The Fed Is a Coin Flip on Polymarket: 49.5% Hike, 49.5% Hold, and Kalshi Agrees to Half a Cent

At 2026-09-06 07:00 UTC on 6 September, Polymarket's September Fed decision event priced a 25 basis-point hike at 49.5% and no change at 49.5%, with a 25 bp cut at 0.4%. $99 million has traded on the event since it opened and $2.87 million of it in the last day. Kalshi's contracts on the same meeting closed their last hour at 49.0% for the hike, 49.0% for no change and 1.0% for a cut.

That is not two markets disagreeing. It is two markets saying the same thing with the precision of a coin, ten days before the Federal Open Market Committee meets on 15 and 16 September. The interesting number is not either probability but the distance between them, because that distance is what an arbitrage bot would be paid to close, and it has sat inside the fee floor for most of the month.

Polymarket charges takers C × feeRate × p × (1 − p), with a 0.04 rate on politics and economics markets at the time of writing, and charges makers nothing; Kalshi charges takers 0.07 × C × p × (1 − p). At 50 cents that is 1.0 cent per contract on Polymarket and 1.75 on Kalshi: a taker on both legs pays about 2.75 cents per dollar before slippage. The largest daily gap between the venues on the hike contract in the last thirty days was 3.5 cents, inside that floor.

In July, Vultax watched the same two venues on matched Fed-rate contracts every 85 seconds for a day and found that the one exactly matched pair held a 5.7-cent gap that never closed; the September decision contracts are better matched, resolve on the same announcement, and are correspondingly tighter. Where there has been a lean it has been Kalshi's: on most days this month its hike contract traded a cent or two above Polymarket's, consistent with a US-regulated venue whose traders weight the public case for a hike that Fed governors have been making, against a global venue that priced the hold more heavily until this week.

Over the last seven days Vultax's pipeline observed 13,659 fills worth $5.90 million on the three September contracts, from 5,120 wallets. 131 of those wallets belong to the round-the-clock cohort from the cadence study, 2.6% of the wallets on the contract, and they made 19.8% of the trades and 8.3% of the notional. Across all of Polymarket that cohort makes 58% of trades; on the Fed markets it makes a fifth. By Polymarket's standards the Fed is unusually human.

The fills arrive in bursts around the data. Friday's jobs report is the tallest hour of the week; the CPI print on 10 September will be the next.

Four things decide whether the coin lands.

• CPI, 10 September - the last macro print before the meeting; on the equivalent release before the July decision the hike contract moved eight cents.
• The gap, not the level - if Polymarket and Kalshi separate by more than three cents on the same contract, one of them is being moved by a large order, and the level alone tells you nothing.
• The cut contract - at under a cent it is the cheapest lottery ticket on either venue; a move above two cents means someone believes something the venues do not.
• Post, do not take - on a coin flip the taker fee is the whole edge; a resting order on Polymarket pays no fee at all.

Polymarket prices are the CLOB price history for each contract's Yes token at hourly fidelity; Kalshi prices are hourly candle closes from its public trade API. Both were read at the time stamped above and the page does not update itself; the interactive edition carries the full hourly series. Fill counts come from Vultax's copy of public on-chain fills and undercount fills not yet ingested at read time. Fee rates are as published by each venue on 5 September 2026. Nothing here is a forecast of the decision, and nothing here is financial advice.