On this page11 sections
The short answer
Polymarket opens a new Bitcoin Up or Down market every five minutes and every fifteen. They are the busiest books on the platform, and most of the trading is bots. Our minute-markets guide explains how they work and why the fee matters. This study asks a simpler question: over one week, who actually ended up with the money?
We pulled every fill in every BTC 5- and 15-minute window from November 1 to November 7, 2026: both sides of 10.9 million fills in 2,654 windows. Then we worked out each wallet's result in each window.
- 70% of wallets lost. 23,104 wallets traded. 7,033 made money and 16,071 lost.
- The prices were right. A taker buy at 58¢ won 58.3% of the time. At 98¢ it won 98.3%. Before fees, buyers at every price above 10¢ came out within two percent of break-even.
- The fee decided the week. Takers paid $1.19M in fees. Wallets that mostly take lost $952K, and $808K of that was fees.
- The winners quote all day. The 407 wallets that traded more than 500 windows and rested most of their orders made $270K. They did 52% of all fills.
- The wallets on Polymarket's leaderboards won. Nobody else did. The 96 wallets from our tracked set made $528K. The other 23,008 lost $1.48M.
The week in numbers
| 5-minute | 15-minute | Both | |
|---|---|---|---|
| Windows | 1,991 | 663 | 2,654 |
| Fills | 8.87M | 2.04M | 10.91M |
| Dollars changing hands | $50.1M | $11.7M | $61.8M |
| Taker fees | $935K | $251K | $1.19M |
| Maker rebates (20% of fees) | $187K | $50K | $237K |
| Windows that resolved Up | 979 | 313 | 1,292 |
The trades themselves are zero-sum. Every dollar one wallet wins at resolution, another loses. We checked it: in every window, the wallets' results add up to minus the taker fees, to the cent. So all traders together were down by exactly what the fee took out: $1.19M paid, $237K returned to makers, $949K gone. Some of the rest comes back later as taker-fee rebates, which aren't in this data.
The median wallet traded $91 across 10 windows. Most of the 23,104 are people trying a few windows. Most of the money is bots.
The prices were right
The first question is whether anyone could win by picking a side. We valued every buy at resolution, $1 a share if its side won and nothing if it lost, and grouped the buys by price:
| Taker buys at | Money spent | Average price | Won | Before fees | Fee | After fees |
|---|---|---|---|---|---|---|
| 1–10¢ | $0.47M | 2.4¢ | 2.3% | −5.0% | 6.6% | −11.6% |
| 10–30¢ | $3.10M | 20.3¢ | 20.3% | 0.0% | 5.5% | −5.5% |
| 30–50¢ | $7.72M | 40.3¢ | 41.0% | +1.9% | 4.1% | −2.3% |
| 50–70¢ | $12.38M | 58.0¢ | 58.3% | +0.5% | 2.9% | −2.4% |
| 70–90¢ | $10.96M | 79.8¢ | 78.9% | −1.2% | 1.4% | −2.5% |
| 90–95¢ | $4.39M | 92.3¢ | 92.4% | +0.1% | 0.5% | −0.4% |
| 95¢ and up | $12.54M | 97.9¢ | 98.3% | +0.3% | 0.1% | +0.2% |
"Won" is the share of shares bought whose side won. Before fees, every band above 10¢ is within two percent of break-even, and the cheapest band, lottery tickets at 2¢, lost 5%. These books price a coin flip on Bitcoin's next few minutes about as well as it can be priced. That leaves a taker paying the fee for nothing. Between 30¢ and 70¢, where most of the money trades, the fee is about 3–4% of what the buyer spends. Our fee guide has the curve.
Resting buys tell the same story without the fee. Makers bought $55M worth, and every band but the two cheapest came within about 1.3% of break-even. Their 10–30¢ bids won a little more often than their price: 20.4% at 19.3¢.
When in the window
| Taker buys made | Money spent | Before fees | After fees |
|---|---|---|---|
| More than 10 minutes before the close (15m only) | $2.38M | +0.4% | −2.6% |
| 5–10 minutes before | $5.61M | +0.7% | −2.2% |
| 2–5 minutes before | $24.97M | −0.1% | −2.6% |
| 1–2 minutes before | $11.10M | +0.8% | −0.5% |
| In the last minute | $7.47M | 0.0% | −1.1% |
Buying late is cheaper, not smarter. In the last minutes, prices move away from 50¢ toward 0 or $1, and the fee shrinks with them. Nothing before fees shows that late takers knew more. The sniping guide covers why the race at the close is for speed, not information. Our study of buying at 90¢ and up looks at the same trade across all of Polymarket.
Takers paid, makers collected
We split wallets by how many of their fills rested on the book:
| Wallets that | Wallets | Made money | Result | Fees paid | Maker rebates |
|---|---|---|---|---|---|
| Rest 70% or more of their fills | 3,549 | 35% | +$292K | $160K | $218K |
| Do both | 1,451 | 37% | −$289K | $219K | $16K |
| Take 70% or more of their fills | 18,104 | 29% | −$952K | $808K | $3K |
Even the makers' total is mostly rebates. Their trading made $74K after the fees some of them paid, and the program added $218K. The takers' $952K loss was $144K before fees. Even if every taker got half of its fees back, the top rebate rate, they would still be $548K down.
