "The full event study with a free account" box lists what signing in adds, with Create an account and Sign in.
• Method and limits. A move runs from the close before the event to the close of the horizon session, so the 1-session move spans the announcement; the baseline is every stored session with no condition; coverage dates for ETF closes and continuous futures; the Fed surprise buckets use the 2-year yield change across the decision.
• Cite this data box at the foot, with the URL, as-of date and the calculation sentence.

Title strip "Macro event impact" with three controls on the right: Event (every event type grouped under Fed decisions, CPI and NFP, each with its n), Bucket (All, Hawkish, In line, Dovish; Fed decisions only, greyed for CPI and NFP) and ETF / Futures (which rows the matrix shows: ETF and index closes back to 2000, or continuous back-adjusted futures from 2016).

• KPI strip.
  • Next FOMC, Next CPI, Next NFP: the next dated release of each family with a countdown. "not in the bundle calendar" means no date is stored. Clicking one switches the study to that family.
  • Study: the selected event type, bucket, n and the years covered.
  • Asset median *horizon*: the selected cell's median move after the event, with n and the baseline median.
  • Delta vs baseline: event median minus baseline median, with its 95% bootstrap confidence interval ("CI low to high").
  • Hit rate: the share of positive moves, with the baseline hit rate and p (a two-sided binomial test of the event's hit count against the baseline rate; a small p means the hit rate differs from an ordinary session's by more than chance alone tends to produce).
  • A Selection note under the strip states the surprise thresholds (hawkish at or above, dovish at or below a basis-point change in the 2-year yield across the decision), which rows the matrix shows, and when the bundle was generated.
• Cumulative median path · asset. Every event lined up so the close before the event day is 0; the line is the median position each session from 5 before to 30 after, the shade the middle half (Q1 to Q3). Hover, or use the arrow keys, for a session's median, q1, q3 and n. The asset picker sits in the panel bar.
• Every instance at horizon. One dot per event at the chosen horizon (tabs 1d, 1w, 1m, 3m). With Fed decisions on the All bucket the dots sit in lanes by surprise bucket. Hover or focus a dot for its date, action, surprise and move; click (or Enter) to pin it, Escape or click again to release. The panel says how many instances are plotted and how many have no completed window yet.
• Matrix. Assets (grouped Equity, Rates, Credit, Commodity, FX, Volatility) against the four horizons, the median per cell. Tile colour and text sign are the sign of the median and nothing else; a short amber rule marks n below 10 (anecdotal); "n/o" means not observable. Hover a cell for the baseline comparison in the read-out line; click it to make it the selected asset and horizon.
• Event against baseline · asset. A table per horizon in three sections: the event (n, Median, Q1, Q3, IQR, Min, Max, Hit rate), the baseline (n, Median, Q1, Q3, Hit rate) and the delta (Delta median, CI 95% low and high, p hits). Hover a row label for its definition.
• Event types: first, last, instances and Assets: first close, last close, sessions: coverage tables for every event type and every series.
• Every instance · asset. One row per event, newest first: Date (hover for the base close date), Action, Sched (scheduled meeting or not), Surprise (2-year yield change in bp), Bucket (Fed only), the move at each horizon, Evt mult (the event-day absolute move divided by the median absolute move of the prior 60 sessions) and VIX at the base date. Filters above it: VIX at event (any, above 20, 20 or below), Years from/to, scheduled only, and a Bucket filter for Fed decisions; clear resets them. With a filter on, a line recomputes n, median and hit rate for the filtered rows (no confidence interval; the bootstrap and p are for the whole bucket). Sort picks the column and the direction button toggles High to low / Low to high.
• Surprise, method and disclosure: the surprise method, caveats and payload version.

Every selection (event, bucket, asset, tier, horizon) is written to the address bar, so you can bookmark or share a view; the back button does not replay each click. Long tables page in their panel bar.

Always read n first: a 25bp hike column with a handful of events is a list of cases, not a distribution. "vs base" and Delta vs baseline tell you whether the event session differs from an ordinary one at all. For the release dates and the latest prints themselves, use the Economic calendar; for how markets price the next Fed decision, the Fed hub.

Historical regime analogs

Which stored days had a macro regime most like today's, measured on the four regime factors (Growth, Liquidity, Risk appetite, Volatility), and what nine markets did 5, 20 and 60 trading sessions after each of them. The page states plainly that this is descriptive history with small samples, not a forecast.

• Note at the top: the descriptive-history disclaimer.
• Header (public). "Historical regime analogs" with the date of today's stored regime row. The closed method note has a computed summary sentence, the calculation and the upstream links.
• Asset picker. The nine tracked series: SPY, QQQ, IWM, TLT, HYG, GLD, USO, the dollar index (DXY) and the VIX. The charts and the table's move columns follow the chosen asset.
• What asset did after each analog day (signed in: "asset after each analog day"). One coloured line per analog day, starting at 0 on the analog-day close and joining the three measured horizons (+5, +20, +60 sessions). The dashed ink line is the median of the completed windows; the shading is their minimum to maximum. A missing point is a window that has not closed yet. The day-by-day path between horizons is not drawn. Hover or use the arrow keys for the values.
• How each analog's factors compare with today's (signed in: "Factors: each analog against today"). The four factors in z-score units: solid ink is today, each colour one analog day. The closer an analog's line sits to today's, the smaller its distance.
• Shown chips (one per analog day, nearest first): click to hide or show that day in both charts; hover to highlight it. All and Nearest 3 reset the selection.
• The N nearest stored days. Columns: a swatch (show or hide in the charts), Date, Regime then (the engine's label that day), Distance (Euclidean distance on the four factors, unweighted; smaller is closer, with a bar), the four factor values, and the chosen asset's move at +5, +20 and +60 sessions. Move tiles are tinted by the sign and size of the past move (capped, with a grey dead band near zero); "pending" means the window has not closed. The last row is today, with distance 0.00 and "not yet" in the move columns. Picks are at least seven days apart and the last seven days are excluded, so one episode cannot fill the list. Click a header (public) or use Sort and the direction button (signed in) to reorder.
• What followed those days (signed in: "Every asset after the analog days"). Every asset against the three horizons: the median move, the min-to-max range, n with the count of up and down windows, and a mean only when at least ten windows have completed ("n<10, no mean" otherwise). "no completed window yet" means none of the analogs has reached that horizon. Signed in, click an asset to chart it above.
• Method and limits. How the distance and picks work, how a move is measured (a weekend analog day uses the Friday close), that the engine writes about every six hours so today's row can still change, the stored history span (the engine's history is short, so analogs come from recent months and many lack a 60-session window), and two known changes in the engine's inputs during July 2026.
• Signed in, the KPI strip adds: Regime today (label and smoothed score), Nearest analog (date, distance and its regime then), asset +20 sessions (the median of the completed 20-session windows with n and range) and Stored history (days stored, first and last date, and how many are pickable).

Pick an asset, toggle analog days with the chips or the row swatches, and hover rows to highlight a day in both charts (signed in, click a row to hold the highlight). Nothing is remembered between visits.

Read Distance before the moves: a far analog says little about today. When most cells say "pending" or n is 2 or 3, the table is a list of cases, which the page says itself. The four factor values and today's regime are explained in full on Macro overview.

