The Hike That Moved — and the Hike That Disappeared

The October hike is nearly gone. The December hike isn't.

Between September 25 and October 3, the probability of an October 28 hike fell from 64.2% to 22.1%. That part is straightforward.

The more interesting move was underneath it.

The probability of at least one hike by December 9 slipped only from 92.6% to 85.2%. But the probability of two hikes by then collapsed from 51.0% to 17.9%. Expected tightening fell at every point on the curve: from 16bp to 6bp for October, 36bp to 26bp for December, and 49bp to 36bp for January 2027.

One hike moved. Another disappeared.

That distinction matters. The market did not simply push October's hike into December. It also abandoned a large part of the tightening it had previously expected.

The result is now strikingly close to the Fed's own September projection. The market is pricing about 26bp of additional tightening by December 9, almost exactly one further 25bp hike.

That is a real change in the rate path. It is also a useful correction to the story this series was telling three weeks ago.

The September 16–17 rally in the long end did not mark a lasting disinflationary repricing. It was a bounce. By late September, the long end had returned to—and then exceeded—its earlier highs.

The durable move came somewhere else: in meeting-specific rate pricing.

And there is an important complication. The cash Treasury market did not confirm the move. On October 2, the day of the weak payrolls report, both the 2-year and 10-year yields closed higher than the previous day.

So the market has moved toward a softer Fed path.

Treasuries have not yet agreed.

That tension is the real story of this note.

A forecast being right for the wrong reason is still a problem.

That is why this scorecard tracks both the call and the mechanism behind it. Some of the desk's earlier views held up. Some did not. One was simply wrong. Others remain open because the evidence has not settled the question.

The point is not to defend the earlier articles. It is to see what survived them.

Article 1 — Warsh vs. Wall Street: The NFP Showdown (Aug 6)

Article 2 — The Fed at Crossroads: The Wedge, the Steepener, and the Silence (Aug 12)

Article 3 — Two Authorities, One Price (Aug 30)

Article 4 — The Day After the Fed (Sep 17)

Article 4 ended with the 10-year yield coming off its highest levels in years. That relief did not last.

By late September, the long end had gone back through those levels. The October 2 move makes the point particularly well: the weak payrolls report pushed yields sharply lower initially, but the move completely reversed before the close.

Every yield in the table is a Treasury constant-maturity close from FRED series DGS2, DGS10, and DGS30, retrieved October 9, 2026. Those series copy the Federal Reserve’s H.15 release. The H.15 page dated October 8, 2026 prints October 2 directly: 2-year 4.83, 10-year 5.28, 30-year 5.63. The same October 2 2-year and 10-year were already in the October 5 release. T10Y2Y on October 2 is 0.45, and 5.28 minus 4.83 is 0.45. The October 1 10-year intraday high of 5.342% is still a CNBC report. It is not the H.15 close, which is 5.24.

The official data show the 2-year at 4.83% on October 2 versus 4.78% on October 1, while the 10-year was 5.28% versus 5.24%.

So the headline jobs miss did not leave the Treasury market lower.

It left both the 2-year and 10-year higher.

That is the part of this episode worth remembering.

Sep 29–30 — Williams speaks, Barr pushes back

John Williams spoke at the University at Buffalo on September 29. The New York Fed posted the prepared text that day. Its advisory scheduled the remarks for 2:00 p.m. EDT. The text says: “With the policy action we took at our September meeting, there is no need for urgency.” It also says that one further increase “may be appropriate late this year,” if the economy follows his forecast.

Michael Barr spoke the same day at the Detroit Economic Club. The Board posted the prepared text on September 29. That page does not give a clock time. He said he counted only two of the past twenty months as consistent with 2 percent core PCE inflation, and that he did not yet see a clear trend back to 2 percent.

The market nevertheless moved in Williams' direction. October hike pricing fell sharply before the PCE report and payrolls data arrived.

That chronology is clear.

The attribution is less so.

Sep 30 — August PCE

The BEA release of September 30, 2026, embargoed until 8:30 a.m. EDT, says the PCE price index rose 3.4 percent from a year earlier and 0.3 percent from July. Core PCE rose 3.0 percent from a year earlier and 0.2 percent from July. That same release included the annual update, so recent months were restated together with the new August figures.

The August 26 release, the one in print when the earlier articles were written, had July at 3.7 percent headline and 3.3 percent core. On the September 30 vintage, July is 3.4 percent headline and 3.0 percent core. August did not fall from 3.3 percent inside one series. Those are two vintages. The expected 3.7 and 3.3 figures are not in either BEA release, so they are not used here.

