How to read CME FedWatch probabilities

Start with the meeting, outcome and observation time

A CME FedWatch percentage is a model-derived estimate for a specified federal funds target-rate outcome at a selected Federal Open Market Committee (FOMC) meeting. It uses 30-Day Federal Funds futures prices and assumptions about how rates change. It is not a promise from the Federal Reserve, an analyst poll or a known true probability. [1][5]

Before interpreting the number, identify the meeting date, the target range or set of outcomes, the observation time and the current target used as a baseline. Those labels tell you what the percentage means. This guide explains the reading process with fictional rates; it does not report today's chances of a cut.

Read the event before the percentage

The meeting date is the horizon of the estimate. The observation time tells you when the underlying market information was captured. The user guide lets readers compare estimates for the same meeting at different observation times. A movement between those observations does not mean the meeting date moved. [2]

CME's method aggregates the probability of a cut by adding probabilities for target levels below the current target. The probability of one specified cut size can cover a narrower set of outcomes. For a later meeting, a target level below the observation-time baseline could reflect changes at earlier meetings. It does not, by itself, identify the action taken specifically at that later meeting. [1]

Separate the policy target, daily EFFR and monthly average

The FOMC establishes the federal funds target rate or target range. The effective federal funds rate (EFFR) is a separate measured benchmark. It is not defined as the midpoint of that range. Under its normal transaction-based method, the New York Fed calculates EFFR as a volume-weighted median of overnight federal funds transactions reported through FR 2420. The median identifies the rate at the 50th percentile of dollar volume. [4][6][7]

CBOT Chapter 22 specifies a calendar-month arithmetic average of daily EFFR for final settlement. A day without a published rate, such as a weekend, uses the last preceding published rate. Counting only business days would therefore change the calculation. Before expiry, a futures price implies a monthly average; the final settlement calculation uses the realized daily rates under the contract rules. [3]

Rounding also belongs to a particular stage. The New York Fed rounds the published daily EFFR to the nearest basis point. Chapter 22 rounds the monthly arithmetic average to one tenth of a basis point, with ties rounded up. These rules do not establish how the tool rounds displayed probabilities. [3][7]

A futures price supplies a rate input, not a probability

The quotation relationship is implied monthly average rate = 100 − futures price, with the rate expressed in annualized percent units. A fictional price of 95.10 therefore implies a monthly average EFFR of 4.90% per year. It does not mean a 95.10% probability, a 4.90% probability or interest of 4.90% earned in one month. [3]

Moving from that average to meeting outcomes needs more assumptions. CME's published method uses 25-basis-point increments, a proportional EFFR response to rate changes and a binary-tree construction. One basis point is 0.01 percentage point, so 25 basis points equal 0.25 percentage point. These are features of the described model, not rules that restrict every possible policy decision. [1]

The method uses months with and without meetings to connect starting and ending rates. A monthly average can blend days before and after a decision, so it need not equal the rate expected after that decision. Calendar weights matter. The simple example below isolates this issue; it does not reproduce the tool's complete sequence of meetings and branches.

Follow one fictional two-level example

Assume a 30-calendar-day month with 15 days at a pre-decision EFFR of 5.00% per year and 15 days after a fictional decision. Suppose the futures-implied monthly average is 4.90% per year. These are invented EFFR levels and day counts, not current target rates, range midpoints or an actual FOMC calendar.

The assumed weights reproduce the post-decision mean: 0.20 × 5.00% + 0.80 × 4.75% = 4.80% per year. Combining 15 days at 5.00% with 15 days at that mean gives the 4.90% monthly average. The 80% is conditional on this model's two permitted levels and proportional rate response. It is not an independently observed chance of a real decision.

The mean alone does not identify every possible distribution. A 20% weight on no change and 80% on a 25-basis-point reduction gives the same expected 20-basis-point reduction as 60% on no change and 40% on a 50-basis-point reduction. That second distribution is a mathematical counterexample outside the two-level setup, not an alternative description of FedWatch. Specifying the possible outcomes is part of the calculation.

Compare like measurements

Similar-looking numbers can answer different questions. CME's user guide distinguishes FOMC participant rate projections from a futures-implied year-end rate marker. Neither should automatically be read as the probability of an action at one meeting. [2]

The Fed's dot plot is part of the Summary of Economic Projections (SEP). It records individual FOMC participants' judgments about the appropriate federal funds target-range midpoint or target level at specified year-ends or over the longer run, conditional on their information and policy assessments. The historical 18 March 2026 SEP is used here only to explain this meaning, not as the latest forecast. Participant counts or a median are not probabilities for the next meeting. Nor are they a vote tally, a committee promise or a guaranteed rate path. Anonymous dots do not identify a participant or reveal one person's path across horizons. [8]

In the CME user-guide comparison, participants' year-end projections, a futures-implied year-end rate marker and the model-implied outcome distribution for a selected meeting are three different objects. Sharing a panel or a similar number does not align their inputs, horizons or meanings. Comparing a year-end median with one meeting probability is not a test of either measure's accuracy. [2] [8]

An event-contract price also needs its own price-to-probability qualifications. Our separate prediction-market guide explains those conditions. Sharing a topic such as interest rates does not make two instruments equivalent or establish that any particular market is offered.

Keep the assumptions attached to the estimate

CME describes its calculated probabilities as estimates that depend on the model's assumptions. A change in the futures inputs can change the readout without making the earlier number a promise that was broken. A high model weight also leaves room for other outcomes; it does not establish certainty. [1]

For a useful comparison, keep the selected meeting, outcome set and baseline consistent, and label both observation times. This guide does not test predictive accuracy, rank forecasts or establish the true probability of a policy decision. Understanding the calculation provides no individual legal, age, account or trading eligibility and is not a recommendation to enter a financial position.

Common questions

Does a futures price of 95.10 mean a 95.10% chance?

No. In the fictional example, 100 − 95.10 gives a monthly average EFFR of 4.90% per year. Turning a monthly rate input into outcome probabilities requires additional assumptions. The quotation itself is not a probability.

Is a lower rate at a later meeting the same as a cut at that meeting?

No. A later meeting's target-level distribution can reflect changes at earlier meetings. Define whether the question concerns a level relative to today's baseline, a particular cut size or the action specifically at that later meeting.

Can the monthly average alone reveal the probability of a cut?

Not without assumptions about the possible rate levels and their timing. Different outcome distributions can have the same mean. The fictional 80% weight follows from the stipulated two-level model, not from the average in isolation.

Does a high percentage guarantee the Fed's decision?

No. It is an estimate for defined outcomes under a model, not a policy commitment or guaranteed result. Read the horizon and assumptions even when one outcome has most of the displayed weight.

Sources and scope

Sources checked on 18 October 2026. New York Fed documentation supports the daily EFFR definition; CBOT Chapter 22 supports monthly quotation and settlement. CME documentation supports the described FedWatch model and readout. All worked rates, prices and day counts are fictional. Access dates are not publication dates; no live probabilities or predictive-accuracy test are supplied.

1. CME FedWatch methodology
2. CME FedWatch user guide
3. CBOT Chapter22:30-Day Federal Funds Futures
4. New York Fed:Effective Federal Funds Rate
5. CME FedWatch Tool
6. Federal Reserve:FOMC
7. New York Fed:Additional Information about Reference Rates
8. Federal Reserve: March 18, 2026 Summary of Economic Projections (historical vintage)