Polymarket Ditched Single-Price Settlement for TWAP — Here's What Changed for Traders
On August 7, 2026, Polymarket's crypto up/down markets stopped resolving on a single price snapshot. Here's what a TWAP is, why Polymarket needed one, and what actually changed for anyone trading these markets.
The change, in one line
As of August 7, 2026, 00:00 UTC, Polymarket's crypto up/down markets no longer settle on the price at one exact instant. They resolve instead using a time-weighted average price (TWAP) computed by Chainlink over a short window before expiry — 30 seconds for 5-minute markets, 60 seconds for both 15-minute and 4-hour markets. Polymarket's developer team laid out the details in a post on X on July 30, 2026, and backed it with a full technical spec in its documentation the same week (Polymarket Developers on X; Polymarket Docs: Chainlink TWAP Prices).
On paper, this sounds like a minor plumbing change, only that it wasn't. For anyone trading these markets, running a bot against them, or providing liquidity, it changed the game.
What a TWAP actually is
A time-weighted average price does what it says on the label: instead of checking an asset's price once, at a single tick, it averages the price across a window of time. If Bitcoin spikes for two seconds right before a market closes and then reverts, a snapshot-based system captures that spike in full. A TWAP-based system mostly doesn't as the distortion gets diluted by the rest of the window.
None of this is new, exactly as decentralized exchanges have leaned on TWAP oracles for years, for the same reason, which is that they make manipulation expensive. A bad actor has to sustain an artificial price for the entire window rather than spiking it for an instant, which multiplies the capital required to move a settlement in their favor. Polymarket wasn't inventing a mechanism here — it was importing one that DeFi had already stress-tested.
Why Polymarket needed this
Polymarket launched its 5-minute crypto up/down contracts on February 12, 2026. They took off. According to an academic study of the market's design — "Settlement Manipulation in Prediction Markets," by researchers at Stanford's Department of Management Science and Engineering and Singapore Management University — Polymarket's five- and fifteen-minute crypto contracts traded over $4 billion cumulatively within months of launch (Dai, Jia & Yu, arXiv:2606.31675).
The same study is also the clearest account of why the fix was necessary. The researchers analyzed roughly two months of five-minute Bitcoin contracts and found a repeating pattern: unusually large orders on Binance in the final seconds before a Polymarket contract settled, followed by a rapid price reversal right after. They identified 821 wallets whose trading activity matched this signature, collectively earning about $8.2 million during settlement windows they classified as manipulated. Excluding market makers, 93% of the losses in those windows landed on retail traders — a figure the paper puts at roughly $7.61 million (Dai, Jia & Yu, arXiv:2606.31675). The researchers also found the pattern was largely absent in Polymarket's 15-minute contracts, which pointed toward a structural fix rather than a cosmetic one: lengthen — or average — the settlement window, and the exploit loses most of its economics.
Put plainly: someone with enough capital could nudge Bitcoin's price on a major exchange for a handful of seconds, collect a Polymarket payout timed to that exact instant, and unwind the position immediately after. The TWAP change was Polymarket's direct answer to that finding, not a feature it shipped ahead of the problem.
The mechanics of the rollout
Per Polymarket's own developer documentation, here's the specification that took effect:
Market duration | TWAP window |
|---|---|
5-minute crypto markets | 30 seconds |
15-minute crypto markets | 60 seconds |
4-hour crypto markets | 60 seconds Sponsored |
The averaging window scales with the length of the market rather than staying fixed across all durations (Polymarket Developers on X).
The price feed itself runs on Chainlink Data Streams, which brought dedicated TWAP feeds — in 30-second and 60-second variants — to mainnet on July 31, 2026, giving developers roughly a week to integrate before the settlement change actually took effect. Chainlink continued soak-testing the feeds through August 4, so early integrators had to treat that window as provisional rather than final (Polymarket Docs: Chainlink TWAP Prices).
