HowPredictionMarketsPricedthe2025FedCutPath

The Fed cut rates three times in 2025. Polymarket's expected-cuts series closed its final trading day at 2.99 — 0.01 cuts away from the realized outcome. Kalshi finished the year at 3.18, also resolving YES on the 3-cuts bucket. Two independent venues, two different market structures, arriving at the same answer through 347 trading days of disagreement and revision.

We compared how two independent prediction markets priced the same forward-looking macro question — how many times will the Fed cut rates in 2025? — across the full year of trading:

• Kalshi — CFTC-regulated US prediction-market exchange, count-bucket markets
• Polymarket — onchain prediction-market venue on Polygon, count-bucket markets

Both venues ran parallel ladders covering 0 through 8+ cuts, with daily trading from late December 2024 through resolution in December 2025. From each venue's strike ladder we computed the daily implied expected number of cuts in 2025 — a probability-weighted mean across all buckets.

Comparable notional across the period: ~$8.9M Kalshi, ~$5.8M Polymarket — the most balanced cross-venue liquidity we've seen anywhere in the dataset. Both resolved YES on the "3 cuts" bucket.

Explore Prediction Markets data

1. Two independent venues, one trajectory. Across 347 trading days, Kalshi and Polymarket's implied E[N cuts] tracked each other closely. Both started near 2.5–3.0 in late December 2024 and reconverged at 3.0 by year-end, with similar broad arcs in between: a February dip during the recession scare, a recovery through spring, drift through summer, and a final climb into resolution.
2. The February recession scare was sharper on Kalshi than Polymarket. Both venues fell in early February, but the magnitudes differed: Kalshi dropped as low as 1.4–1.6 on several days while Polymarket mostly stayed at 2.0–2.5, briefly touching 1.69 on Feb 9. That divergence is itself a finding — Kalshi's thinner book repriced harder on the same news flow, which is a real cross-venue microstructure observation.
3. Polymarket priced slightly higher than Kalshi most of the year, but the spread varied. On most days, Polymarket sat above Kalshi by 0 to 0.5 cuts. The spread compressed near zero by September–October and flipped sign briefly in November. Directionally stable enough to be a tradeable signal, but not a constant.
4. Both venues converged to the right answer, with different precision. By early October, both had stabilized in the 2.8–2.9 range. Polymarket finished its last trading day (Dec 10) at 2.99 — within 0.01 cuts of the realized outcome of 3. Kalshi closed December at 3.18, slightly above 3 but still resolving YES on the 3-cuts bucket.

E[N cuts] is a probability-weighted mean across the count distribution, not a forecast that the Fed will cut "2.5 times." The same expected value can come from a tight distribution around 2–3, a bimodal split across 1 and 4, or a flat distribution across 1–4. The cross-venue agreement story we're showing is about first moments; a follow-up could compare full distributions across venues, where the February divergence in finding 2 suggests there's more to find than the mean alone reveals.

We also don't have an off-chain benchmark in this version. Adding CME Fed Funds Futures-implied expected cuts (from the ZQF26 contract) would provide a third independent estimate from the $300T+ rates market — a natural follow-up piece.

This analysis required daily VWAP pricing across 9+ strike buckets on two venues, with parsed count labels and normalized probability distributions. The new dataset makes this straightforward — without it, doing this comparison requires reconciling two vendor APIs, different strike-label conventions, and reconstructing the probability distribution from raw trade data.

The same methodology extends to cross-venue basis monitoring (the Polymarket–Kalshi spread on macro questions varies enough across the year to be worth tracking), distribution-shape analysis (comparing variance, skew, and bimodality across venues, particularly during stress events), and any multi-bucket macro question priced on both venues (annual counts, year-end levels, range markets).

• Daily implied E[N cuts] both venues — the main series
• Count-distribution market inventory — bucket-level notional and trading windows