How often they trade
| Wallets that traded | Wallets | Made money | Result |
|---|---|---|---|
| 1–5 windows | 8,960 | 32% | −$63K |
| 6–50 windows | 8,336 | 28% | −$67K |
| 51–500 windows | 4,965 | 30% | −$899K |
| More than 500 windows | 843 | 44% | +$79K |
The 51–500 group lost the most, $899K. These are regular traders that aren't full-time quoting bots: most of them take liquidity and pay the fee every time. More than 500 windows is about one window in five for the whole week, which only an automated account does. Split by role, the 407 makers in that group made $270K. The 333 takers in it lost $170K.
The wallets that won every day
2,740 wallets traded on all seven days. 372 of them were up on at least six, and 127 on all seven. Those 372 made $367K together. The steadiest ones are quoting bots:
| Wallet | Result | Of which maker rebates | Fills resting | Traded | Days up |
|---|---|---|---|---|---|
| mo-money | +$30.1K | $2.4K | 98% | $591K | 7 of 7 |
| almach | +$23.7K | $2.3K | 98% | $573K | 6 of 7 |
| bosona | +$23.3K | $2.6K | 98% | $643K | 6 of 7 |
| 0x18c1…2183 | +$20.9K | $9.5K | 99% | $2.23M | 7 of 7 |
| CrouchPredictions | +$16.0K | $6.6K | 88% | $1.65M | 6 of 7 |
| 1000monkeys | +$14.8K | $5.1K | 92% | $1.11M | 7 of 7 |
| ply-pm0 | +$12.6K | $4.7K | 88% | $1.02M | 6 of 7 |
The margins are thin: 0.9% to 5% of the money traded. The two largest by volume, 0x18c1…2183 and CrouchPredictions, earned close to half of their week from rebates.
Not every consistent winner quotes. Of the 372, 210 take most of their liquidity, and they tend to trade less and win more per trade. Some of the week's biggest single wins were bets, not systems: 76marsh made $41.9K from 46 windows as a pure taker, and was up on two of the four days it traded. A good week for a wallet like that is a result, not yet a record.
The leaderboard wallets against everyone else
96 of the wallets we track from Polymarket's leaderboards traded these windows:
| Wallets | Made money | Result | |
|---|---|---|---|
| Wallets we track from the leaderboards | 96 | 69 | +$528K |
| Everyone else | 23,008 | 6,964 | −$1.48M |
The leaderboard is made of the wallets that win here, so this isn't a surprise. It does show how lopsided the books are: 0.4% of the wallets made $528K while the other 99.6% lost $1.48M. The tracked set isn't all winners, though. uninformedfloww lost $21.0K and paid $15.1K in fees doing it. 0xb55f…64d4 lost $21.2K.
The bots you can buy
Three sellers in our BTC 5- and 15-minute list run their bot on a public account. Their BTC windows that week, before taker-fee rebates:
| Bot | Account | Traded | Taker fees | Result | Days up |
|---|---|---|---|---|---|
| PolymTradeBot | dangerously-skip-permissions | $1.37M | $29.4K | +$1.8K | 3 of 7 |
| GoodFallen | GoodFallen | $0.70M | $14.2K | −$1.0K | 4 of 7 |
| StepAhead | stepahead | $0.40M | $4.7K | +$2.2K | 2 of 6 |
The result includes our estimate of maker rebates. All three were close to break-even, and the fee rebate decides which side of it they land on. The PolymTradeBot account sits at Polymarket's Diamond tier, per our rebates study, so about $12.9K of its $29.4K in fees comes back. A buyer running the same code on a new account starts with no tier and gets none of it back. These accounts also trade other coins and timeframes, which this sample doesn't cover.
PolymTradeBot
Editor's ChoiceAutomated Python bot for Polymarket's 5- and 15-minute Bitcoin & Ethereum up/down markets. Trades live or in paper mode with configurable risk controls, self-hosted from full source.
GoodFallen
Editor's ChoiceSelf-hosted Python bot for Polymarket's 5- and 15-minute crypto Up or Down windows, sold as full source for $399 once and backed by the public Polymarket account its seller runs it on, @GoodFallen.
StepAhead
Editor's ChoiceSelf-hosted Python bot for Polymarket's 5- and 15-minute crypto Up or Down windows, sold as full source for $399 once, or for $899 with installation on your server and a private monitoring dashboard.
What this means for you
- If you trade these windows by hand, the price is already right and the fee is about 3% of your stake near 50¢. Over many windows that is a steady loss, and 70% of the wallets in our week took it.
- If you build a taker bot, test it against the fee and not against the price. An edge of 1% before fees, which would be a very good signal here, still loses 2% at 50¢. Our backtesting guide shows how to model it.
- If you build a maker bot, the rebates are real but small: 20% of the fees in the windows you fill, shared with every other maker. The quoting bots above made 1–5% of what they traded, on hundreds of thousands of dollars a week.
