Every five minutes Polymarket opens a new question: will Bitcoin be higher at the end of this round than at its start? That is 288 rounds a day, each a small market of its own. The same runs every 15 minutes, hourly and daily, for Bitcoin, Ethereum, Solana and XRP.
We tracked these markets wallet by wallet, from the blockchain: 9,197 resolved Bitcoin 5-minute rounds from 19 October to 20 November 2026, plus every other series. This piece covers how a round is decided, how close the rounds really are, and what the numbers say about the people who trade them. The short version of the last part: winning most of your rounds and making money are two very different things here.
If you want to watch a round while you read, it is live at orcalayer.com/up-down.
How a round is decided
A round has a price to beat and a final price, and both are averages, not single ticks:
- Price to beat: the Chainlink BTC/USD price averaged over the 60 seconds just before the round starts (it is the previous round's final).
- Final price: the same 60-second average over the last minute of the round.
- Up wins if the final average is at or above the price to beat; a tie goes to Up. Otherwise Down wins.
Two things follow from the averaging. A spike in the very last second moves the result much less than it moves the chart: it is one second out of sixty. And a round is not settled until the last minute is over, because every second of it still counts.
We compute the price to beat, the final and the result ourselves, from the same Chainlink feed, and compare them with Polymarket's own resolution afterwards. Since we started on 17 November, that is 2,069 rounds across all series, with the same result every time.
How close the rounds get
For the 871 Bitcoin 5-minute rounds since 17 November we have both averages to the cent:
| Distance between the final and the price to beat | Share of rounds |
|---|---|
| under $10 | 9.3% |
| under $25 | 23.9% |
| median distance | $58.88 |
Almost one round in ten is decided by less than ten dollars on a ~$63,000 asset, and one in four by less than twenty-five. Over these 32 days Up won 48.6% of the 9,197 rounds, essentially a coin flip.
Because of the 60-second average, the outcome stays open late. We project the result second by second; one second before the end, the projection is within $1.47 of the final in 99% of rounds. That still leaves about 1.6% of rounds closer than the projection's own error. Those rounds genuinely cannot be called until they end.
Who trades them
A median Bitcoin 5-minute round draws $22,400 of volume, 831 wallets and 1,430 trades. Over the same 32 days the series traded $230 million. The 15-minute Bitcoin rounds are about half that size per round; Ethereum, Solana and XRP rounds are much smaller (a median 15-minute XRP round trades a few hundred dollars).
That crowd is mixed: market-making bots that quote both sides of the book, directional bots that trade one side automatically, and a large number of people trading by hand. The live page shows the split by wallet type for each round as it happens.
The win-rate trap, in its purest form
Here is the part that surprised us. We took every wallet that played at least 20 Up/Down rounds in the last 30 days and left out the market-making bots: 22,827 wallets.
- Only 22.6% of them made money over the 30 days.
- 45% of them won 55% or more of their rounds.
- Of those high-win-rate wallets, 73% lost money.
How do you win most of your rounds and still lose? By what you pay for them. The median entry price of the high-win-rate wallets is 64¢; the median entry of everyone else is around 33¢. Buying the side that is already favoured wins often, and a share bought at 64¢ makes 36¢ when it wins and loses 64¢ when it does not. At that price you need to win 64% of the time just to break even, before fees.
At the extreme it looks like this. Real wallets from the same 30 days:
| Rounds | Won | Average entry | 30-day P&L | Volume |
|---|---|---|---|---|
| 2,940 | 99.0% | 98.8¢ | −$3,664 | $7.4M |
| 3,888 | 79.2% | 84.6¢ | −$76,609 | $3.2M |
| 6,881 | 98.4% | 98.7¢ | −$3,832 | $2.0M |
| 333 | 78.4% | 78.3¢ | −$66,268 | $0.9M |
A 99% win rate, nearly three thousand rounds, seven million dollars traded, and a loss. Each win pays about a cent per share; one loss costs ninety-nine. It is the same mechanic we described in why a 99% win rate is a trap, compressed into five-minute rounds where it plays out hundreds of times a week.
Taken together, the profitable wallets of the group made $2.76 million in 30 days; the losing ones lost $7.99 million. Of the net $5.2 million, $3.55 million is the trading fees these wallets paid; most of the rest went to the market-making bots, which made about $1.5 million on the spread before their own fees.
What to look at instead
If win rate misleads, what tells you something about a wallet in these markets?
- Edge, or P&L per dollar traded. It folds the entry price in. A wallet that wins 52% of its rounds while buying at 40¢ is doing far better than one that wins 79% at 85¢.
- Entry price and timing. Whether a wallet took its side before the round started or chased the price in the last minute says more than how often it was right.
- Size and consistency. Thirty days and dozens of rounds, not one lucky afternoon.
The Crypto Up/Down page shows these live for every wallet in the current round. Each wallet's number of rounds sits next to its trades, and its 30-day Edge and win rate come with a Pro plan (the first three rows of the wallets table are open to everyone). You also see who was already in before the start, how the smart-money side is leaning, and a second-by-second projection of the 60-second average. Big trades on Binance and Coinbase are marked on the chart.
How we counted
- Data: every Polymarket fill in these markets, read from the Polygon blockchain; Chainlink prices from the feed Polymarket resolves on; Polymarket's market data to cross-check the result of each round.
- Window: Bitcoin 5-minute rounds from 19 October to 20 November 2026 (9,197 resolved rounds); prices to the cent from 17 November (871 rounds).
- Wallet numbers: the 30 days to 20 November 2026, wallets with 20 or more Up/Down rounds across all series. P&L counts each round at its resolution price and includes trading fees. Market-making bots are left out, so a two-sided quoting strategy does not blur the picture.
- What this is not: a trading signal. OrcaLayer is an analytics tool and does not place trades. Past rounds tell you how these markets behave, not which way the next one goes.