Companies and deals

Peer comparison, valuation, deal maths and repurchases for one company at a time. In the menu they sit under Tools › Cross-Asset & Macro (Merger Arbitrage under Tools › Positioning & Flows).

Relative value hub

The starting point for comparing one company with the rest of its industry group. You pick a company here; the comparison itself opens on the company page of the tool.

• Open a company. A search box ("Ticker or company, e.g. AAPL", listed equities only) and six worked examples: NVDA against semiconductors, JPM against diversified banks, LLY against pharmaceuticals, XOM against integrated oil and gas, SAP (a euro reporter) against application software, MSFT against systems software. Signed in, the examples are a small table with Ticker, Industry group and Example columns.
• What the desk shows (signed in only). Six one-line descriptions of the signed-in desk: Peer matrix (31 metrics in nine column groups, each cell shaded by its peer percentile), Distribution, History, Peers (an editable peer set), Provenance (every cell opens its period, currency, form type and accession) and Shareable state (columns, metric and peers live in the URL).
• How the comparison is kept honest. The five rules every figure follows:
  1. Every fundamental figure for one company comes from one annual filing, so ratios share a period and a currency. P/E combines the latest market price with annual earnings; one-year return uses market prices, dated separately.
  2. Nothing is trailing twelve months. Quarterly (10-Q) figures are year to date and the fourth quarter is not a reported fact, so neither is used.
  3. Nothing is converted between currencies. A peer reporting in EUR or JPY still compares on margins and returns but is refused from a revenue or a market cap comparison.
  4. No median, quartile or percentile is published under five peers, and the n is printed beside every one.
  5. A blank cell always carries its reason. It is never a zero.
• Coverage line. "Covers N companies." with a note, counted live each time it is read (companies that carry annual SEC fundamentals and a peer group). On the public page a Cite this data block follows, naming SEC EDGAR XBRL company facts as the upstream and explaining how the peer group is formed (the GICS sub-industry where the company is GICS-classified, otherwise the data vendor's industry label).

• Type a ticker or company name and pick a result, or press Enter: the tool opens at /tools/relative-value/<TICKER>.
• Click an example chip (public) or an example row (signed in).
• Signed in, the title strip also carries the row of sibling links to the other cross-asset and macro tools.

The fastest route is the command bar: RV plus a ticker skips this page. For the valuation multiples only (P/E, P/S, P/B, EV multiples) against a peer set you choose, use EQRV.

Relative value for one company

Answers "how does this company compare with the rest of its industry on growth, margins, balance sheet, cash and valuation", using one full fiscal year per company from one annual SEC filing.

• Title. "<Company> (<TICKER>) relative value" with the RV code tag, and the line "Compared with the rest of its industry group on the figures from its last annual filing."
• Against N <industry> peers. A table with columns Metric, the ticker, Peer median, Percentile and n (how many peers carried a comparable value). Eight rows: Revenue FY, Revenue YoY, Gross margin, Operating margin, Net margin, Return on equity, P / E and 1 year return. The panel's top right names the subject's form type, its fiscal year end and its reporting currency. A percentile of 80th means the company's value is above 80% of its peers on that metric; it is a position in the peer spread, not a judgement.
• Blanks. "n/a" means no comparable figure; "n/m" means the base was too small for a meaningful percentage (for example growth from a near-zero year). Hover a blank for its reason. A blank median or percentile means fewer than five peers carried the value.
• Source line. The source, the fiscal period (or the range of fiscal periods across the peers), the price date, the rules (no trailing twelve months, no currency conversion) and a link to the subject's annual filing on SEC EDGAR (the accession number).
• Peer set, N companies. The peer tickers, then "Left out:" with each excluded peer and its reason.
• What could not be compared. Counts of refused cells with the reason for each.
• Coverage. If there is no comparable peer group the page says so ("Kresmion has no comparable peer group for <TICKER>..."): a peer table needs an industry classification, an annual filing with fiscal-year figures and at least five peers with the same.
• A sign-up panel ("Open the relative value desk with a free account") describes the signed-in desk.

• Title strip. "<TICKER> relative value", with a context line: peer count and bucket ("custom set" when you edited it), the fiscal year and form type, and the reporting currency ("never converted").
• KPI strip. Operating margin, Revenue YoY, P / E, 1 year return and Market cap, each with its percentile among peers ("Nth pctl of n") and a small meter. Hover for the peer median. When a growth figure is missing from the matrix it is computed from the two filed fiscal years and labelled with both years.
• <Group> against the peer set (the matrix). One column group at a time: Size, Growth, Profitability, Balance sheet, Cash, Per share, Valuation, Price performance or Positioning (31 metrics in all). Each row is a company; the subject is marked. The cell tint is the rank within the column (one hue, stronger = higher rank), never good or bad. Tags beside a peer's ticker flag a fiscal year that ends far from the subject's (hover for the days) or a different reporting currency. Hover a cell for its period, form type, accession and as-of date, or for the refusal reason.
• <Metric> through time. The selected metric across fiscal years for the subject against the peer band (median and quartiles), one annual filing per point.
• Distribution: <metric>. Every peer as a tick on one axis; the box is the interquartile range, the dashed line the median and the accent line the subject. The readout gives n, median and quartiles.
• Rank: <metric>. Every company ranked on the selected metric, with the median, the interquartile range, n and how many were blank in the panel bar.
• Filing: <TICKER>. For the selected row: company, industry, the annual filing (form and accession), fiscal year end, fiscal year end versus the subject in days, filed date, currency and last close.
• Peer set. "In the set (N)" and "Left out (N)" with the reason for each.

• Column group picker (matrix panel bar): choose a group, or one of four presets: Cheapest on P/E, Highest margin, Most levered (net debt to operating cash flow), Best cash conversion (buybacks plus dividends over operating cash flow). A preset sets the group and the sort together.
• Click a column header to sort on it; the sorted metric drives the History, Distribution and Rank panels.
• Click a row in the matrix, the rank list or the peer list, or a tick on the distribution, to show that company in the Filing panel.
• Add a peer (title strip search) adds a company to the set, up to 25; Remove on a peer row drops it; Reset in the Peer set panel returns to the default industry set. Every view recomputes.
• Change subject (title strip search) opens another company.
• Long lists page with the pager in the panel bar. The group, sort, metric, selected row and custom peers all live in the URL, so a copied link reopens the same view.

Read the n before a percentile: a 90th percentile of 6 peers says less than one of 40. A company whose fiscal year ends months from its peers' is still compared, but the tag tells you the years are not aligned. For the same peer logic on valuation multiples with your own peer list, use EQRV; the company's full statements are on its company page.

Equity relative valuation (EQRV)

Answers "is this company priced high or low against its peers on the usual multiples, and how do its growth and margins line up with that", using the last fiscal year from annual filings and the latest close. It has no forward estimates, no EV / EBITDA and no history of the multiples.

