SOFR Futures and Rate Expectations: Implied Rates, Meeting Probabilities, and CORRA Futures

SOFR futures are cash-settled CME contracts priced at 100 minus the Secured Overnight Financing Rate over a reference period: a calendar month for One-Month SOFR futures (SR1), and a quarter running from one IMM date to the next for Three-Month SOFR futures (SR3). Because each price pins down the market’s expected average rate for a specific window, a strip of SOFR futures can be turned into an implied path for the Federal Reserve’s policy rate and, with the FOMC calendar, into meeting-by-meeting probabilities of a move.

In short: an SR1 price of 96.19 implies an average SOFR of 3.81% for that month. If a policy meeting falls inside the month, the average is a day-weighted mix of the rate before and after the decision, so the post-meeting rate can be solved for. Dividing the implied change by 25 basis points gives an expected number of moves, which CME’s FedWatch methodology splits into probabilities for the two nearest outcomes. Those numbers are risk-neutral: they include risk premiums and technical distortions such as month-end funding spikes, and they are not polls of anyone’s forecast. Canada’s equivalents are the Montreal Exchange’s One-Month (COA) and Three-Month (CRA) CORRA futures, which compound the rate daily on an actual/365 basis.

This piece builds on the contract overview in interest rate futures explained. It works from CME Rulebook Chapter 460, CME’s “Understanding SOFR Futures,” CME’s published FedWatch and SOFRWatch methodologies, and the Montreal Exchange’s CORRA futures specifications.

The misconception: “FedWatch shows the odds the Fed will cut”

Headlines quote FedWatch as “markets see a 70% chance of a cut.” Two things are wrong with that framing. First, the standard FedWatch tool is built from 30-Day Federal Funds futures, not SOFR futures; CME publishes a separate tool, SOFRWatch, for SOFR. Second, the probabilities are derived from prices, and prices include compensation for risk. Monika Piazzesi and Eric Swanson (“Futures Prices as Risk-Adjusted Forecasts of Monetary Policy,” Journal of Monetary Economics, 2008) found that excess returns on fed funds futures “have been positive on average and strongly countercyclical,” predictable with indicators such as employment growth, so futures-implied paths tend to overpredict rates in recessions. The probabilities are a useful translation of prices, not a measured belief.

SOFR itself

SOFR is published each business day by the Federal Reserve Bank of New York as a volume-weighted median of overnight repo transactions secured by Treasuries. It was first published on April 3, 2018. For November 9, 2026, the New York Fed published SOFR at 3.88%, on about $3.1 trillion of volume, with the federal funds target range at 3.75% to 4.00%, where the FOMC held it on October 28, 2026.

SOFR tracks the policy range but not exactly. It is a secured rate and moves with Treasury collateral supply and dealer balance sheets. From November 3 to November 9, 2026 it printed 3.97%, 3.94%, 3.92%, 3.89% and 3.88% on successive business days (New York Fed data): a 9 basis point drift with no policy change, as month-start funding pressure faded. That noise matters for extracting probabilities, as we show below.

The two CME contracts

From CME Rulebook Chapter 460 (SR3) and CME’s “Understanding SOFR Futures” (for SR1):

	Three-Month SOFR (SR3)	One-Month SOFR (SR1)
Rate	Compounded daily SOFR over the Reference Quarter	Arithmetic average of daily SOFR over the contract month
Contract size	$2,500 x IMM Index ($25 per basis point)	$4,167 x IMM Index ($41.67 per basis point)
Period	Third Wednesday of a month (included) to the third Wednesday three months later (excluded)	The calendar month
Price	100 minus R	100 minus R
Minimum tick	0.005 ($12.50); 0.0025 ($6.25) with four months or less to termination of trading	0.005 ($20.835); 0.0025 ($10.4175) for the nearby month
Last trading day	Business day before the third Wednesday of the month the quarter ends	Last business day of the month

Both contracts are cash-settled: no repo is delivered, and open positions are closed out against the final settlement price through ordinary variation margin (the mechanics are in physical vs cash settlement). Two details trip people up.

Naming. The rulebook calls the month in which an SR3 Reference Quarter ends the “delivery month,” but CME’s own explainer notes that, by market convention, a contract is named for the month in which the quarter begins. The “December 2026” SR3 covers Wednesday, December 16, 2026 to Wednesday, March 17, 2027.
Compounding. Rule 46003.A. defines R as the product of (1 + d/360 x r) over each business day, minus 1, scaled by 360/D, where weekends and holidays carry the prior business day’s rate. The result is rounded to 0.0001; a value of 3.14155% rounds to 3.1416%, for a final settlement of 96.8584. SR1, by contrast, is a simple average, like fed funds futures.

SR3 is the deep, liquid curve instrument, listed many years out. SR1 is the precision tool for the next few meetings, because its months line up with the calendar months that FOMC meetings fall in.

