On August 7, 2026 at 00:00 UTC, Polymarket stops resolving its crypto up/down markets on a single snapshot price. From that moment, those markets settle on a time-weighted average price, computed by Chainlink over a window that scales with the length of the market. If your bot’s edge depends on what the price is at one exact instant, that edge has a deadline, and it is next week.
This is the verified spec, what the change does to each kind of strategy, the three things Polymarket has not documented yet, and what to do in the three-day gap between the feeds going live and the rules changing. Everything here is dated, and where something is unknown, this post says so rather than guessing.
Polymarket TWAP settlement: the spec
From Polymarket’s developer announcement of July 30, 2026:
| Market duration | TWAP window | Previously |
|---|---|---|
| All crypto 5-minute markets | 30-second TWAP | Single price at expiry |
| All crypto 15-minute markets | 60-second TWAP | Single price at expiry |
| All crypto 4-hour markets | 60-second TWAP | Single price at expiry |
Effective August 7, 2026 at 00:00 UTC. Chainlink TWAP testnet feeds are available now, and mainnet feeds plus delivery through Polymarket’s Real-Time Data Streaming service arrive August 4, 2026. Polymarket is adding $1 million in liquidity rewards across impacted markets through August. The stated purpose is protecting market integrity in these markets.
One warning before you read anything else about this change. Polymarket signalled a TWAP move around July 12 that covered 5-minute contracts only, and some coverage from that week described it as a five-minute average with 15-minute markets to follow later. Those numbers are wrong now. The July 30 announcement is the specification: three durations, 30 and 60 second windows, one date.
What actually changes in Polymarket resolution
Until August 7, these contracts work as the academic literature describes them: the strike is the oracle price sampled at the contract’s open, the settlement value is the oracle price at its close, and the up side wins only if the close print exceeds the strike. Two instants, each a single number.
After August 7, at least the close becomes an average instead of an instant. For a 5-minute market, the settlement value is a 30-second time-weighted average. For 15-minute and 4-hour markets, 60 seconds.
That single sentence is the whole change, and it is enough to reprice several strategies, because a number you cannot move at one instant is a different object from a number you can.
Why Polymarket did this
Because the single-print design was measurably exploitable, and someone measured it.
A working paper dated June 30, 2026 by David Dai and Ruizhe Jia of Stanford and Shihao Yu of Singapore Management University, titled “Settlement Manipulation in Prediction Markets”, studied Polymarket’s 5-minute Bitcoin contract from its February 12, 2026 launch. The finding: after launch, spot order flow spiked at settlement times and reversed afterwards, the signature of temporary price pressure rather than information. Coverage of the paper reports 821 wallets capturing roughly $8.2 million, with the losses falling mostly on retail. The same signature was much weaker in the 15-minute contracts, and the authors’ own recommendation was to lengthen the contract horizon.
Note what Polymarket did with that recommendation, because it is the interesting part. It did not lengthen the horizon. The 5-minute market survives; what changed is the shape of the number it settles against. Polymarket kept its most popular product (the paper notes 5-minute and 15-minute crypto up/down markets traded over $4 billion within months of launch) and attacked the exploit instead of the product. Whether a 30-second window is enough is now an empirical question, and the honest answer is that nobody knows yet.
Polymarket TWAP strategy impact, by strategy type
The mechanism is simple enough to reason about directly. Moving a snapshot requires dislocating the price for one instant. Moving a 30-second average requires holding that dislocation for 30 seconds while every arbitrageur on the other side fades you. The cost is not slightly higher, it is a different order of problem, and it scales with the window.
What follows is reasoning from that mechanism, not a measured result. Nobody has post-change data yet, and any page claiming otherwise before August 7 is guessing.
Settlement snipers and final-seconds pushes. This is the strategy the change was built to end. A push that has to survive the whole window costs far more and is far easier to fade. The paper’s own evidence that the effect was much weaker at longer horizons is the closest thing to a preview we have.
Latency arbitrage against stale quotes. Not directly targeted, and not dead. If your edge is reading the underlying move and hitting a Polymarket quote that has not repriced yet, the settlement mechanism was never what you were exploiting. What does change is the other side of that trade: makers carry less risk of a settlement flipping on one print, which lets them quote tighter and reprice with more confidence. Expect the stale-quote windows you were harvesting to get narrower rather than disappear.
