By Kresmion Research, July 28, 2026
The Federal Open Market Committee announces its decision on Wednesday July 29 at 2:00 pm Eastern. The economists surveyed for it expect no change in rates. That would leave the upper bound of the target range at 3.75 percent, where it has been since April.
The people betting money on it disagree by a wide margin. The Polymarket contract on a 25 basis point increase at this meeting is quoted at 26.30 percent this morning, bid 26.20 and offered 26.40. Kalshi, a separate venue, prices the same outcome at 26.50 percent. Two weeks ago the same contract was near 8 percent.
The explanation that has travelled with the repricing is inflation. A piece published on July 24 by The Motley Fool and syndicated through Yahoo Finance, The Probability of a July Fed Rate Hike Has Tripled Over the Last Week, gave three causes: the Strait of Hormuz and its effect on oil, price stickiness in core PCE, and the cost of the artificial intelligence data centre build out.
Two of those three are testable against data that has printed since. One of them holds up. The other does not, and neither does the timing story that would replace it.
Key takeaways
| Measure | Value | Date | Context |
|---|---|---|---|
| Polymarket, 25bp increase on July 29 | 26.30 percent | July 28, 08:50 UTC | 7.97 percent on July 3 |
| Kalshi, same outcome | 26.50 percent | July 28, 08:50 UTC | 8.7 million contracts of open interest |
| Surveyed economist forecast | 3.75 percent, unchanged | July 29 event | Same as the previous meeting |
| Headline CPI, year over year | 3.5 percent | July 14 print | Forecast 3.8 percent, prior 4.2 percent |
| Core PCE, year over year | 3.4 percent | June 25 print | Prior 3.3 percent, no update before the meeting |
| Initial jobless claims | 187,000 | July 23 print | Forecast 212,000, prior 209,000 |
| 2 year Treasury yield | 4.37 percent | July 23 close | 4.16 percent on July 16 |
| Brent crude, continuous | 87.83 dollars | July 27 close | Still 28 percent above July 1 |
What the market actually did
Kresmion captures the Polymarket contract on a 25 basis point increase at the July meeting. The daily averages of the quoted probability:
| Date | Probability of a 25bp increase |
|---|---|
| July 3 | 7.97 percent |
| July 14 | 7.30 percent |
| July 15 | 5.40 percent |
| July 16 | 3.63 percent |
| July 17 | 4.31 percent |
| July 18 | 4.87 percent |
| July 19 | 5.43 percent |
| July 20 | 6.55 percent |
| July 21 | 11.61 percent |
| July 22 | 18.40 percent |
| July 23 | 24.55 percent |
| July 24 | 24.20 percent |
| July 25 | 24.42 percent |
| July 26 | 18.52 percent |
| July 27 | 22.58 percent |
| July 28 | 26.30 percent, live quote at 08:50 UTC |
Where you start changes the story, so both starting points are given. Measured from the July 16 low of 3.63 percent the contract has risen 22.7 points and multiplied by seven. But July 16 is the lowest reading in the entire recorded series, a trough reached two days after a soft CPI report. Measured from July 3, the first day of capture, the move is 7.97 to 26.30, a rise of 18.3 points and a little over three times. The second number is the fairer one.
There is a gap in the record worth stating. Kresmion has 16 distinct capture days for this contract, July 3 and then July 14 onward. There is no June history, so this contract has no long term base rate of its own.
The move is not confined to Wednesday. A separate Polymarket contract asking whether the Federal Reserve raises rates at any point in 2026, with 4.85 million dollars of lifetime volume, was at 51.50 percent on July 16 and is at 77.50 percent now. It trades on the same venue, so it is not independent confirmation, but it shows a path being repriced rather than a single meeting.
Down the calendar, the September contract puts an increase at 55.50 percent against 35.00 percent for no change, making an increase the single most likely September outcome where a week ago no change was the favourite. October reverts to nothing, with no change at 60.50 percent against 24.50 percent. Those two contracts are much smaller than July, at 1.09 million and 77,692 dollars of lifetime volume respectively, so the October reading in particular should be treated as thin.
How unusual is a 20 point week
The contract's one week change is 19.85 points. That invites the word extraordinary, and it does not quite earn it.
Across liquid macro contracts on Polymarket, using 183 market days from 30 contracts with at least 500,000 dollars of volume, the median absolute seven day move is 2.00 points and the ninetieth percentile is 18.70 points. A move of 19.85 points or larger occurs on 6.6 percent of macro market days. This week sits at the 93rd percentile of its own category. It is a top decile week, roughly one in fifteen, not a freak.