How FedWatch turns prices into probabilities

CME’s “Understanding the CME Group FedWatch Tool Methodology” lays out the procedure using fed funds futures. It assumes that “rate hikes/cuts are uniformly sized in increments of 25bps” and that the rate reacts proportionally. The steps:

Find an anchor month with no meeting. Its futures price gives the average rate for the whole month, which must equal both the rate at the end of the previous month and at the start of the next.
Solve each meeting month. If N days fall before the change takes effect and M after, then the month’s average = (N x start + M x end) / (N + M). Knowing the average and one end, solve for the other.
Convert to moves. Divide (end - start) by 0.25. Split the result into its whole part and fraction: 2.1103 expected hikes means an 88.97% probability of 50 basis points and 11.03% of 75.
Chain meetings. Multiply the probabilities across meetings to build a binary tree of cumulative outcomes.

The methodology also specifies direction: rates from an anchor month propagate “forward in time for only one month and backwards for as many months needed until another non-FOMC anchor month is reached.”

We checked our implementation against CME’s own example for the September 21, 2022 meeting: an October 2022 price of 96.9400 (3.06% average), a September price of 97.4475 (2.5525%), N = 21 and M = 9. The start-of-September rate solves to (2.5525 - 9/30 x 3.06) / (21/30) = 2.3350%, a change of 72.5 basis points, or 2.9 hikes: 10% for 50 basis points and 90% for 75. That matches the CME document.

SOFRWatch: the extra assumption

SOFR futures add a step, because SOFR is not the policy rate. CME’s SOFRWatch methodology holds “the basis (or spread) between the lower end of the Fed’s policy rate range and the O/N SOFR” constant. It measures that spread as the latest gap between the range floor and the effective fed funds rate, plus the market-implied EFFR-SOFR spread from SR1 versus fed funds futures over the next three months. CME’s example: a range of 5.00% to 5.25%, a 14 basis point floor-to-EFFR spread, a 2 basis point EFFR-SOFR spread, and a 25 basis point hike give an implied SOFR of 5.00 + 0.14 + 0.02 + 0.25 = 5.41%.

Worked example: a SOFR strip around three FOMC meetings

The prices below are hypothetical, chosen to be consistent with tick sizes and to make the arithmetic visible. They are not market prices: they price a path of cuts, whereas the FOMC’s two dissents on October 28, 2026 were in favour of a hike. The meeting dates are real: the Federal Reserve’s calendar lists December 8-9, 2026, and (tentatively, until confirmed at the prior meeting) January 26-27 and March 16-17, 2027. Decisions take effect the day after the announcement, so December has 9 days at the old rate and 22 at the new one; January 27 and 4; March 17 and 14.

Start from SOFR at 3.88%, its November 9 fixing, 13 basis points above the 3.75% floor, and assume that spread holds (the SOFRWatch assumption). November is a month with no meeting and serves as the first anchor; February 2027 is the second.

SR1 contract	Price	Implied average SOFR
December 2026	96.1900	3.8100%
January 2027	96.2400	3.7600%
February 2027	96.3700	3.6300%
March 2027	96.3950	3.6050%

December 9 meeting (forward from the November anchor). End-of-December SOFR = (31 x 3.8100 - 9 x 3.88) / 22 = 3.7814%. The change is -9.86 basis points, or -0.3945 of a 25 basis point move: 39.5% probability of a cut, 60.5% no change.

January 27 meeting (backward from the February anchor). End-of-January SOFR equals February’s average, 3.6300%. Start-of-January = (31 x 3.7600 - 4 x 3.6300) / 27 = 3.7793%. The change is -14.93 basis points, or -0.5970 moves: 59.7% cut, 40.3% no change.

Notice that December’s implied end (3.7814%) and January’s implied start (3.7793%) differ by 0.2 basis points. In theory they are the same day’s rate. In practice each contract trades on its own and in ticks, so the strip is never perfectly consistent; FedWatch’s propagation rule exists to decide which price wins.

March 17 meeting (forward from the February anchor). Start-of-March = 3.6300%. End-of-March = (31 x 3.6050 - 17 x 3.6300) / 14 = 3.5746%, a change of -5.54 basis points, or -0.2214 moves: 22.1% cut.

Chaining the three meetings gives cumulative probabilities for the policy range:

Cumulative change	After December	After January	After March
No change	60.5%	24.4%	19.0%
-25 bp	39.5%	52.0%	45.9%
-50 bp		23.6%	29.9%
-75 bp			5.2%

The expected change by the end of March is 0.3945 + 0.5970 + 0.2214 = 1.21 cuts, about 30 basis points.

The same path through an SR3 price

If SOFR follows the expected path (3.88% to December 9, 3.78% to January 27, 3.63% to March 17), compounding the daily rate through the December 2026 SR3 Reference Quarter (91 days, with US holidays on December 25, January 1, January 18 and February 15) gives R = 3.7178% and a fair price of 96.2822. The simple average over the same days is 3.7009%. Compounding adds about 1.7 basis points at these rate levels, enough to matter for anyone backing meeting probabilities out of SR3 instead of SR1.