Wire reporting across this window was inconsistent — variously citing ~47%, ~35%, and as low as ~25% for the October meeting on different days, all attributed to “CME.” Those intraday secondary characterizations are not used here. The figures in this note come from CME FedWatch’s own target-rate probability tables, captured with timestamps (§07, Figure 1), which is the only basis on which the week’s moves can be compared like for like.

Oct 2 — September payrolls

The BLS Employment Situation for September 2026, USDL-26-1549, was embargoed until 8:30 a.m. ET on Friday, October 2. Payroll employment rose 29,000. The unemployment rate was 4.2 percent. Average hourly earnings rose 5 cents, or 0.1 percent, to $37.81, and 3.0 percent over the year. August payrolls were revised from +162,000 to +133,000. The 90,000 consensus is not in the BLS release. It stays a calendar figure, not an official one.

Article 4 proposed a simple test: if the next genuine dovish repricing arrived, the front end should lead rather than the long end.

That test has now fired—but only in the futures market.

The October 28 hike probability fell from 64.2% to 22.1%. That is a substantial change in the market's view of the next meeting.

Cash Treasuries tell a different story.

On October 2, the 2-year yield fell sharply after the payrolls release, then reversed and closed at 4.83%, five basis points above October 1. The 10-year did the same, finishing at 5.28%, four basis points higher.

So we have two markets telling two different stories.

Meeting-specific futures turned dovish. Cash Treasuries did not.

That is more interesting than trying to force the two into one narrative.

The simplest explanation is that the market changed its view of the Fed's near-term reaction function without yet changing its view of the forces driving the long end.

That leaves the central question open: does the cash curve eventually follow the futures, or do futures pricing move back toward the cash market?

We do not know yet.

The August gap is real. Core CPI was 2.4 percent and core PCE was 3.0 percent. That is 60 basis points.

The BLS release of September 11 says core CPI was 2.5 percent over the year ending July and 2.4 percent over the year ending August. The monthly core index rose 0.3 percent in August after 0.2 percent in July.

The PCE comparison in the first draft of this section was too clean.

On August 26, BEA put July core PCE at 3.3 percent. On September 30, it put August core PCE at 3.0 percent and revised July core PCE to 3.0 percent as well. August did not drop 30 basis points from the revised July reading.

The gap is narrower than the 70–80 basis points in the earlier articles. That part holds. The earlier call that the gap would narrow because CPI would rise does not. CPI eased, from 2.5 percent to 2.4 percent.

The line that PCE did almost all of the narrowing in the latest month does not hold either. Part of the change from 3.3 percent to 3.0 percent is the annual revision published on September 30, not a one-month decline.

The new observations are interesting on their own.

The September 16 FOMC produced an 8bp curve gap: the 2-year rose 7bp while the 30-year fell 1bp.

The three sessions from September 29 through October 1 were smaller individually—6bp, 6bp and 7bp—but they added up to a 19bp cumulative move, with the 2-year falling 14bp while the 30-year rose 5bp.

That tells us something about persistence. A sequence of moderate moves can matter more than any single headline.

What it does not tell us is whether Article 3’s result survives a new sample. Article 3 reported that policy events averaged about five times the curve gap of scheduled data releases, and it reported a very small p-value. This note does not confirm either number. The test was not rerun.

Article 3 reported a 9–15bp range for its four policy events and an 8bp ceiling for information events. September 16 lands exactly on that 8bp ceiling.

Interesting? Yes.

A result? Not yet.

The original event calendar, classifications, exclusions and full sample have not been rebuilt here. Until they are, the September 16 observation should remain an observation rather than being used to overturn—or confirm—the original result.

Philip Jefferson’s prepared remarks were released for delivery at 1:30 p.m. EDT on October 1 at the University of Virginia’s Darden School. The curve gap that day is the close-to-close change in the 2-year and the 30-year. It is not a measured reaction to the speech. The same is true of Williams on September 29 and the PCE release on September 30. The dates line up. The daily closes do not say how much of each move belonged to each event.

The three FedWatch tables give us something useful because they put the same market measure at three different points on the curve.

The important number is not simply the October probability.

Expected tightening fell by roughly 10–13bp at every meeting shown.

October fell from 16.1bp to 5.5bp.

December fell from 35.9bp to 25.8bp.

