If you trade or build bots on Polymarket’s crypto Up/Down markets, you’ve probably already heard the headline: as of August 7, 2026, settlement switched from a single price snapshot to a time-weighted average (TWAP). Prices get averaged over a 30-to-60 second window now instead of read once at the final tick. Manipulation problem solved, everyone moves on.
Except it isn’t solved. It’s half-solved. And the half that’s still open is the half nobody’s writing about.
Let me walk you through what actually changed, why it happened, and the one detail buried in the live market rules that changes how you should think about these contracts going forward.
First, the part that’s real
Polymarket’s short-duration crypto markets - 5-minute, 15-minute, and 4-hour Up/Down contracts on BTC, ETH, SOL, and a handful of other assets - used to resolve on a single price read at the exact moment the contract expired. If you could move that one price for even a second, you could decide the outcome of a market with real money riding on it.
People did exactly that. A research paper out of Stanford and Singapore Management University studied about 16,000 five-minute Bitcoin contracts and found something uncomfortable: roughly 1 in every 300 wallets that traded these markets was running this exact play, collectively pocketing around $8.2 million while breaking even on everything else they did. The signature was unmistakable - a spike in order flow in the final ten seconds before expiry, then an immediate reversion once the payout landed. That’s not someone trading on information. That’s someone paying to move a number.
Here’s the part that should make you angry if you’re a regular trader on this platform: 93% of the losses from this landed on retail. This wasn’t whales fighting whales. It was a small number of sophisticated wallets quietly taxing everyone else.
So Polymarket fixed it - for the close. Settlement now averages price over the final 30 seconds (5-minute markets) or 60 seconds (15-minute and 4-hour markets), computed by Chainlink. Moving a single tick doesn’t do anything anymore, because that tick is now just one of many samples getting blended into an average. To actually influence the outcome now, you’d need to sustain a price dislocation across the entire window, while every arbitrageur watching gets to trade against you the whole time. That’s not a small speed bump. That’s a completely different, much more expensive game.
Good. Genuinely good. This needed to happen.
The part nobody’s saying out loud
Here’s what almost every writeup of this update - including some genuinely thorough, technically detailed ones - leaves out: a Polymarket Up/Down contract has two price points, not one.
There’s the open - the “price to beat,” read once at the moment the market starts. And there’s the close - the resolution price, read at the moment it ends.
The August 7 update only touched the close.
The open is still exactly what it always was: a single, instantaneous snapshot. No averaging, no window, no cost to sustain a move. If you read the live market rules carefully, they say a market resolves “Up” if the TWAP over the window is at or above the price at the beginning of the range - not an average at the beginning. The price. One number, one instant, unchanged from before this whole update happened.
Think about what that actually means. Polymarket took the exploit that was measurably costing retail traders millions of dollars, and fixed exactly one end of the lever. The other end is sitting right there, just as pullable as it was in July.
Now, before you go thinking this is some obvious remaining exploit everyone’s about to pile onto - it’s genuinely a worse trade than the old one. Pushing a close is surgical: you already hold your position, you know exactly which side benefits you, and the payoff lands immediately. Pushing an open is speculative: you’re setting the strike before the order book has even formed around it, and then you still need the entire rest of the market to go your way. Same lever, meaningfully worse odds. That’s probably exactly why it wasn’t the first thing anyone fixed - it’s a much less attractive target than the close ever was.
But “less attractive” isn’t “closed.” And there’s one loose thread even the most careful analysis I found couldn’t answer: for back-to-back short markets, a new contract’s open lands at the exact clock boundary where the previous contract’s TWAP window just finished closing. Whether that adjacency creates anything usable - some kind of information leakage or positioning edge - is completely untested. Nobody has data on it yet. It’s just the most obvious place to look if you’re trying to find where the residual edge in these markets actually lives.
Why this distinction should change how you build
If you’re running a bot, or even just trading these markets manually, here’s the practical shift this creates.
The old game was: predict a single number at a single future instant. The new game is: predict an average of many numbers across a window - which is a genuinely different forecasting problem, not just a harder version of the old one.
And there’s a trap hiding inside that shift that’s easy to miss. Once the averaging window opens, you’re not purely forecasting anymore - you’re partially watching the thing you’re trying to predict unfold in real time. Every tick that prints inside the window makes your estimate of the final average mechanically more accurate, whether or not you’re actually good at this. If you backtest a strategy and it looks like it gets more accurate as expiry approaches, that might just be arithmetic - you’re literally seeing more of the average - not evidence of a real edge. Separate those two things explicitly, or you’ll fool yourself with a backtest that looks great and means nothing.
The strategies that die here are the ones built purely around timing - landing one order at one precise instant and hoping nobody notices. The strategies that survive, and arguably get better, are the ones built around genuinely reading the underlying asset faster than the market reprices, because that edge was never about the settlement mechanism in the first place.
And one more thing worth knowing if you use any copy-trading tool on this platform: the wallets that looked most impressive on 5-minute crypto leaderboards were disproportionately the ones running the strategy that just died. Their track record isn’t a lie, exactly - it’s just a record of a game that doesn’t exist anymore. Treat any pre-August-7 performance history on these markets as historical trivia, not a signal.
Where this leaves things
Polymarket did something real here. The documented exploit - the one that was quietly taxing retail traders to the tune of millions of dollars - is genuinely dead at the point it was happening. That’s worth acknowledging plainly instead of being cynical about it.
But “we fixed the exploitable part” and “we fixed the exploit” aren’t the same claim, and the gap between them is sitting in plain sight in the rules text if you read it closely enough. One end of every contract on this platform is still a single, ungamed-against number. Whether anyone builds something around that before it gets closed too is genuinely an open question - and it’s the one worth watching, not the part everyone’s already written about.
Sources: Polymarket’s July 30, 2026 developer announcement; live post-launch market rules verified August 7, 2026; Dai, Jia & Yu, “Settlement Manipulation in Prediction Markets” (working paper, June 30, 2026). This is analysis and commentary, not trading or financial advice - do your own research before risking money on any of this.