What distorts the probabilities

Funding noise inside the month. Suppose traders expect SOFR to average 2 basis points higher across December than the policy spread implies, because of year-end balance sheet constraints. That 2 basis points is spread over 31 days but solved into the 22 post-meeting days, biasing the implied end-of-December rate up by 2 x 31/22 = 2.8 basis points and cutting the implied probability of a December cut by about 11 percentage points. Small funding effects become large probability effects whenever the post-meeting window is short.
The spread assumption. SOFRWatch holds the spread to the range floor fixed. The November 2026 drift from 3.97% to 3.88% shows how much it can move with no policy change.
Risk premiums. Piazzesi and Swanson’s result means the implied path embeds a premium that varies with the business cycle. Over a few weeks it is small; over a year it is not.
The 25 basis point grid. A market pricing a 15 basis point move gets reported as a 60% chance of 25 basis points. If the committee might move by 50, or between meetings, the binary tree cannot say so.
Futures versus forwards. Futures are marked to market daily (see how futures margin works), so a long position gains cash when rates fall (when reinvestment rates are lower) and pays when rates rise. That makes futures slightly less valuable than forward rate agreements, so futures-implied rates sit a little above forward rates. The gap, called the convexity bias, is negligible for the next few months and grows with maturity.

CORRA futures in Canada

The Canadian Overnight Repo Rate Average (CORRA) is Canada’s overnight risk-free benchmark. The Bank of Canada announced in July 2019 that it would become CORRA’s administrator when an enhanced methodology took effect, and the Montreal Exchange’s analysis supporting the new contracts describes that enhancement as a trimmed median of overnight general collateral repo trades in Government of Canada securities. With CDOR’s final publication on June 28, 2024, CORRA futures became the Canadian short-rate futures market. (For how the Montreal Exchange and its clearing house, CDCC, fit into the Canadian market, see how futures exchanges work.)

From the Montreal Exchange’s specifications:

	One-Month CORRA (COA)	Three-Month CORRA (CRA)
Rate	Compounded daily CORRA over the contract month	Compounded daily CORRA over an IMM Reference Quarter
Size	C$2,500 x Index (C$25 per basis point)	C$2,500 x Index (C$25 per basis point)
Day count	Actual/365	Actual/365
Calculation period	First business day of the month to (not including) the first business day of the next	Third Wednesday to third Wednesday
Listing	Up to the nearest 7 months	Nearest 12 quarterly months
Tick	0.0025 (C$6.25) nearest; 0.005 (C$12.50) others	0.0025 (C$6.25) nearest; 0.005 (C$12.50) others
Last trading day	Last business day of the month	Business day before the third Wednesday of the delivery month
Clearing	CDCC	CDCC

Three differences from the US contracts change the arithmetic. One-month CORRA is compounded, not averaged. The day count is 365, not 360. And the Bank of Canada sets its target for the overnight rate on eight fixed dates a year, with the change taking effect, as the Bank states, “the day after its announcement” (the practice since 2021).

Worked example: the December 9, 2026 Bank of Canada decision

On October 28, 2026, the Bank of Canada raised its target for the overnight rate by 25 basis points to 2.50%. Its next scheduled announcement is December 9, 2026. The December COA calculation period runs from Tuesday, December 1, 2026 to Monday, January 4, 2027 (34 days, with Toronto bank holidays on December 25 and 28 and January 1). A decision on December 9 takes effect December 10: 9 days at the old rate and 25 at the new.

Assume CORRA fixes at the 2.50% target until the decision. Compounding daily:

If the Bank holds: R = 2.5027% (compounding alone adds a quarter of a basis point).
If the Bank hikes 25 basis points: R = 2.6869%.

The full hike is worth 18.4 basis points on the contract. A hypothetical COA price of 97.4600 implies R = 2.5400%, so the implied probability of a hike is (2.5400 - 2.5027) / 0.1842 = 20.2%. Skipping the compounding and treating 2.50% as the no-change value would give 21.7%, an error of 1.5 percentage points from the compounding convention alone.

The same CORRA-to-target spread caveat applies as for SOFR: CORRA is a repo rate and does not sit exactly on the target every day.

The short version

SOFR and CORRA futures settle to realized overnight repo rates, averaged or compounded over a month or an IMM quarter, and are quoted as 100 minus that rate. Each month’s price fixes an expected average; with known meeting dates and effective-day conventions, the average can be split into pre- and post-meeting rates, converted into a number of 25 basis point moves, and chained into a probability tree, as CME’s FedWatch methodology does. The method is exact arithmetic on top of three assumptions: a stable spread between the overnight rate and the policy rate, a 25 basis point grid, and no risk premium. Each one can fail, and a 2 basis point funding distortion can swing a reported probability by 10 points. Professionals read the strip as a price of rate risk, cross-check SR1 against SR3 and fed funds futures, and treat headline probabilities as a translation, not a forecast.