• Title strip. "Equity relative valuation (EQRV)", with a context line: the subject against N peers in its industry group, the span of fiscal years and the date of the closes. With no ticker in the URL the page opens on AAPL.
• Subject and peers. The subject search, an "Add a peer ticker" search and the current peer list as chips. The panel bar says "industry peers (N)" or "your peer set (N of 25)".
• KPI strip (valuation). One tile per valuation multiple: P / E, P / S, P / B, EV / revenue, EV / operating income, FCF yield and Dividend yield. Each shows the subject's value, how far it sits from the peer median ("vs peer median"), and "pctl N · n N". "under 5 peers" means no median is published. Hover a tile for the filing year, the close date and the median.
• Where does it sit? A sorted bar for every company in the set on one metric (pick it in the panel bar), the subject in its own colour, the dashed line the peer median and a box for the middle half. Hover a bar for the company and value; the readout otherwise gives the peer median, quartiles and n. A line below counts companies with no value.
• Two metrics across the set. A scatter of any two ratio metrics (x and y pickers; defaults are operating margin across and EV / revenue up). Hover a dot for both values.
• Peer table. One row per company with FY end and every metric, then a peer median row with the number of peers carrying each value. Tint is the distance from the column's middle, not a verdict. Hover a blank for its reason. A gap tag on FY end means that company's fiscal year ends far from the subject's.
• What is refused, and what is not carried. Peers left out (a fiscal year ending more than 200 days from the subject's, or no annual filing with figures), tickers not ingested, counts of blank cells with reasons, measures the tool does not carry, and the universe size.

• Subject search: pick a company to change the subject (keeps a custom peer list if you have one).
• Add a peer ticker: adds a company; the × on a chip removes it. Use industry group drops your custom list.
• Save peer set stores the current peers as a saved model named "EQRV <TICKER> peers"; a status line confirms "Saved to Saved models."
• Sort picker and High to low / Low to high on the peer table.
• The subject and custom peers live in the URL (?t= and ?peers=), so a link reopens the same set.

EQRV, the relative value desk and the screener read the same cells, so a P/E here matches the other two. Put a margin on one axis and a multiple on the other in the scatter to see whether the companies priced highest are also the most profitable.

Merger calculator (MRGC)

Works out the acquirer's pro forma earnings per share for a deal scenario you state, against its standalone EPS, from each company's last annual filing and latest close. It is a calculator of your assumptions, not a Kresmion view: nothing here says whether a deal is likely or desirable.

• KPI strip.
  • Standalone EPS: acquirer net income over diluted shares, last fiscal year.
  • Pro forma EPS: the same fiscal year with your deal applied.
  • EPS change: pro forma against standalone, in percent (+ is accretion, minus is dilution). "n/a" when the standalone EPS is not positive.
  • Purchase price: offer per share times target shares, with the cash and stock split.
  • New acquirer shares: shares issued for the stock leg, and the acquirer holders' share of the combined company after.
  • Breakeven synergies: the pre-tax annual synergies at which the EPS change is zero.
  • Offer P/E: offer per share over the target's diluted EPS.
  • Acquirer P/E at issue: issue price over the acquirer's standalone EPS.
• Your scenario note. States that the offer, the cash and stock split, the rate, the tax rate and the synergies are yours and everything else is read from a named filing or the latest close.
• Your scenario panel. Acquirer and Target searches (company names and the target's close shown beside them) and one control per assumption, each a number box plus a slider with a hint: Premium % (or Offer $/sh), Cash %, Rate on cash % (pre-tax financing rate), Tax rate %, Synergies $/yr (blank counts as 0) and Issue price $ (blank uses the acquirer's close).
• EPS change by cash share and offer price. One line per offer price (yours, and 0.90x, 0.95x, 1.05x, 1.10x of it) across cash shares of 0, 25, 50, 75 and 100%. A ring marks your scenario.
• The same grid, as numbers. The same table as a heat grid: rows are offer prices, columns cash shares, cells the EPS change in percent; your cash share is framed.
• Not computed / Inputs to fix. Appears when the inputs are not like with like (for example the two companies report in different currencies) or an input is out of range, with one reason per line.
• Every input and where it comes from. Each input with its Value, Source (filing, latest close, your assumption, derived, not used) and Filing or date (form and accession, or the close date). The foot lists what is not modelled (purchase accounting, fees, debt refinancing) and the size of the eligible universe (US dollar 10-K filers).

• Pick both tickers, choose Premium % or Offer $ for how you state the offer, set the cash share, rate and tax rate (all three are required), then Calculate.
• Moving a slider recomputes the figures in your browser at once; with Re-read as I adjust on (the default) the server re-checks the scenario 0.8 seconds after you stop.
• Save scenario stores it as a saved model; the status reads "Saved to Saved models."
• Every input lives in the URL, so a copied link reopens the same scenario.

Read Breakeven synergies first: it tells you how much cost saving the deal needs before EPS stops falling. For deals that have actually been filed, start from the Merger arbitrage board and enter its terms here.

Merger arbitrage board

Lists pending acquisitions of listed companies read from SEC filings, with the terms each filing states, the spread between the deal value and the target's price, and the prices prediction and options markets put on completion, each with its source. A deal appears once its 8-K, merger proxy, S-4 or tender offer is filed.

• Summary strip. On the board (and pending in total), Cash price stated, Exchange ratio stated (stock or mixed), Spread computed, Polymarket match (a live market naming the target) and Options-implied (target in the options universe). On the public page, while no deal on the board has a Polymarket match, the Polymarket match tile, the Polymarket filter choice and the Polymarket and Gap columns are left out; the same goes for Options-implied (tile, Options choice, Options-implied and Gap columns) while no target is in the options universe. Each comes back with the first match, and the page's dated summary still gives both counts.
• Colour. Each deal carries a swatch for its consideration: blue for cash, violet for stock or mixed, amber for election, aggregate or not parsed.
• Spread against time to the stated close. One dot per deal: across, days to the last day of the close period the filing names; up, the spread. Click a dot to open that deal.
• The prices behind each spread, with a free account (public page). An account box right under the first chart: "A free account adds the target's and the acquirer's last prices behind each spread, with their dates, and the Kalshi price wherever a Kalshi market names the deal, to this board.", with Create a free account and Sign in.
• When do the stated close periods end? Pending deals counted by the month their stated close period ends, stacked by consideration; signed in, past months are faded and a note counts deals with no calendar end ("early 2027", fiscal periods).
• Deal timeline (and, signed in, Deal figures). For the selected deal: announcement, agreement, stated close (last day) and outside date on a line, then Spread, Annualised, Premium, Deal value per share, Polymarket YES, Options-implied, Termination fee (payable by the target), Reverse fee (payable by the acquirer), Filings read and Stated close.
• Pending deals table. Target (ticker and name), Acquirer, Terms (filing) such as "$X cash + N ACQ sh + CVR (not valued)", Announced, Stated close, Spread, Annualised, Premium, Polymarket, Options-implied (Options), Gap, Fees (T = termination fee, R = reverse fee) and the source filing.
  • Spread = deal value per share over the target's last price, minus one. A last price older than the stated number of days is not used.
  • Annualised = spread times 365 over the days to the stated close (simple). Periods with no calendar end are not annualised.
  • Premium = deal value at announcement over the target's close on the last session before the announcement, minus one.
  • Polymarket = the YES price of a live market whose question names the target and an M&A word, with the question beneath.
  • Options-implied = the risk-neutral probability, from options chain snapshots, that the target trades at or above 98% of the deal value at the first expiry on or after the stated close. Risk-neutral is not a real-world probability.
  • Gap = Polymarket minus options-implied, in percentage points.
  • "n/a", "not parsed" or "not stated" means the filing does not state the term or the value could not be computed; hover for the reason. Long notes under a value wrap within their cell, so the whole table fits the panel.
• Closed or withdrawn in the last 60 days. Target, Acquirer, Status (completed or withdrawn, with the form that settled it, such as 8-K or Form 25) and Date.
• Kalshi and last prices (signed in). In the board's order: Target last, Acquirer last, Deal value, Polymarket, Kalshi (YES price and market ticker), Options-implied, Poly minus Kalshi and Kalshi minus options in points. Signed out this panel says these figures are shown to signed-in readers, with a Sign in link.
• Method. The rules above in full, including that the parser is deterministic, contingent value rights and holder-election deals are not valued, partial tenders, SPAC combinations and private targets are excluded, and a deal leaves the board on completion (8-K Item 2.01), termination (Item 1.02) or 400 days with no new filing. No spread history is stored, so no spread-over-time line is drawn.