The size of the market matters, and two different figures are involved. Polymarket reports 21.5 million dollars of lifetime volume on the 25 basis point contract, 31.0 million on the no change contract, and 109.0 million across the five July outcomes together, of which 2.38 million traded in the past 24 hours. Separately, Kresmion has captured the individual trade tape since July 14, a sample of roughly 13 percent of that venue volume: 47,056 trades worth 14.0 million dollars across the five contracts, and 11,785 trades worth 2.30 million on the 25 basis point contract. Every participant figure below is computed on that captured sample. On Kalshi the matching contract shows 8.7 million contracts of open interest and the no change side 8.9 million, each settling at one dollar.
The inflation data mostly went the other way, with three exceptions
The July CPI report landed on July 14. Every line came in below forecast.
| Series | Actual | Forecast | Previous |
|---|---|---|---|
| Inflation rate, year over year | 3.5 percent | 3.8 percent | 4.2 percent |
| Core inflation rate, year over year | 2.6 percent | 2.8 percent | 2.9 percent |
| Inflation rate, month over month | negative 0.4 percent | negative 0.1 percent | 0.5 percent |
| Core inflation rate, month over month | 0.0 percent | 0.2 percent | 0.2 percent |
Producer prices followed the next day. Headline PPI fell 0.3 percent on the month against a forecast of no change and a prior of 0.6 percent. Consumers moved the same way. The University of Michigan preliminary survey on July 17 put one year inflation expectations at 4.2 percent, down from 4.6 percent, with the five year figure unchanged at 3.3 percent.
Three series ran the other way and all three belong in the record. Import prices rose 0.3 percent on the month against a forecast of a 0.7 percent fall, with the annual rate up from 6.7 to 7.1 percent. Core producer prices rose 4.7 percent year over year against 4.6 percent previously, missing forecast while still accelerating. And the prices paid component of the same Philadelphia Fed survey quoted later in this note rose from 53.2 to 53.9.
The strongest case against this reading
That case is core PCE, which is the measure the Federal Reserve actually targets, and it has not printed since any of this began.
At its last reading on June 25 the core PCE price index was 3.4 percent year over year, up from 3.3. The headline PCE index was 4.1 percent, up from 3.8. The quarterly annualised measure came in at 4.6 percent against 2.9 percent the quarter before.
Those are the freshest numbers on the Federal Reserve's preferred gauge, they accelerated on every horizon, and the committee decides on Wednesday without an update. The next print lands on July 30, the day after. So the syndicated explanation this note is testing is correct about core PCE, and a reader who weights the Fed's own target measure above CPI should discount much of what follows. Excluding that series because it falls outside a convenient window would be exactly the error this note is trying to avoid.
The oil leg, and a control that actually works
Oil is the one leg of the inflation case that can be tested cleanly, because a market prices it directly.
Brent rose from 68.40 dollars on July 1 to 96.23 on July 23 on the roll adjusted continuous series, a 40.7 percent move, then fell to 87.83 by July 27, down 8.73 percent from the peak. WTI ran the same shape, 89.01 to 81.94, down 7.94 percent. A note on those figures: July 27 is a contract roll boundary in both, so raw front month closes are not comparable across it and the percentages above use the roll corrected series. The raw closes would show a larger fall that is partly mechanical.
It would be tempting to call that a collapse of the inflation impulse that failed to unwind the hike bet. It is not. Brent is still 28 percent above where it started the month and sits near the top of its range. An 8.7 percent retracement from a spike high does not remove an oil driven inflation impulse, and the oil thesis does not predict that hike odds should fall while oil stays this elevated. That test does not falsify what it appears to falsify.
A better test exists on the same venue, in the same window, among the same traders. Polymarket runs a contract on whether WTI reaches 95 dollars during July. It is a pure bet on the oil spike and nothing else.
| Date | WTI reaches 95 dollars | Fed raises 25bp |
|---|---|---|
| July 16 | 10.78 percent | 3.63 percent |
| July 23 | 59.23 percent | 24.55 percent |
| July 25 | 41.06 percent | 24.42 percent |
| July 26 | 19.44 percent | 18.52 percent |
| July 27 | 10.04 percent | 22.58 percent |
| July 28, live | 4.35 percent | 26.30 percent |
The oil bet round tripped. It went from 10.78 percent to 59.23 and back to 4.35, ending well below where it started. Over the same five sessions the Fed contract fell once and then made a new high. If the hike bid were a repackaged oil trade, the two should have died together, because the people holding them are drawn from the same pool on the same platform. One died and the other did not.
That is a genuine control with real variance, and it is the strongest single piece of evidence here. It does not prove what the repricing is about. It does show what it is not merely a proxy for.
