Every time a Fed meeting gets close, someone shows me two numbers for the same event and asks which one is right. The CME's FedWatch tool says the odds of a cut at the next meeting are one thing, and the event contract on the same meeting is trading at something else, sometimes ten points apart, and the natural reaction is that one of them has to be wrong. Usually neither is. They are measuring slightly different things with slightly different machinery, and once you understand what each one is actually doing, the gap tells you something instead of just annoying you.
Where the futures number comes from
Tools like FedWatch do not poll traders. They back a probability out of the price of fed funds futures, and that distinction is the whole story. Fed funds futures settle on the average daily effective fed funds rate over the delivery month. So the price of a given contract is a bet on what the average rate will be across that entire month, not on what the Fed does on any single day.
That averaging is where the arithmetic gets fiddly. If a meeting happens partway through a month, the settlement rate is a blend of the days before the decision at the old rate and the days after at the new rate. FedWatch and similar tools invert that. They take the current implied average, they know how many days sit on each side of the meeting, and they solve for the rate change that would produce the observed average. Then they translate that expected change into a probability distribution across the possible outcomes, usually assuming the Fed moves in standard increments.
None of that is a poll of opinion. It is a mechanical decomposition of a futures price into an implied path. That matters because the futures price carries baggage that has nothing to do with anyone's view on the meeting.
Why the event contract disagrees
An event contract on a Fed meeting is a much cleaner object. It pays out on a specific, binary outcome, a cut of a certain size at a certain meeting, and its price is more or less a direct probability. There is no averaging convention, no term structure, no month-blending. So when it disagrees with the futures-implied number, the difference is almost always coming from the futures side, not the event side. A few forces push them apart.
- Term premium. Futures prices embed a small risk premium for holding rate exposure over time. That premium is not a probability, but the decomposition treats the whole price as if it were expectation. So the implied odds can be biased, and the bias grows the further out the meeting is.
- Averaging artifacts. For meetings that fall early or late in a month, the number of days on each side of the decision changes how much a rate move shifts the monthly average. A small mis-specification in day counts, or a month with an unusual calendar, distorts the back-out. Meetings near a month boundary are the worst offenders.
- The standard-increment assumption. The futures math usually assumes the Fed moves in fixed steps. When the market is genuinely split between, say, a hold and a larger-than-usual move, the two-outcome decomposition smears that into a probability that does not match how the event contract, which can price each specific size separately, sees it.
- Retail flow and thin books. Event contracts trade in a different venue with a different crowd. Retail money crowds into the obvious outcome, and near a meeting the book can be thin enough that the last print is not a clean consensus. That is a real dislocation on the contract side, and it is the one you can sometimes trade.
When the gap is tradeable and when it is noise
Here is the rule of thumb I use. If the gap is coming from term premium or averaging conventions, it is an artifact and you should not fade it. Those are structural features of how the futures number is constructed, and the futures market is deep and efficient, so the futures side is not mispriced. It is just measuring a blended monthly average with a risk premium baked in. Trading against that is trading against your own misunderstanding of the tool.
If the gap is coming from retail flow piling into a thin event-contract book, that is a genuine dislocation, and it can be worth a position if the size and the liquidity justify it. The tell is usually the shape. Artifact-driven gaps are stable and grow smoothly with time to the meeting. Flow-driven gaps appear suddenly, often after a headline, and they sit on one specific outcome while the neighboring outcomes stay sane.
A quick way to sort one from the other:
- Pull the futures-implied odds and the event-contract price for the same meeting and the same outcome.
- Check where the meeting sits in its month. If it is near a boundary, discount a chunk of the gap as an averaging artifact before you do anything.
- Look at how far out the meeting is. The more distant it is, the more of the gap is term premium, and term premium is not your edge.
- Look at the event-contract book depth and recent prints. A wide spread and a jumpy last price on one outcome is flow, not consensus.
- Only after all of that, ask whether the residual gap is big enough, and the book deep enough, to justify a trade after fees.
The failure mode I see most often is someone treating FedWatch as ground truth and the event contract as the mispriced thing, then selling the event contract every time it prints richer. Against a thin retail book that occasionally works and feels like genius. Then a real repricing comes, the futures number was the one lagging because it averages a whole month and the event contract moved first on fresh information, and the position bleeds. The futures decomposition is not gospel. It is a model output with known biases, and near a meeting it can be the slower of the two.
Which number to trust for which job
For the shape of the expected rate path over the next year, trust the futures-implied curve. It is deep, it is liquid, and the term-premium bias is a known quantity you can adjust for roughly. For a clean probability on one specific decision at one specific meeting, the event contract is the more direct read, as long as you have sanity-checked the book. For anything close to a meeting where fresh information is landing, watch which one moves first, because that tells you which crowd is actually processing the news rather than sitting on a stale average.
When I want both side by side, plus the whale and disclosure flow that tends to front-run the obvious repricings, I keep them on one screen in Blockcircle so I am comparing the futures-implied path and the event-contract price against each other instead of squinting at two tabs. The point is not that one source wins. It is that they answer different questions, and most of the confusion comes from asking a monthly-average futures contract for a single-day probability it was never built to give you.