• Jump to a target or acquirer: a ticker search that filters the board to deals naming that company; the ticker button beside it clears the filter.
• Consideration filter (All, Cash, Stock or mixed, Other) and Market price filter (Any market, Polymarket, Options, No market). On the public page a choice with no match on the board is left out, and the whole Market price filter is hidden when neither market has one.
• Filter by name or ticker box on the table. Public: click a column header to sort (default: announcement date, newest first). Signed in: the sort pills Newest, Spread, Annualised, Premium, Close.
• Click a row or a dot to open its timeline and figures.

Sort by Annualised to compare deals with different close dates on one scale, then check the Stated close column: a deal with no calendar end has no annualised figure at all. A wide Gap means the two markets price completion differently; it is a difference between two prices, not a Kresmion estimate.

Buybacks hub

The starting point for a company's share repurchases: what it paid to buy back its stock, how many shares that removed, and which listed companies connected to it have filed a position in its shares.

• Open a company. A search box ("Search a company or ticker") and six worked examples with what each page shows: AAPL (large cash repurchases each year), JPM (a program authorization in its newest 10-K), LOW (accelerated share repurchase settlements and remaining capacity), XOM (treasury shares acquired each quarter), SYM (its largest customer also holds a Schedule 13D) and VNOM (a customer whose 13D percent rose between filings).
• What each page shows (public) or What a company page shows (signed in): cash paid per fiscal year and quarter, the cover-page share count over time, a fiscal-year table with the share-count cross-check, program authorizations and accelerated repurchases, and a graph of connected listed companies that filed a Schedule 13D or 13G.
• How to read it (signed in): every figure is a value an issuer tagged in an SEC filing or a filed 13D/13G; derived figures say derived; filed positions and relationships only, nothing indicates that any company intends a transaction.

Search a company and pick it, or click an example; both open /tools/buybacks/<TICKER>.

The company page is also the Ownership view of the company workspace's buyback block; see Ownership view.

Buybacks for one company

Shows one company's repurchases from SEC XBRL as filed, whether its share count actually fell, its tagged buyback programmes, and the listed companies linked to it that filed a Schedule 13D or 13G on its stock. There is no deal history and no predicted acquirer.

• Summary (KPIs). Cash paid in the latest fiscal year; the last four quarters (derived from year-to-date figures when the filer tags only those, and labelled so); share count change over 1 year and over 3 years (cover-page shares outstanding); remaining authorization as tagged. Hover a figure for its dates and form.
• How much did it spend buying back stock? / Buyback spend. Bars per fiscal year or per quarter. Public: toggles Fiscal years / Quarters and Cash paid / Shares bought (shares are tagged per fiscal year only). Signed in: one picker with FY cash, FY shares and Qtr cash. Faded bars are derived quarters. Hover a bar for its value.
• Did the share count fall? / Share count. Cover-page shares outstanding over time.
• Fiscal years, with the share-count cross-check / By fiscal year. FY end (linked to the filing), Cash paid, Shares bought, Avg price* (derived: the equity-statement value over shares repurchased), Net change (cover count on the annual report against the previous one; "basis changed" when the share basis changed), Other* (derived: net change plus shares repurchased, which captures issuance, awards, conversions and timing) and the filing. A starred column is derived.
• Program authorizations and Accelerated share repurchases (signed in: one panel with Auth N / ASR N tabs). Each authorization with its amount and period, and each ASR fact as tagged; "Not tagged." when the filer does not tag them.
• Who else holds a filed position / How the filers connect. A graph centred on the company: teal lines are customer or supplier disclosures, pink lines are a Schedule 13D or 13G on the issuer (thicker for 13D), amber nodes are companies that are both connected and holding. The note counts companies drawn and what was left out (passive institutions with no other link, filers not matched to a listed company, ambiguous names).
• Other filers (signed in). The graph's companies as a list: Company, Link (how it is connected), Latest % (percent of class in the latest 13D/13G) and Since first (percent of class up, down or unchanged since the first filing, with the change in points).
• Method and limits. What each figure is, what is derived, coverage and source.

Use the search in the header ("Open another company") to switch company. Click a node in the graph (or a row in Other filers) to open that company's buybacks page. Pagers in the panel bars page long tables.

Compare Net change with Shares bought: a company can buy back many shares and still see its count rise if it issues more through awards or conversions; the Other column shows how much. The graph shows filed relationships only and is not a list of buyers.

Insiders and Congress

Screens over SEC Form 4 insider filings and congressional trade disclosures, under Tools › Insiders & Congress.

Insider clusters

Lists every cluster in the daily snapshot: three or more distinct insiders filing the same direction (open-market buys or sells) on one ticker within 30 days, from SEC Form 4. One executive buying is one decision; several insiders acting together in the same window is what this page isolates.

• At this snapshot (public) or the KPI strip (signed in): new buy and sell clusters today, the top buy and sell clusters, counts and dollar value of buy and sell clusters, and (signed in) how many buys include a senior officer. A NEW chip marks a cluster appearing in a snapshot for the first time.
• Cluster table (public "All clusters", signed in "All clusters"). Ticker and company, Dir (buy or sell), Insiders, Total $ (Value), % mcap (cluster value as a share of market cap), % stake (value-weighted share of the prior stake sold, or added on a buy), Senior (CEO, CFO, President, Chair or Chairman among the filers), Strength (a score comparable only within one direction), Since (price move from the window start to the latest close), Age and Last trade. A 10% owner chip means 10% owners contributed half or more of the value: often a fund rather than an executive. "no market cap" and "no holdings data" mean the figure is not on file.
• Cluster detail drawer (open a row). Window, insiders and trades, total value, % of market cap, % of stake, 10% owner share, price since window start and since last trade, strength and age; the senior roles; a price chart over the window; every trade (Insider, Date, Code P or S, Shares, Price, Value, Held after, 10b5-1 plan flag where the filing carries it, Filing link); and Prior episodes on this ticker and direction with "What happened next" (the 30-day return after each, or "outcome not tracked yet").
• Track record. For buy and sell clusters, buckets (all, senior role involved or not, by size: buys $1M and above or under, sells $10M and above or under) with n, Matured (episodes old enough for a 30-day return), Med +7d, Med +30d, Hit (share that rose over 30 days for buys, fell for sells) and Excess (median 30-day return minus the SPY median over the same windows). Below, the most recent episodes with +7d and +30d returns. Every episode is counted, losers included; entry is the first close after the cluster first appeared. These are historical outcomes, not a forecast.
• By sector (public) or Buy against sell, by sector and What the snapshot is made of (signed in): cluster counts and value per sector, and the share of dollar value buy against sell or buys by sector.
• Method and sources. The rule, the strength score version, the filing lag (Form 4 is due within two business days of the trade), and that Section 16 disclosure does not imply impropriety.