January fell from 48.6bp to 35.6bp.

The market did not simply move one hike from October to December.

It reduced the total amount of tightening it expected.

The December distribution makes that especially clear. The probability of two hikes fell from 51.0% to 17.9%, while the probability of one hike rose from 41.6% to 67.3%.

That is the heart of the move.

The market did not lose its expectation of another hike. It lost its expectation of two.

By December 9, expected tightening is now about 26bp—almost exactly one 25bp hike, which is where the Fed's September median projection puts it.

But the starting point matters.

The market reached that 26bp from roughly 36bp a week earlier. It did not sit there all along.

That makes the convergence meaningful.

One earlier figure should not be lined up against these. Article 3's August 31 reading, about 37bp by December 9, used the old target rate and still counted the September meeting as ahead.

The base case is still a hold on October 28 followed by one further 25bp hike in December.

But this is no longer a simple “October moved to December” story.

The market has done two things at once: it has sharply reduced the chance of an October hike and substantially reduced the chance of getting two more hikes by December.

The first move is about timing.

The second is about the amount of tightening the market thinks is coming.

The second is more important.

The main risk to this view is inflation. A materially hotter September CPI report, renewed oil pressure, or a further rise in long-end yields despite soft economic data would change the picture.

So would a renewed deterioration in the credibility of the December hike itself.

For now, the market is aligned with the Fed's September dot plot.

The cash Treasury market is not.

That is where the next signal has to come from.

There are still several things this note does not claim.

The FedWatch “1 WEEK” figures are CME retrospective calculations rather than archived snapshots of the screen on September 25. They are useful for a like-for-like comparison, but they should not be confused with contemporaneous tick-by-tick pricing.

The Article 3 event study has also not been rebuilt here. Its thresholds, event classifications and statistical tests remain prior findings.

The September 16, September 29, September 30 and October 1 curve gaps are therefore best viewed as new observations, not a new statistical test.

Finally, the sequence of Fed communication and economic releases shows when the repricing happened. It does not prove which event caused which portion of it.

That distinction is important enough to state directly: the market moved; the daily data do not tell us exactly why.

The three FedWatch captures carry two different timestamps — Oct 28 and Jan 27 at 08:38:47 CT, Dec 9 at 12:49:57 CT on the same day — and the Dec 9 table’s “1 DAY” column is labelled 1 Oct where the other two read 2 Oct. The “1 WEEK” columns all read 25 Sep, so the cross-meeting comparison in §07 is consistent; the intraday gap is noted for completeness and does not affect it.

The October 2 yields are direct H.15 and FRED observations, read on October 9, 2026. The 10-year is DGS10 at 5.28, not a figure derived from the 2-year plus the spread. The 30-year that an earlier draft left blank is 5.63.

August PCE and August CPI are taken from the BEA release of September 30 and the BLS release of September 11. The July PCE figures of 3.7 percent headline and 3.3 percent core are from the BEA release of August 26, which the September 30 annual update superseded. Survey expectations for those releases are not in the agency tables and are not used as facts.

Williams’s September 29 text is the New York Fed prepared remarks, scheduled for 2:00 p.m. EDT. Barr’s September 29 text is the Board’s prepared remarks; that page has no clock time. Jefferson’s October 1 text was released for delivery at 1:30 p.m. EDT. These are prepared texts, not verified transcripts of the words as spoken.

The September 24 Treasury buyback operation's size was not independently confirmed in sources reviewed for this note.

Article 3's falsification condition on an explicit Fed filter for intervention-vs-information price moves remains untested; this note found no new evidence either way.

CME FedWatch’s 22.1% October-28 hike probability differs from the 13–18% cited in contemporaneous wire reporting (Kalshi ~18%, Coindesk ~13%). This note uses CME’s own published figure. The gap is most likely instrument (fed funds futures vs. event contracts) or snapshot timing, but was not run to ground.

This note went through two external editorial audit rounds before publication. Per the series' standing practice, what changed is listed rather than silently absorbed — including corrections to this note's own drafts.

Carried forward from prior articles

Article 2's base case (“hold in September”) is marked FALSIFIED by the actual September 16 hike; its predicted mechanism for the CPI/PCE wedge narrowing (“from the CPI side rising”) is marked WRONG on direction — the narrowing came from PCE falling instead (§01, §05).

Article 3's downgrade of the “July 29 = credibility repricing” claim to an open question is carried forward unchanged; this note takes no new position on it. Its event-study thresholds are used as context in §06 but not re-derived, so nothing in this note scores against them.