Market makers. Structurally better off. Less settlement-flip risk, plus $1 million in liquidity rewards through August aimed at exactly this seat. The inventory risk of a real market move during the window is unchanged, because a TWAP still moves when the market really moves.
Copy trading. Worse, and worth saying plainly to anyone running it. The wallets that looked most worth copying in 5-minute crypto were disproportionately running the strategy that stops working. A track record built before August 7 is a record of a game with different rules.
The strategy that replaces sniping. Once the window opens, the settlement value stops being a future unknown and starts being a partially-determined running average. That turns a speed problem into a forecasting problem: estimating where the average lands given what has already printed. Speed still helps, but it is speed applied to tracking and quoting rather than to landing one order at one instant.
Three things Polymarket has not documented
This is the part other coverage will skip, and it matters more than any strategy opinion, because you cannot finish an integration without these answers.
Where the window sits. Nobody outside Polymarket knows whether the 30 seconds look back from the close, start at the close, or straddle it. The docs consistently call these lookback windows, which points to looking back from the boundary, and every comparable venue averages the period before expiry. That is an inference, not a statement, and the point at which your outcome becomes effectively determined moves depending on the answer.
Whether the open is averaged too. The strike was a single print under the old design. The announcement addresses resolution and does not say whether the open gets the same treatment. If the strike stays a single print, then a snapshot remains in the contract, at the other end.
How the TWAP is computed. Chainlink computes and signs the value, and Polymarket’s docs state plainly that Chainlink does not currently publish the feed’s sampling boundaries, weighting, rounding or missing-input behaviour, and that developers should not try to reproduce the value without a specification. So the settlement number is an authoritative black box. You verify it, you do not recompute it.
Treat all three as open until a post-August-7 market’s own rules text answers them. We will update this page when it does.
Integrating the Polymarket TWAP feeds, and the three-day gap
The feeds arrive August 4. The rules change August 7. Those three days are a gift, and the correct use of them is a rehearsal.
You can consume TWAP prices two ways. Chainlink Data Streams gives you the signed reports directly, with the SDK handling authentication and reconnection. Polymarket’s Real-Time Data Streaming WebSocket relays the same Chainlink-computed values without credentials, using the topics crypto_prices_twap_thirty and crypto_prices_twap_sixty, filtered with an exact compact JSON form such as {"symbol":"btc/usd"} and kept alive with a PING text frame every five seconds. Feed identifiers change at mainnet launch, so pull them from Polymarket’s Chainlink TWAP documentation on the day rather than from any blog, including this one.
The operational details that will actually bite you:
- The 30 and 60 second figures are lookback windows, not publication rates. Do not infer the window from how often updates arrive, and do not use update frequency as a health check. Use the observation timestamp for freshness.
- There is no snapshot, no history and no replay. Subscriptions start with the next update. A disconnect at the wrong moment leaves a hole you cannot backfill from the stream, so your reconnect path needs Data Streams as the authoritative source.
- Before launch, subscriptions may return “topic not found”, and a rejected pre-launch subscription may not retry on an open socket. Recreate the subscription after launch rather than assuming it heals.
- Decoding is not verifying. The decode step parses the report; it does not verify the signatures. For anything settlement-sensitive, follow Chainlink’s verification requirements.
- Your clock is a dependency again. The SDK signs each request and expects your server clock within five seconds of Chainlink’s. A drifting box fails in ways that look random, which is the same failure mode we have written about before in a different context.
- Preserve the exact value. The feed carries a full-accuracy fixed-point number alongside a display value. Keep the exact one as an integer or decimal string. Rounding a settlement price through a floating point type is how you produce a bot that is right about the market and wrong about the outcome.
A staged migration plan for the next week
- Now through August 3. Integrate against testnet. Build the freshness threshold, the gap fallback, the reconnect path and signature verification. This is the whole job; the mainnet swap is mostly configuration.
- August 4. Swap in the mainnet feed identifiers the moment they are published, and re-confirm the report schema and price scale before reusing your decoder. Do not assume testnet identifiers map to mainnet ones.
- August 4 to 6. Run your settlement logic in shadow mode against the live mainnet TWAP while the old rules still apply. Log both the old snapshot print and the new average for every market you trade, and compare. This is the only window in which you can measure the difference on live data with nothing at stake.