• Direction (Buy, Sell, All), Min insiders, Min $ (any, $100K, $1M, $10M), Sector, toggles Senior only, No 10% owner and NEW only, and a ticker or company search. Reset clears them. Filters live in the URL.
• Public: click a column header to sort. Signed in: a sort picker (Strength, Value, Insiders, % mcap, Price since, Last trade) and an order button.
• Signed in, click a sector bar to filter the table, or a bar in New buy clusters, Strongest buy clusters or Strongest sell clusters to open it.
• Click any row (or a hero entry) to open the detail drawer; close it with its close button.

Turn on No 10% owner and Senior only together to keep the clusters driven by officers. Strength compares clusters of the same direction; do not rank a buy against a sell on it. For every Form 4 trade, not only clusters, use the insider trades feed at /intel/insider-trades.

Insider clusters for one company

Answers "have insiders at this company traded together, when, and what followed".

• Title. "<TICKER> insider buying and selling clusters", with a link to the company's Form 4 filings on EDGAR.
• Latest cluster per direction. Direction, Status, Trade dates, Insiders, Trades, Filed value (shares times the Form 4 price, summed), Senior titles among the filers, First detected, and Latest filing read by.
• The trades behind these clusters. Insider, Direction, Code (P purchase, S sale), Trade date, On EDGAR by, Shares, Price, Value and a Form 4 link.
• Full <TICKER> history after sign-in. Explains what an account adds, with Create an account and Sign in.
• Method and limitations. How filed value and senior titles are matched, with links to the full methodology and to all clusters.

• KPI strip. Episodes on file (buy and sell), Latest episode (first seen), Latest value, 30d after latest (return over the 30 days after the latest episode's entry close, or "not tracked yet"), Open-market trades in the last 24 months (codes P and S).
• Cluster episodes. First seen, Dir, Window, Ins (distinct insiders), Total $, Str (strength score), Senior, and 30d after.
• Open-market trades. Insider and role (with "10% owner" where flagged), Date, Code, Shares, Price, Value, Held after, % stake (trade as a share of the insider's holding), 10b5-1 and a Filing link.
• How an episode is counted. The rule and its limits.

Tables page with the pager in the panel bar. Hover a column header for its definition.

A 10b5-1 flag means the trade was made under a pre-arranged plan; filings before April 2023 do not carry the flag.

Congressional trade anomalies

Flags two kinds of unusual trades in US House STOCK Act disclosures: trades three or more times a member's typical size, and clusters of three or more members trading the same ticker in the same direction within 30 days. Senate filings are not ingested. Disclosure does not imply impropriety.

• Sizes are estimates. Disclosures give a dollar bracket; each is turned into its midpoint (for example "$1,001 - $15,000" becomes $8,000).
• KPI strip (signed in). Outlier trades (last 60 days), Largest multiple, Members flagged, Multi-member clusters.
• Multi-member clusters (public) / Clusters by estimated total and Cluster detail (signed in). Ticker, Direction, Members (distinct members in the 30-day window), Trades, Est. $ (estimated total) and Window. When none qualify, both pages say no 3 or more members fall within one 30-day span among trades dated in the last 75 days (the longer look-back leaves room for the 45-day disclosure delay).
• Single-member outliers (last 60 days). Member, Chamber, Ticker, Type (purchase or sale), Date, Est. $ (Est. size), Typical $ (the member's median trade over their disclosed history; a member needs at least 12 trades to have one), and × (Multiple: estimated size over typical size). Signed in also adds On file (the member's disclosed trades).
• How far above their typical size (signed in). Outlier trades as bars ranked by multiple, coloured by purchase or sale.
• Who and which way (signed in). Two rings: outlier trades by member and by type.
• Method. The thresholds above, no party segmentation, and where the full disclosures live (the Clerk of the House site).

• Signed in: sort outliers with Multiple / Size / Date / Member; filter with the "Member or ticker" search. Click a bar in the multiples chart to filter the outlier table to that member (click again to clear). Click a cluster bar to highlight it in Cluster detail.
• Public: read only; the tables are in ranked order.

A large multiple from a member who usually trades small amounts can still be a modest dollar figure: read Est. $ beside ×. Because sizes are bracket midpoints, two trades in the same bracket show the same estimate.

Supply chains

The semiconductor tracker, under Tools › Supply Chains.

Semiconductor supply chain

Following how much moved between two countries at each production stage of the semiconductor chain last month (customs values), who holds each chokepoint, where the named fabs and plants are, what policy and physical events are dated near the chain, and which listed companies sit on it. A corridor number is a declared customs value, never a price. Where a source has not delivered, the page says so in words rather than showing 0.

• Title strip. "Semiconductor supply chain", with a context line naming the chain and "as of time UTC" ("syncing" while it refreshes). On the right: the Chain picker (the chip archetype), the five view buttons (Overview, Corridors, Chokepoints, Facilities, Exposure) and About the data.
• KPI strip.
  • Next dated event: a countdown to the next dated policy or release event, with its date; hover for its title. "nothing dated ahead" when none is on file.
  • Top mover: the corridor whose latest month is most unusual against its own 36-month history, with its year-on-year change; hover for the label, y/y and z-score.
  • Live corridors: corridors printing numbers out of every corridor declared for the chain ("n / m").
  • Policy actions and Physical events: how many dated items are on file.
  • Facilities: seeded sites, with how many are in production.
• Overview (one screen, four panels):
  • Movers (tab Movers): corridors ranked by the size of the 36-month z-score, then the size of the y/y change. Direction is never part of the rank; colour carries the sign. Columns: Corridor, Value (monthly customs value in USD, latest month), y/y, z (standard deviations from the trailing 36-month mean) and a 24m sparkline. Tab Early (when available): Early reads, official flashes that land before the customs month closes, with Line, Value, y/y and Per day (working-day adjusted y/y where the source printed one), all as the source printed them.
  • Chain (tab Chain): one row per production stage with Stage, Leading holders (up to three countries with their share; hover for the basis and date), Tickers (coloured by today's change), Hop to next (the customs corridor from this stage to the next; "no customs flow" or "end of chain" otherwise) and its y/y. Tab Exposure: tickers ranked by how many stages they are named at (Ticker, Stages, Where, 1D live change).
  • Selected corridor: the picked corridor's monthly series since 2015 (dashed line: the mirror figure the partner country reports; ticks: dated policy events), with Latest, m/m ("masked Jan-Mar": the Lunar New Year moves those months between years), y/y, 3m-avg y/y (trailing three months against the same three a year earlier), z, 36m, Rank (this print among all observations of the same calendar month), HS codes and Next update. A Value / Qty switch appears when quantities exist. Not seasonally adjusted. "Not shown: this corridor's figures are not verified for display" means the source's reuse terms are unconfirmed.
  • Risk: tabs Policy (dated export-control and licensing actions: Date, Jur jurisdiction, Action), Physical (earthquakes and other events near tracked sites: Date, Event, Nearest site) and Ind. (reservoir storage and the chain's economic series: Series, Last, y/y, Period, Trend). Each list pages.
• Corridors. Every corridor grouped by stage, with Corridor, HS, Latest, Period, m/m, y/y, 3m y/y, z36, Share (this corridor as a percentage of the reporter's own total for the same codes and flow, blank where that total is not published), Rank and Next. Beside it, Corridor detail for the picked row: the series and statistics above plus the mirror comparison, weight and unit value ("not a price"), a price-versus-volume decomposition, restatements, the policy actions touching the corridor, the sites at its two ends, "What prints like this meant" (how similar past prints were followed) and the same month in prior years.
• Chokepoints. Who holds each step: Chokepoint, Holder, Share (the source's own share) and As of. Every row carries a source and date.
• Facilities. Tab Table: Facility, Owner, Ticker, Country, Stage / kind, Node, Wafer, capacity (wafer starts per month, only where a source publishes it), Status and As of, with Stage, Country and Status filter chips. Tab Map: every seeded site as a pin on a dark world map, sized by capacity where one is sourced, coloured by status: production (sky blue), ramp (violet), construction (amber), announced (light grey), idled (slate), closed (dark slate). Hover a pin for owner, stage, node, wafer size, capacity and dated status. The map sidebar counts the drawn sites by country (ISO code), one stacked bar per country split by status, largest first; the header reads "n sites in m countries". There are no flow arcs: corridors are country totals, not routes.
• Exposure. Tickers routed through the chain (or one stage): Ticker, Stages, Seeded sites, Corridors (the hops at those stages), Short interest (FINRA, newest settlement), 13F filers (tracked filers, newest quarter), Insider (Form 4 clusters), Congress 90d (House transaction reports, trailing 90 days by trade date), Dealer gamma (modelled from the listed options chain, not disclosed positioning) and Top customer (as disclosed, percent of revenue).
• About the data. Data status (what each feed delivered and when the next print is due), Coverage (what is tracked and what is not: no shipment-level tracking, no prices, nothing behind a subscription, corridors marked "not shown"), and the Attribution lines each source requires (US Census Bureau, Eurostat, Taiwan customs and export orders, Korea customs, Japan e-Stat, Taiwan Water Resources Agency, USGS).