Article 4's oil/long-end explanation for the September 16–17 reversal is reaffirmed as accurate for those two sessions specifically, and explicitly marked as NOT durable beyond them (§02) — the long end fully round-tripped to new highs within two weeks.

Corrected before publication

OCTOBER-28 PROBABILITY. An earlier draft reported 13–18%, sourced from wire characterizations. CME FedWatch's own published figure is 22.1% (no change 77.9%), against 64.2% / 35.8% a week earlier. The CME figure is now used throughout; the wire discrepancy is logged in Limitations.

CENTRAL THESIS. An earlier draft argued the December path was essentially unmoved (79.7% → 85.2% across Oct 1–Oct 3) and framed the week as a pure timing reassignment. That window was too narrow. Measured Sep 25 → Oct 3, expected tightening fell ~10–13bp at every meeting and P(two hikes by December) collapsed from 51.0% to 17.9%. Subtitle, §00, §03, §07 and Figure 1 were rewritten: one hike moved, a second was priced out.

OCTOBER 2 CASH YIELDS. Revised once closing rather than intraday data were obtained: both the 2-year and 10-year closed October 2 higher than October 1, a full reversal of the initial post-NFP decline. An earlier draft's “partly retraced” understated this. The audit also correctly caught that the draft compared an October 1 constant-maturity yield against an October 2 par-curve yield from a different provider, and the 2-year figure was corrected from 4.84% (an OTC-quote print) to 4.83%.

CONFIRMED (Oct 5). At the time of the audit the October 2 row could only be sourced from a secondary aggregator, because H.15 had not yet published it; the note flagged those values as provisional and stated what the official release should show. The H.15 release of October 5, 2026 confirms them exactly — DGS2 at 4.83 and T10Y2Y at 0.45, placing the 10-year at 5.28. The provisional language has been removed and §02 now rests on official constant-maturity data.

§06 FINDINGS DEMOTED. An earlier draft presented “September 16 landed exactly on Article 3's information-event ceiling” as a result and claimed the Fedspeak-as-policy-event extension was withdrawn on evidence. Both depend on a threshold this note inherits but never re-derived. §06 now leads with the two observations that stand on their own — the FOMC's front-end concentration, and the compounding of three individually small sessions into a −19bp weekly gap — and labels the threshold comparison an observation pending replication. The §01 scorecard rows were demoted to match.

Also corrected: an unsourced ~5.68% intraday 30-year high (removed from §02 and §08); the September payrolls consensus tightened from a “~84–90K” range to the 90K on the release capture; an Article 1 scorecard entry that wrongly cited the Oct 2 equity rally on a payrolls miss as a counter-example to mechanical NFP trading (it is the mechanical prediction); a §04 claim that the front end was leading “for the first time this cycle,” which ignored Article 2's own July 29 finding.

Figure 1 was rebuilt twice — first to remove a fabricated September 16 data point and show cross-source disagreement as explicit ranges, then again once the three CME FedWatch tables were obtained, replacing the estimated series entirely with values read directly from them.

What this check did, and what it did not

Checked against the original releases on October 9, 2026: the September Employment Situation (BLS, USDL-26-1549), the August and July Personal Income and Outlays releases (BEA), the August CPI release (BLS), the Williams, Barr, and Jefferson prepared texts, and H.15/FRED constant-maturity yields through October 2. The source list now matches those documents. Article 3’s event study was not rebuilt. Its fivefold comparison and its p-value stay Article 3’s findings. This note does not confirm them.

Federal Reserve Bank of New York, John C. Williams, “Unwavering Dedication,” prepared remarks, September 29, 2026. Media advisory: University at Buffalo, 2:00 p.m. EDT.

Board of Governors, Michael S. Barr, “Economic Conditions and Monetary Policy,” prepared remarks at the Detroit Economic Club, September 29, 2026. No delivery time is printed on the Board page.

Board of Governors, Philip N. Jefferson, “The U.S. Economy and Monetary Policy,” for release on delivery at 1:30 p.m. EDT, October 1, 2026, Darden School of Business, University of Virginia.

Board of Governors, FOMC statement and Summary of Economic Projections, September 16, 2026. Williams’s prepared text describes that decision as a quarter-point increase to 3-3/4 to 4 percent.