- If you buy a bot, ask for its record in fills and fees, not a profit chart. The sellers' own accounts here were around break-even before rebates, and a new account doesn't get the seller's rebate tier.
- For the last-second sweep at 98–99¢, the edge is a fraction of a percent. It pays only at very large size and very low latency.
BTC 15-Min Bot
Open-source Python bot for Polymarket's 15-minute Bitcoin up/down markets. A 7-phase engine fuses spike detection, Fear & Greed sentiment and cross-exchange divergence via weighted voting.
Uruguabot
Verified PartnerSelf-hosted Python bot for Polymarket's 5-minute BTC and ETH up/down markets, sold with its complete live trading record — losses included.
How to check a window yourself
Every fill in a window is public. Find the market's conditionId from its slug, then pull its trades with both sides:
curl -s "https://gamma-api.polymarket.com/events?slug=btc-updown-15m-1794200400"
curl -s "https://data-api.polymarket.com/v2/trades?condition=0xCONDITION&taker_only=false&limit=500"The slug ends in the window's start time in Unix seconds. taker_only=false returns both sides of every fill: the taker's row and one row for each maker it matched. Run it again with taker_only=true to see which rows are takers. Follow next_cursor until has_more is false.
What this can't tell you
- It is one week. November 1–7, 2026. A week with a big Bitcoin move would look different, and the BTC windows are only part of the crypto markets.
- It is Bitcoin only. ETH, SOL and XRP windows, and the hourly and 4-hour ones, aren't in it.
- Taker-fee rebates are missing. They go up to 50% of fees for the busiest accounts. We showed above that they don't rescue the takers as a group.
- Maker rebates are our estimate. We split 20% of each window's taker fees among its makers by fee-weighted fill size, as Polymarket's docs describe. Actual payouts may differ a little.
- A wallet isn't always one trader. One operator can run many wallets, and one wallet can run several strategies.
Method
- Windows. Every BTC 5-minute (
btc-updown-5m-) and 15-minute (btc-updown-15m-) market that opened from November 1, 2026, 00:00 UTC to November 8, 00:00 UTC. 2,654 of 2,688 had a resolved market. Result from Gamma'soutcomePrices. - Fills. /v2/trades by
condition,taker_only=falsefor every fill andtaker_only=trueto mark the takers, all pages. - Result per wallet and window. Cash from sells, minus cash paid for buys, plus shares of the winning side, minus taker fees. Splits and merges don't appear in the trade feed, and they don't need to: they cancel out at resolution. In every window, the sum over wallets equals minus the taker fees to within one cent, which shows no fills were missed.
- Fees. Each market's
feeSchedule: rate 0.07 × shares × p × (1 − p), takers only. Maker rebates: the market'srebateRate(20%) of its taker fees, split by each maker's fee-weighted filled size. - Buy value. Each buy valued at resolution: $1 a share if its side won, nothing if it lost, whatever the buyer did with the shares later.
- Script.
scripts/minute-market-winners.mjsin our repository does the pull. It is plain Node with no keys.
Frequently asked questions
- Who makes money on Polymarket's 15-minute Bitcoin markets?
- Mostly the bots that quote. In the 2,654 BTC 5- and 15-minute windows of November 1–7, 2026, the 407 wallets that traded more than 500 windows and rested at least 70% of their orders made $270K together. The 18,104 wallets that mostly took liquidity lost $952K. Overall, 7,033 of the 23,104 wallets that traded made money, and 16,071 lost.
- Are Polymarket's Bitcoin Up or Down prices accurate?
- Very close to it, in our week of data. Takers who bought at an average 58¢ saw their side win 58.3% of the time. At 80¢ it won 78.9%, and at 98¢ it won 98.3%. In every band above 10¢, the buyers' result before fees was within two percent of break-even. The fee is what turns that into a loss.
- How much is the fee on Polymarket's 5-minute and 15-minute Bitcoin markets?
- The taker pays 0.07 × shares × p × (1 − p), so 1.75¢ a share at 50¢. In our week that was 2.9% of the money spent on buys between 50¢ and 70¢, 4.1% between 30¢ and 50¢, and 0.14% at 95¢ and up. Makers pay nothing and share 20% of the takers' fees as rebates. Takers paid $1.19M in fees over the week, and $237K of it went back to makers.
- Is it profitable to buy at 95¢ or more in the last minute of a Bitcoin window?
- Only just. Taker buys at 95¢ and up, $12.5M of them, won 98.3% of the time at an average price of 97.9¢, which comes to +0.18% after fees. Resting buys at the same prices made +0.30%. It is close to fair, which is why the bots that do it trade millions of dollars a week for a small total.
- Do the Polymarket bots sold to the public make money in the BTC windows?
- Roughly break-even in our week, before taker-fee rebates. The account behind PolymTradeBot traded $1.37M in the BTC windows, paid $29.4K in taker fees and ended $1.8K up, counting our estimate of its maker rebates. The account behind GoodFallen was $1.0K down, and StepAhead's was $2.2K up. Rebates on those fees decide the result, and they depend on the account's volume tier, which a new buyer doesn't have.