• Chain picker: switches the archetype; the picked corridor and stage clear with it.
• View buttons and About the data (press again to leave it). The view, chain, corridor and stage are kept in the address (?tab=, ?chip=, ?corridor=, ?stage=), so a link reopens the same state; switching does not add Back-button steps.
• A row in Movers or a Chain hop opens that corridor in Selected corridor. Corridor table headers y/y and z36 sort.
• Facilities Map: toggle each status chip (with its count) to hide or show those pins; type in Site, owner, ticker to filter as you type; click a country in the sidebar to fit the map to its sites, World to return.
• Exposure: All stages or one stage chip; a filter box ("Filter by ticker, stage or customer"); click a header to sort.

Start on Overview: Top mover and Movers tell you which corridor printed out of line, the Chain panel shows where it sits, Risk shows what is dated around it. For the disclosed supplier and customer graph of one company, use Supply chain explorer instead.

Working tools

Tools you use to size, ask, generate and model, rather than read: under Tools › Trading and Data › Models & Data.

Position size calculator

Working out how many units, shares, lots or contracts a trade can be for a given account size, risk per trade and stop distance, with the notional, margin, leverage, break-even and profit or loss at up to five targets. It is sizing arithmetic on the numbers you enter: long and short are your inputs, and nothing on the page says whether to take a trade.

• Instrument. The symbol you picked, a kind badge (crypto, stock, index, fx spot, future), its name, the last price with its change, and the time of that price in UTC. The price is the latest daily bar from the Kresmion price feed; while the order type is Market it is re-read every 15 seconds (a small dot pulses on each read). The signed-in layout prints the stamp as "daily bar date" when the bar is a daily close. For a futures contract a spec line follows ("exchange name: contract unit, quoted in unit. Tick size = $value, 1.00 = $value"); for an FX pair, the pip size and "1 lot = 100,000 base". "No price for symbol" means the feed has no quote: try another asset class tab or use a Limit entry.
• Result strip (KPIs).
  • Position size: units in the instrument's own measure (coins, shares, index points, FX base units, or contracts when a futures spec is on file). For futures the sub-line gives the whole contracts and the risk at that whole number; for FX the lot count.
  • Notional: size times entry, in US dollars, with "% of account".
  • Risk at stop: the dollars you lose if the stop is hit (account × risk per trade), with the percentage and account it came from.
  • R:R (public) or Reward to risk (signed in): distance to the first target divided by distance to the stop, written 1:n; the sub-line gives the final target's ratio when there are several.
  • Leverage / Leverage used: notional divided by account. It turns amber above 5x.
  • Margin: notional divided by the leverage you set. For futures it reads "set by the exchange": exchange margin is not modelled.
• Check note. A warning line appears when the inputs cannot produce a size (for example a stop on the wrong side of the entry).
• Levels table. Rows: Entry, Break-even (only when a fee is set: the price that covers both sides' commission), Stop (red) and TP 1 to TP 5 (green). Columns: Level, Price, Distance % (from the entry), P&L (loss at the stop, gross profit at each target; hover a target's P&L for the figure net of fees) and R:R. The stop row's last column shows the stop in pips (FX) or ticks (futures). A price on the wrong side of the entry is marked.
• Levels on the price axis (signed in only). A vertical price axis with the stop, entry, break-even and targets, the loss zone (entry to stop) tinted red and the gain zone (entry to last target) tinted green, with the P&L at each level.
• Contract details (FX and futures only). FX: standard, mini and micro lots, pip value per lot, stop in pips. Futures: contracts, whole contracts, risk at the whole number, tick value, stop in ticks. "Contract spec: not on file, sizing in units" means no exchange spec is stored for that contract.
• Exposure. Notional as a share of the account, with a bar that fills at 500% (a mark at 100%). Grey up to 100% (unleveraged), amber to 500%, red beyond.
• How the size is worked out (signed in) / the method note under the title (public): size = (account × risk per trade) ÷ loss per unit between entry and stop; reward to risk = distance to a target ÷ distance to the stop; break-even adds both sides' commission.

• Pick an instrument: the asset-class tabs (All, Stocks, Crypto, FX, Futures, Index) and a search box ("Search stocks, crypto, FX, futures"). With an empty box the list shows "Popular" names for the tab. Arrow keys move through results, Enter picks. The calculator opens on BTC.
• Direction: Long or Short. Order: Market (entry follows the latest price, labelled "Entry · latest price") or Limit (type your entry).
• Account (USD) and Risk per trade (%), or one of the Quick risk chips: 0.25%, 0.5%, 1%, 1.5%, 2%.
• Fees and leverage (a fold: "Edit" / "Hide" when signed in): Fee per side in dollars for the whole position, and Leverage (1x to 500x, default 30x). The fold's title shows the current values.
• Stop and targets: type either the price or the distance %. The one you typed is locked (bright, tooltip "Set by you"); the other follows the entry ("Moves with entry"). Defaults: stop 2.5%, targets 2.5%, 5% and 7.5%. + Add target adds rows up to five; × removes one. On a narrow screen signed in, an Edit by switch (Price / Distance %) picks which column you edit.
• Save stores every input in this browser; the calculator reopens with them next time. Copy summary copies a plain-text summary (symbol, direction, account, risk, entry, size, notional, stop and each target with P&L and R:R) to the clipboard. Reset levels puts the stop, targets and fees back to the defaults.