What ran hot in the same window
Something changed in the third week of July, and a cluster of activity and labour readings beat forecast by wide margins in it.
| Date | Series | Actual | Forecast |
|---|---|---|---|
| July 15 | NY Empire State manufacturing | 15.6 | 8.8 |
| July 16 | Philadelphia Fed manufacturing | 41.4 | 13.0 |
| July 16 | Initial jobless claims | 208,000 | 217,000 |
| July 17 | Housing starts | 1.427 million | 1.310 million |
| July 17 | Michigan consumer sentiment | 54.4 | 51.0 |
| July 23 | Initial jobless claims | 187,000 | 212,000 |
The Philadelphia Fed print was more than three times its forecast. Housing starts rose 19.0 percent on the month after falling 15.2 percent. Claims fell to 187,000, the lowest of the six readings Kresmion holds, which begin on June 18 at 226,000 and run 215,000, 215,000, 215,000, 208,000.
Now the part that stops this from being an explanation. Attributing the repricing to those prints does not survive a day by day check. Of the 20.9 points the contract added between July 16 and July 23, only about a third landed on days carrying a US activity beat. The two largest single day moves, 5.06 points on July 21 and 6.79 points on July 22, fell on days with no US activity release at all. The Philadelphia Fed blowout printed on July 16, the day the contract hit its low. And on July 27, when durable goods missed badly, the probability rose four points anyway.
So the honest statement is that the activity data and the repricing occupy the same fortnight, and that Kresmion cannot show one caused the other. With nine to fifteen daily observations, no attribution of this move to any data series is establishable, including the oil attribution rejected above. Anyone claiming otherwise, in either direction, is fitting a story to a fortnight.
What does stand on its own is an arithmetic point that needs no timing story. The effective federal funds rate was 3.63 percent in June and headline CPI is running at 3.5 percent, so the real policy rate is about 0.13 percentage points, close to zero. A committee that believes policy is already near neutral, watching employment and activity firm, can consider an increase without any inflation surprise at all. That makes a hike arguable. It does not make it likely.
The cash market moved too
Prediction markets are a single source here. Polymarket and Kalshi trade the identical event and are arbitraged against each other, so agreement at 26.30 and 26.50 percent shows the arbitrage works rather than that two independent judgements coincided.
The separate reading is the cash Treasury market, with different participants and a different instrument. The 2 year yield rose from 4.16 percent on July 16 to 4.37 percent on July 23, an increase of 21 basis points across five sessions. Monday's auction says the same with live money: the 2 year note sold on July 27 at a stop of 4.315 percent against 4.189 percent at the June 23 auction.
Two qualifications, both of which cut against reading too much into this.
The move is not a front end story alone. Over the same five sessions the 10 year rose 14 basis points, so roughly two thirds of the front end move was a selloff of the whole curve. The 2s10s spread compressed from 41 to 34 basis points, and those 7 basis points are the only part attributable to the front specifically.
The move is also less rare than a short lens suggests. Against the preceding year, where 250 rolling five session windows average 0.79 basis points with a standard deviation of 8.56, 21 basis points is 2.36 standard deviations and only one other window in the year was as large, the 22 basis points to March 24. Widen the lens and it shrinks. Across the last ten years, 88 of 2,500 windows were at least this large, the 96.5th percentile. Across the full series back to 1976, 967 of 12,528 were, the 92.3rd percentile. The past year has been unusually quiet in the front end, with a five session standard deviation of 8.56 basis points against 19.20 over the full history, so calling this a two sigma event would be true and would also be an artifact of the window. It is the largest move in a year and that is as far as it goes.
One further limit. It would be natural to ask whether the hike probability and the 2 year move together day by day, and Kresmion cannot answer that. The tick capture begins on July 14 and the Federal Reserve's yield series publishes only through July 24, so the two overlap on nine sessions inside a single month. Nine observations from one episode cannot support a correlation and none is offered. Note also that the contract has added 6.65 points in the last 24 hours, a leg no Treasury print yet covers.
One trader, or ten thousand
A 20 point move in an event contract invites the objection that somebody pushed it.
In the trade tape Kresmion captured since July 14, across the five July contracts, there are 47,056 trades from 10,057 distinct wallets. On the 25 basis point contract alone there are 11,785 trades from 2,671 distinct wallets against 2.30 million dollars of captured notional.
The concentration in full, rather than only its most flattering line: the largest wallet is 8.96 percent of that flow, the top four are 32.9 percent and the top ten are 48.2 percent. One wallet put through 195,348 dollars in a single trade. That is concentrated, and it is still a long way from one actor moving a market.
The flow is two sided today. Buying of the no change contract has reached 173,195 dollars this morning across 478 trades, with a further 211,780 dollars spent selling the hike outcome, together about 9 percent of the contract's 24 hour volume. The price rose 6.65 points anyway.