- August 7. The rules change at 00:00 UTC. Size down for a few sessions. Your fill assumptions, your edge estimates and your risk model were all fitted to a different settlement mechanism.
- After. Re-derive your edge from post-change data before scaling back up. The temptation will be to assume the old model still holds with a small correction. It might not.
What infrastructure fixes after the Polymarket TWAP change
Our honest read, as the company selling the servers.
This change does not make a fast, well-placed box less useful, but it does change what you use it for. The pitch that speed lets you win a race to a single settlement print is over on August 7, and anyone still selling that after next week is selling something they have not read about. What survives is unglamorous and real: a stable low-latency path for quoting and cancelling, a machine that stays up through a window you cannot pause, clean clock discipline for a feed that now cares about your clock, and enough headroom to run a second data path in shadow while you migrate.
What it does not fix is the part that matters most this month. If your strategy was the snapshot, no server anywhere restores it. Rebuild the strategy, then measure the path. The method for measuring the path properly has not changed, our browser-based latency checker still gives you a zero-install first look, and neither has our position on providers who quote you a fixed millisecond number from a sales page.
For the wider picture of where the order engine sits and what drives these markets, our Polymarket crypto trading infrastructure guide covers it, and it is being updated for this change. If your client predates the April exchange upgrade, fix that first with the V2 migration guide, because a bot that cannot place an order has no opinion about settlement. Both our Polymarket locations come with a free 3-day demo if you want to test the path yourself before August 7.
Frequently Asked Questions
TWAP stands for time-weighted average price: instead of reading the price once, you average it across a defined window of time. On Polymarket, from August 7, 2026, crypto up/down markets settle against a Chainlink-computed TWAP rather than a single price at expiry. The window is 30 seconds for 5-minute markets and 60 seconds for 15-minute and 4-hour markets.
Crypto up/down markets stop resolving on a single snapshot price and begin resolving on a Chainlink-computed time-weighted average. The window scales with market length: 30 seconds for 5-minute markets, 60 seconds for 15-minute and 4-hour markets. The change takes effect at 00:00 UTC.
The announcement names all crypto 5-minute, 15-minute and 4-hour markets. It does not name other durations, and Polymarket has not published a per-asset list, so confirm against the rules text of the specific market you trade.
It raises the cost substantially, because a manipulator now has to sustain a price dislocation across the whole averaging window instead of moving one print. Whether that is enough at 30 seconds is an open question, and the honest answer is that there is no post-change data yet.
It ends the specific game of racing to influence or capture a single settlement print. It does not end low-latency trading, because quoting, cancelling and reacting to the underlying still reward speed. The edge shifts from sniping an instant toward tracking and forecasting a moving average.
Two routes: Chainlink Data Streams directly with credentials, or Polymarket’s Real-Time Data Streaming WebSocket without them, subscribing to the 30-second or 60-second TWAP topic with an exact symbol filter and a five-second PING heartbeat. Mainnet feeds arrive August 4, 2026, and current identifiers and SDK versions are in Polymarket’s documentation.
Polymarket has not documented this as of July 31, 2026. The documentation calls the 30 and 60 second figures lookback windows, which suggests they look back from the boundary, and comparable venues average the period before expiry, but that is inference rather than a published statement. Check the rules text of a live market after August 7.
The shape of one is visible, though nobody has post-change data yet. Once the averaging window opens, the settlement value stops being unknown and becomes a partially-determined running average, so the work moves from landing one order at one instant to estimating where that average lands and quoting against it. Treat any specific claim about profitability, including ours, as untested until there is live data after August 7.
If any part of your edge depends on the settlement being a single price at a single instant, yes, and the useful work is testing rather than guessing. Run your settlement logic in shadow against the live mainnet feed between August 4 and 6, log the old print and the new average side by side, and size down for the first sessions after the switch.
Specifications on this page come from Polymarket’s developer announcement of July 30, 2026 and Polymarket’s Chainlink TWAP documentation, verified July 31, 2026. Research figures come from the Dai, Jia and Yu working paper of June 30, 2026. Fast-moving details, including feed identifiers and the exact window mechanics, change; Polymarket’s own documentation is always the current authority. This is infrastructure and educational content, not trading or financial advice.