Set the stop by price when you have a level in mind and by percent when you think in distance; the lock keeps your choice when the live price moves. For futures, read the whole-contract line: the risk at a whole number of contracts can differ from your target risk. The same calculator also sits in the terminal canvas as a widget.

Ask the Data

A guided query builder over four Kresmion datasets: pick a dataset, add filters from dropdowns, sort and run. It is not a chat box and accepts no free text questions or SQL.

Only these four datasets, with these fields:

It cannot join datasets, group, sum or average, compute anything new, read prices, filings text or any other dataset, or return more than 100 rows a run. Text filters offer is (exact, any case) and contains; numbers offer is, ≥ and ≤; dates offer ≥ and ≤ (type YYYY-MM-DD). Up to 6 filters, always combined with AND. The convergence dataset only holds events that are still active. The 13F dataset spans every stored quarter, so filter on Quarter to read one.

• KPI strip. Before a run: Rows returned ("not run yet"), the selected Dataset, and Datasets (how many, "up to 100 rows a run"). After a run: Rows returned (the rows this run returned: "every matching row" under the limit, or "hit the n-row limit, more rows may match"), Dataset, Filters ("every row" or "combined with AND") and Sorted by with the direction. They describe the run you made, not unsent edits.
• Result. One column per field, values as stored. Tickers are chips; dates show as stored, timestamps as day and time in UTC. "--" is an empty value. When a run is capped a line says "Capped at the n-row limit; raise the limit or narrow the filters."

• Dataset: changing it clears filters, sort and the result.
• + Add filter, then pick the field, the operator and the value; changing the field resets the operator. × removes a filter. Filters with an empty value are ignored.
• Sort by and Descending / Ascending; Row limit (1 to 100, default 50).
• Run query, or press Enter in a value box. The result pages in its bar.
• Opening /tools/ask?ticker=NVDA (what the entity menu does) picks the first dataset with a ticker field and adds a Ticker filter for you.

To list one politician's trades, pick Congressional trades, filter Politician contains a surname, sort by Trade date. For the largest 13F positions in one name, filter Ticker is the ticker and Quarter is the quarter label as stored, sorted by Market value. The full congress, contracts and 13F desks are Institutional holdings and the intelligence pages.

Report Generator

Generating a PDF report, with charts, from Kresmion's intelligence and macro data, downloading it, and having the Portfolio Risk report emailed to you on a schedule.

• Title strip. "Report generator", context "Server-rendered PDF reports ... · 5 reports per user per 24 hours".
• KPI strip. Last 24 hours ("n / 5": reports you requested in the rolling 24 hours the limit counts), Last report (its name, status, page count and time) and Email schedules ("n / 5" active).
• Pick a report. Seven tiles, each with its description (every report is available during the free beta; no plan tag is shown while billing is off):
  • Portfolio Risk Report: VaR, drawdown profile, correlation matrix, event log and macro context for a basket.
  • Asset Deep Dive: one ticker's price action, events, institutional positioning, insider trades and OSINT.
  • Macro Regime Briefing: the current regime, its score trajectory, the factor breakdown and the top COT extremes.
  • COT Positioning Report: every tracked CFTC futures market on the latest report week, sorted by 52-week z-score, with a trajectory chart for the top eight.
  • Capital Flows Digest: the quarter's largest 13F increases, reductions, exits and new positions, ranked by dollar change.
  • Watchlist Morning Brief: one summary per watchlist asset with price moves, recent alerts, related prediction markets and the regime.
  • Prediction Market Microstructure: cross-venue divergence, biggest repricings, flow flags, crowd calibration and wallet flows (crowd estimates, not advice).
• Configure. The picked report's settings (below), then status: "Generating your report..." (usually 15 to 30 seconds), "Report ready, n pages." or "Generation failed" with the reason.
• Recent reports. Your last 20 requests: Report (a code chip such as RISK, DIVE, MACRO, COT, 13F, WATCH, PM, and the name), Status, Pages, Requested (UTC) and File: Download while we hold the PDF, "File removed" for a report made before October 2026 (reports were then deleted after their first download), "--" for one still generating or failed.
• Email schedules. Your Portfolio Risk report emailed as a PDF attachment: Schedule (for example "weekly (Mon) at 07:00 UTC"), State (active or paused), Last sent ("never sent yet") and Delete.

• Click a tile (Generate →); it changes to "Configuring below". Close on the configure panel cancels.
• Settings per report:
  • Portfolio Risk: Asset selection (My vault holdings, weighted by USD value with dust excluded; Use my watchlist, equal weight; Custom basket, search and add up to 20 assets, each with an optional relative weight), Lookback period (30D, 60D, 90D, 180D), and an optional AI narrative (three paragraphs written by a language model, about 10 seconds more).
  • Asset Deep Dive: Symbol (search by ticker or name, arrows and Enter to pick, clear to reset), Lookback period, optional AI narrative.
  • Macro Regime Briefing: Score trajectory lookback (days) 30 to 365 (default 90), optional analyst paragraph.
  • COT Positioning: Top-8 trajectory lookback (weeks) 8 to 104 (default 26), optional analyst paragraph.
  • Capital Flows Digest: Quarter (Latest filed, Previous, Two back).
  • Watchlist Morning Brief and Prediction Market Microstructure: no settings.
• Generate report →, then Download PDF (saved as kresmion_<report>_<date>.pdf) or Generate another. On failure, Try again. We keep every report you generate: Download PDF works more than once, and Download on its Recent reports row saves it again at any later visit. Only you can download your reports. A failed download says so at the top of Recent reports.
• The limit is 5 reports in 24 hours per user. Schedules: Every Daily, Weekly (pick the day) or Monthly (day 1 to 28) at an hour in UTC, then Add schedule (up to 5 active).

Put your basket in the watchlist first and the Portfolio Risk and Morning Brief reports pick it up with no typing. For an interactive version of VaR and correlation, use Risk models.

Risk models

Running three pre-built models (portfolio VaR, signal momentum, a correlation matrix) on live data, building your own ranking or alert model, and downloading Excel templates that call the Kresmion Excel add-in.