Two cautions on those figures. They describe trading, not positions, which the venue does not publish per wallet, so a participant who bought and sold appears large while holding nothing. And they cover roughly an eighth of the venue's reported volume on these contracts.
A note on what is missing. The natural next test is a control on an unrelated central bank. The nearest candidate, the Polymarket contract on the Bank of Japan holding rates in July, fell from 98.50 percent on July 22 to 97.10 on July 24 before recovering to 98.60 today. It moved on the same two days the Fed contract peaked, which if anything points to a global rate repricing rather than something specific to Washington. It is not used as a control here, and the case against a single actor rests on the wallet distribution instead.
The evidence against
The single most likely outcome on Wednesday is still nothing. No change trades at 73.75 percent on Polymarket and 73.50 percent on Kalshi. Roughly three quarters of the money says the committee sits still, and a 26 percent probability that resolves at zero is not a wrong price.
Core PCE, set out above, is the strongest single argument on the other side and it is unrefuted.
The activity story has holes. Durable goods orders published on July 27 rose 0.3 percent against a forecast of 2.5 percent, though the core capital goods proxy in the same release beat at 0.9 against 0.8 from a prior of negative 1.9. Retail sales in July came in on forecast at 0.2 percent with the control group in line at 0.5, so retail was flat rather than weak. Pending home sales fell 5.4 percent against a forecast of negative 0.5. Preliminary building permits fell 3.0 percent, revised to negative 2.6 in the final.
The aggregate has been going the wrong way for the whole episode. Kresmion's economic surprise index has fallen steadily through July, from 0.123 on July 9 to 0.095 on July 11, 0.051 on July 14, negative 0.021 on July 20 and 0.001 today. Across the entire repricing the country moved from running ahead of forecast to running level with it.
There is no Summary of Economic Projections at this meeting. June carried one, July does not, so there will be no updated dot plot and the statement language carries more weight than usual.
What would change the read
Wednesday settles the July contract but not the question. The falsification condition sits one meeting further out.
If the repricing reflects a genuine revision to the policy path, the September contract should hold an increase as its most likely outcome even if the committee does nothing on Wednesday. If September falls back below no change on a July hold, what was priced this week was the risk of one meeting rather than a changed path. The September contract on a 25 basis point increase is at 55.50 percent this morning, on 1.09 million dollars of lifetime volume. The committee went on to hold that Wednesday with three dissents in favor of an increase, and what the decision did to a record crowded dollar position is covered in the August 2 note.
Data notes
Prediction market prices are read from Polymarket and Kalshi as of 08:50 UTC on July 28, 2026, with the historical series from Kresmion's own tick capture rather than a vendor summary. That capture begins on July 3 for the July contract and July 14 for the trade tape, and covers roughly 13 percent of reported venue volume, so all participant figures describe trading rather than open positions. Economic actuals, forecasts and priors come from the macro calendar. Treasury yields are Federal Reserve daily series, which publish with a lag, and every yield is labelled with the date it belongs to rather than presented as a live quote. Crude oil percentages use roll adjusted continuous series because July 27 is a contract roll boundary in both Brent and WTI; the oil series runs through July 27, one day behind the prediction market figures. Where a move is called large it is measured against more than one window and every window is named, because the same move can look extreme against a quiet year and ordinary against a long history. Daily probabilities are means of all captured quotes on that date, except July 28 which is a single live quote, and they include weekend dates on which the contract traded thinly.
Frequently asked questions
Is the Federal Reserve going to raise rates on Wednesday?
Nobody knows, and this note does not predict it. The interest is that a liquid market prices the chance at about 26 percent while the surveyed economist forecast is for no change.
Why does a prediction market disagree with economists?
They are different instruments. A survey collects considered forecasts from a fixed panel. A market prices continuously and is moved by whoever is willing to trade. Neither is automatically right.
Does the CPI report rule out a rate increase?
No. A committee can tighten because activity and employment are firm with the real policy rate near zero, without an inflation surprise. And core PCE, which the Fed targets directly, was accelerating at its last print.
How reliable is a Polymarket price?
It depends on the contract. Polymarket reports 21.5 million dollars of lifetime volume on this one. Kresmion captured 2.30 million dollars of its actual trade tape since July 14, and in that portion 2,671 distinct wallets traded it, with the largest accounting for 8.96 percent of the flow and the top ten for 48.2 percent. A second venue quotes the same outcome within 0.2 points.
What is the single number to watch after Wednesday?
The September contract on a 25 basis point increase, at 55.50 percent this morning. Whether it holds above the no change outcome after the July decision separates a path revision from a one meeting scare.