Bureau of Labor Statistics, The Employment Situation — September 2026, USDL-26-1549, embargoed until 8:30 a.m. ET, Friday, October 2, 2026. Payrolls +29,000; unemployment rate 4.2 percent; average hourly earnings +0.1 percent to $37.81 and +3.0 percent over the year; August payroll change revised from +162,000 to +133,000.

Bureau of Economic Analysis, Personal Income and Outlays, August 2026, embargoed until 8:30 a.m. EDT, Wednesday, September 30, 2026, including the 2026 annual update. PCE price index +3.4 percent from a year ago and +0.3 percent on the month; core PCE +3.0 percent from a year ago and +0.2 percent on the month. On this vintage, July is also +3.4 percent headline and +3.0 percent core.

Bureau of Economic Analysis, Personal Income and Outlays, July 2026, embargoed until 8:30 a.m. EDT, Wednesday, August 26, 2026. This is the superseded vintage: PCE price index +3.7 percent from a year ago; core PCE +3.3 percent from a year ago. It is cited only to show what was in print before the September 30 revision.

Bureau of Labor Statistics, Consumer Price Index — August 2026, USDL-26-1496, embargoed until 8:30 a.m. ET, Friday, September 11, 2026. Core CPI +2.4 percent over the year ending August, after +2.5 percent over the year ending July; core CPI +0.3 percent in August after +0.2 percent in July.

CME FedWatch and Kalshi prediction-market data as reported by CNBC, Bloomberg, Yahoo Finance, Coindesk, BNN Bloomberg and DeFiRate, Sep 29–Oct 3, 2026.

CNBC: “Labor market faltered in September…”; “10-year Treasury yield dives/ticks higher…”; “Fed's preferred gauge showed core inflation at 3.0%…”; “Traders now see little chance of a Fed rate hike in October…” (Sep 30–Oct 2, 2026).

Bloomberg: “US Firms Add Just 29,000 Jobs, Unemployment Rate Ticks Up,” Oct 2, 2026.

Yahoo Finance / BNN Bloomberg: “September 2026 jobs report”; “Fed may skip October but pull U.S. rate hike trigger in December,” Oct 2, 2026.

Benzinga, Coindesk, Seeking Alpha, DeFiRate: October-28 FedWatch/Kalshi probability snapshots, Oct 2–3, 2026.

Trading Economics was consulted while drafting and is not the source for the CPI, PCE, or Treasury yields in this note.

FRED, Federal Reserve Bank of St. Louis, series DGS2, DGS10, DGS30, and T10Y2Y, retrieved October 9, 2026. The series run through October 6. October 1 and October 2 read: DGS2 4.78 and 4.83; DGS10 5.24 and 5.28; DGS30 5.61 and 5.63; T10Y2Y 0.46 and 0.45. September 15 through September 30 in the §02 table are the same series.

Board of Governors, H.15 Selected Interest Rates (Daily), release page dated October 8, 2026. The October 2 column prints the 2-year at 4.83, the 10-year at 5.28, and the 30-year at 5.63. The October 5 release was the first business-day H.15 after October 2 and is the release cited in the October 5 correction. The older line in this list, which stopped the yield series on October 1, was wrong and has been removed.

Secondary Treasury pages (Stockmarketwatch, Spotthemoney, Trading Economics, Macrotrends) are not the source for any yield in this note.

CNBC: “Treasury yields are little changed as traders await Fed comments”; “30-year Treasury bond yield scales to highest level since 2002”; “S&P 500 closes higher to start October…” (Sep 16, 29–30, Oct 1, 2026).

Quartz / CNBC: Treasury buyback operation reporting, Aug 19 and Sep 9, 2026.

Prior Regime Intelligence articles: Warsh vs. Wall Street (Aug 6, 2026); The Fed at Crossroads (Aug 12, 2026); Two Authorities, One Price (Aug 30, 2026); The Day After the Fed (Sep 17, 2026) — all regimeintelligence.com/articles.

CME FedWatch target-rate probability tables, captured 3 Oct 2026: 28 Oct 2026 meeting (contract ZQV6, 08:38:47 CT); 9 Dec 2026 meeting (ZQZ6, 12:49:57 CT); 27 Jan 2027 meeting (ZQF7, 08:38:47 CT). Each table’s NOW / 1 DAY / 1 WEEK / 1 MONTH columns used as published.

An economic-calendar capture supplied the 90,000 payrolls consensus used in an earlier draft. That consensus is not in the BLS release and is not used as a fact in this note.