• Every model is one panel: its name, the question it answers, the inputs in one row, How it works (method, when to use it, limitations), and the result with the time it ran.
• Portfolio VaR. "How much could this portfolio lose on a bad day?" The result is the 1-day VaR in dollars at your confidence ("At 95% confidence, this portfolio is not expected to lose more than $x in a single trading day under normal market conditions"), from historical simulation over your lookback, then Largest contributors: Asset, Share of risk bar, Loss share (dollars) and % of risk. It is a normal-market estimate, not a stress test.
• Signal momentum. What prices did after each Kresmion event type fired over the last 90 days (fixed window). Columns: Signal, Direction (the type's direction tag, Up or Down), 1D, 5D, 30D (the share of events after which price moved that way over that horizon) and N 5D (matured 5-day samples; "(low)" and greyed under 10). Rows sort by the 5D rate into Highest 5D win rate (top six) and Lowest 5D win rate (the three weakest). Types with no 5D reading are left out. A past win rate is a record, not a forecast.
• Watchlist correlation matrix. Pearson correlation of daily log returns over your lookback, for 2 to 20 tickers. The lead line gives the average cross-asset correlation (mean of absolute off-diagonal values) labelled Low (under 0.4), Elevated (0.4 to 0.7) or "High, diversification breaking down" (0.7 and above), and the most correlated pair. The Matrix colours each cell: 0.95 and above "move as one", 0.6 to 0.95 high, 0.2 to 0.6 some, -0.2 to 0.2 independent, below -0.6 "hedge each other". Pairs lists every pair most correlated first, with a strength bar.
• Build your own and Your custom models. Your saved custom models: Model, Setup, Last run. A result is a table of assets: # (or Alert for alert models), Asset, and Sharpe, Return (annualised), Vol (annualised) or Signals depending on the measure.
• Excel templates. Portfolio Risk Template, Signal Tracker Template and Correlation Workbook, each with what it does and the add-in functions it uses (for example KRESMION.PRICE(ticker, field), KRESMION.SIGNAL(ticker, field, days), KRESMION.REGIME(field)). The Correlation Workbook works without the add-in.

• Portfolio VaR: one row per position (ticker and weight; + Position, × to remove). The weights must sum to 1.00 ("Sum of weights = n (ok)"). Confidence 0.90, 0.95 or 0.99; Lookback (days) 30 to 2,520; Portfolio value ($). It opens with BTC-USD 0.5, SPY 0.3, GLD 0.2. Save as model stores this setup in Saved models ("Saved to Saved models.").
• Correlation: tickers comma-separated (opens with BTC-USD, ETH-USD, SPY, GLD, TLT) and the lookback. Matrix / Pairs switches the view when the panel is wide enough; the pairs list pages.
• Signal momentum: just Run model.
• Build your own, four steps: 1 · Name, 2 · Universe (My watchlist or a custom comma-separated list), 3 · Measure (Momentum, Sharpe ratio, Volatility, Max drawdown, Signal frequency) over the last 14, 30, 60, 90 or 180 days, 4 · Output (Rank assets, or Alert me when the measure drops below / rises above a threshold). There is no schedule: a model runs when you click. Save & run saves and runs; in the list, Run reruns and × deletes (after a confirmation). Custom models live in this list, not on Saved models.
• Run model on any panel; edit an input and run again in place.

The add-in that the templates call installs from API keys settings.

Saved models

The list of model configurations you saved from other tools (a Portfolio VaR setup from Risk models, a peer set from Equity relative valuation, a deal scenario from the Merger calculator), to rerun, open, rename or delete. Up to 20 per user.

• KPI strip. Saved ("n / 20"), Last run (the most recently run model and when) and Types (distinct model types).
• Your saved models. Model (name), Type (for example Portfolio VaR, EQRV peer set, Merger scenario), Last run and Updated (UTC; "never" if not yet), then the row actions.
• Run result. "Result: name", with the run time. Fields lists every value the model returned as a field path and value (for example contribution_by_asset[0].ticker), paged; JSON prints the same payload whole. A result is kept for this visit only: "Not run in this session" until you click Run.

• Click a row to show its result; Run computes it (and picks the row).
• Open (peer sets and deal scenarios only) opens the model in its own tool with its inputs filled in, where the result is laid out properly.
• Rename opens a line above the table: type, then Enter or Save; Escape or Cancel leaves it.
• Delete asks for confirmation; it cannot be undone.

Custom models built under Build your own on Risk models are listed there, not here.

Where to find it

• See which asset class led over the last month → Tools › Cross-Asset & Macro › Market Rotation, window 1M (Market rotation)
• See which coins, currencies or commodities lead their class → Market rotation, the class tab, Bars or Quadrant (Market rotation)
• Check whether defensive or cyclical sectors lead → Sector rotation, Leadership line (signed in: Cyclical minus defensive KPI)
• Find the most unusual one-day moves for each asset's own volatility → Relative strength, sort by Rank (1D z)
• Chart the ratio of two assets → Relative strength, open a row, Compare with... (Relative strength)
• Find asset pairs whose correlation broke down → Correlation regime
• See the strongest and weakest major currency today → Currency strength
• See how far an asset sits below its all-time high → Drawdown tracker, Jump to
• Browse every free public tool → Tools hub, /tools
• See whether the US yield curve is inverted → the badge on the public yield curve page, or the US 2s10s KPI signed in (Yield curves)
• Compare today's Treasury curve with the curve on any past day → US curve panel, Now vs then, date box (signed in) (Yield curves)
• List every past 2y/10y inversion and how long it lasted → Inversions panel (Yield curves)
• Rank government 10-year yields or their spread over the Bund → Sovereign 10Y panel (Yield curves)
• Set an alert on the 2s10s spread → the bell beside the inversion badge (public) or in the title strip (signed in) (Yield curves)
• See how SPY or Treasuries usually moved after CPI releases or Fed holds → headline matrix (Macro event impact)
• Inspect every past Fed decision behind a statistic, with its 2-year yield surprise → Every instance table (signed in) (Macro event impact)
• Find past days whose macro regime looked like today's → Historical regime analogs
• Check how the latest 10-year note auction ranked on bid-to-cover → auction table (Treasury auctions)
• See how much Treasury debt matures within a year → Debt maturing within 12 months KPI and the maturity ladder (Treasury auctions)
• Compare a company with its industry peers → Relative value (RV <TICKER>)
• See where a stock's P/E sits among its peers → EQRV (EQRV <TICKER>)
• Work out whether a deal adds to or dilutes EPS → Merger calculator (MRGC)
• Find pending takeovers and their spreads → Merger arbitrage board (MARB)
• See the market-implied odds a deal completes → Merger arbitrage board, Polymarket and Options-implied columns
• See how much a company spent on buybacks → Buybacks for one company (BUYP <TICKER>)
• Check whether a company's share count actually fell → Buybacks for one company, Net change column
• Find which listed companies filed a 13D or 13G on a stock → Buybacks for one company, graph
• Find stocks where several insiders bought at once → Insider clusters
• See past insider clusters on one company and what followed → Insider clusters for one company
• Spot unusually large trades by members of Congress → Congressional trade anomalies
• Size a trade from account risk and a stop → Tools › Trading › Position Calculator (Position size calculator)
• Find which semiconductor trade corridor moved most unusually last month → Tools › Supply Chains › Semiconductor Supply Chain, Movers (Semiconductor supply chain)
• Find where a company's fabs are on a map → Semiconductor supply chain › Facilities › Map (Semiconductor supply chain)
• See export-control actions dated around the chip chain → Semiconductor supply chain, Risk › Policy (Semiconductor supply chain)
• List one politician's trades, or the biggest contract awards, with filters → Tools › Trading › Ask the Data (Ask the Data)
• Generate a PDF deep dive on one ticker → Tools › Trading › Report Generator (Report Generator)
• Get a Portfolio Risk report emailed every week → Report Generator, Email schedules (Report Generator)
• Compute a portfolio's 1-day VaR or a correlation matrix → Data › Models & Data › Risk Models (Risk models)
• Download the Excel add-in templates → Cmd-K XLTP, bottom of Risk models (Risk models)
• Rerun or rename a saved VaR setup or peer set → Data › Models & Data › Saved Models (Saved models)