Developers ended up with two supported ways to pull the data, and both remain the standard paths:
Directly through Chainlink Data Streams, using standard or sponsored credentials, via the Chainlink Data Streams catalog.
Through Polymarket's Real-Time Data Streaming (RTDS) service, a public WebSocket that relays the same Chainlink-computed values without requiring separate Chainlink credentials. RTDS went live on August 4, 2026, three days ahead of the settlement rules actually changing (Polymarket Docs: Chainlink TWAP Prices).
One detail worth flagging for anyone building against this: Chainlink hasn't published the exact sampling boundaries, weighting, or rounding behind the TWAP calculation, and Polymarket's own docs are blunt about it — don't try to reproduce the number yourself. Treat the feed as something you consume and verify, not something you recalculate (Polymarket Docs: Chainlink TWAP Prices).
The liquidity incentive
To keep market depth intact while participants adjusted, Polymarket committed $1 million in liquidity rewards across affected markets through August 2026, distributed as part of the same rollout (Polymarket Developers on X). Polymarket's own announcement didn't publish a breakdown of how that pool was split across the 5-minute, 15-minute, and 4-hour markets — some outlets reported specific per-market figures in the days after the announcement, but since that detail didn't come from Polymarket directly, it's worth treating with a healthy dose of skepticism until confirmed in Polymarket's market-maker documentation, which is the authoritative source for current terms.
Should you trust it? A few open questions
TWAP wasn't a solved problem the moment it shipped, and it's worth sitting with a few things the upgrade didn't fully answer — the kind of questions worth re-checking against Polymarket's current rules text rather than assuming were tidied up somewhere along the way.
For one, TWAP raises the cost of manipulation; it doesn't eliminate it. A well-capitalized actor can still try to sustain an artificial price across a 30- or 60-second window, especially in thinner markets. The fix changes the economics of the exploit — it doesn't make the exploit structurally impossible.
There's also a question the original announcement left hanging: whether the opening price of a contract, the strike it's measured against, gets the same TWAP treatment as the close. Polymarket's rollout was explicit about how markets resolve, but noticeably quieter on whether the open is averaged too. If only one side of a contract is smoothed, a snapshot vulnerability arguably survives at the other end.
Then there's the dependency question. Settlement now runs through an external oracle network with its own uptime, latency, and update cadence — and Chainlink's documentation is candid that the computation methodology behind these specific feeds isn't fully public. Polymarket, and by extension its traders, are trusting infrastructure they can verify but can't independently reproduce.
And it's fair to note the pattern here, too. Polymarket adjusted how its short-duration crypto markets settled more than once over the course of 2026, generally after users had already absorbed losses rather than ahead of them. Each individual fix has been reasonable on its own terms. A pattern of reactive patches to core market mechanics is still the kind of thing that costs a platform trust, even when each patch itself is sound engineering.
None of that makes TWAP the wrong call. It's a well-understood, industry-standard mitigation, and adopting it brought Polymarket's fastest markets in line with how DeFi has handled this exact problem for years. But "harder to manipulate" and "impossible to manipulate" are different claims, and it's worth keeping the two straight no matter how far removed you're reading this from the actual launch date.
What this means if you're trading these markets now
If you trade manually, settlement prices reflect an average rather than a single tick — strategies built around timing the exact expiry second stopped working the way they used to as of August 7, 2026.
If you run a bot or integration against Polymarket's crypto markets, Chainlink's TWAP feeds and Polymarket's RTDS service are the two supported data paths. Worth double-checking you're actually pulling from one of them rather than reconstructing settlement prices independently — Polymarket has been explicit that its TWAP methodology isn't something integrators should try to replicate on their own.
And if you're providing liquidity, don't assume the incentive terms described above still apply by the time you're reading this. Check Polymarket's market-maker documentation directly. Reward programs like this one tend to be time-boxed, and they typically get revised or retired on a schedule that has nothing to do with the underlying settlement mechanics.
Primary sources and further reading