You are sure Bitcoin goes up this week. Since September, Polymarket lets you play that with 20× leverage: a 1,000 dollar stake moves a 20,000 dollar position. If Bitcoin rises 5 percent, you have doubled your stake.
That sounds like a simple bet: be right, and you win. But a 20× long does not bet on where Bitcoin ends the week. It bets that on the way there, Bitcoin never falls 2.5 percent below your entry. Leverage pays for the destination, but it loses on the road.
At 20×, a 2.5 percent dip is enough
In early September, Polymarket opened its perps to all users outside the US and Canada: perpetual futures on crypto, stocks, indices and commodities, 67 markets at launch. Bitcoin, gold, oil and the S&P 500 go up to 20× leverage, single stocks up to 10×. By mid-October, DefiLlama counted about 91 million dollars in open positions.
When a position gets force-closed is spelled out in Polymarket's own documentation. At entry you post your stake, 5 percent of the position at 20×. The position is force-closed as soon as your account has fallen to half of that, to 2.5 percent of the position. On a 20× market this maintenance line applies to every position, even if you pick less leverage. Once it is hit, the system sells the position at market.
At 20×, then, exactly 2.5 percent of price movement separates entry from liquidation, and fees and funding pull it a little closer. What is left is at most half your stake, minus whatever the forced sale costs in the market. Your position does not see the recovery that follows.
With less leverage the distance grows fast, because the maintenance line stays at 2.5 percent of the position:
| Leverage | Stake as % of position | Liquidated after a drop of |
|---|---|---|
| 20× | 5 % | 2.5 % |
| 10× | 10 % | 7.5 % |
| 5× | 20 % | 17.5 % |
| 3× | 33.3 % | 30.8 % |
Going from 20× to 10× triples the distance. Halving the leverage here means three times the room. How often Bitcoin actually broke these lines is the next section.
A week at 20× mostly means liquidated
Recomputed on Binance Bitcoin daily candles, January 2024 to 19 October 2026: an entry on every day at the close, then a check whether a daily low on the following days touched the liquidation line.
| Leverage | Held 1 day | Held 1 week | Held 30 days |
|---|---|---|---|
| 20× | 25.3 % | 64.2 % | 80.7 % |
| 10× | 1.6 % | 20.3 % | 50.4 % |
| 5× | 0.0 % | 2.2 % | 17.2 % |
| 3× | 0.0 % | 0.0 % | 2.2 % |
Share of entry days after which a Bitcoin long would have been liquidated. 1,016 one-week windows, rule of Polymarket's 20× markets.
At 20×, one long in four was already liquidated after a single day. It lasted a week after only a little over a third of entry days. 20× is not a weekly trade. It is a day trade with a built-in countdown.
Which leaves the question whether this only hits the cases where you were wrong anyway.
Being right does not protect you
Of the 1,016 one-week windows, 535 ended with Bitcoin higher than at entry. Going long in those weeks meant getting the direction right.
| Leverage | Weeks that ended higher | liquidated anyway |
|---|---|---|
| 20× | 535 | 206 (38.5 %) |
| 10× | 535 | 17 (3.2 %) |
| 5× | 535 | 1 (0.2 %) |
At 20×, you were thrown out in more than one week in three in which you were right. Over 30 days it is starker still: 375 of the 564 months that ended higher would have closed the 20× long first, two out of three.
The reason is Bitcoin's normal swing. A 2.5 percent pullback within a week is not an exception for Bitcoin but the rule, including in weeks that finish in profit. At 20× you have to get not only the direction right, but the day.
Anyone who takes this for a quirk of the last two years will find no comfort in the longer history.
Since 2017 it was worse, not better
Across all Bitcoin daily candles since August 2017, a 20× long would have been liquidated within a week after 69.0 % of entry days. Of the weeks that ended higher, 46.5 % were hit.
So the years since 2024 were the calmer stretch. The numbers above are the friendly version.
Shorting does not help, and funding is the smaller problem
"Then I'll just go short." Since 2024, Bitcoin rose 2.5 percent above the entry within a week after 65.6 % of entry days, practically as often as it fell. The swing hits both sides alike.
"And the financing cost?" Perps cost funding, a payment between the long and short side that keeps the contract price at the spot price. Averaged over Binance, Bybit, OKX and Deribit, it ran at about 3.25 percent a year on the position size over the past year. At 20×, that is roughly 1.25 percent of your stake per week. Polymarket settles hourly and caps it at 4 percent per hour. Funding costs you a little over one percent a week. Liquidation costs you half.
What the math leaves out
The calculation uses Binance spot daily lows. Polymarket liquidates on its own mark price, built from several external price feeds. Short spikes on a single exchange can be flatter there, which would lower the rates somewhat. Fees and funding push the other way: both drain the account and move liquidation in front of the 2.5 percent.
The weekly windows overlap, because a new one starts every day. The 1,016 windows are therefore not 1,016 independent trials; the gap between 20×, 10× and 5× does not depend on that. Other exchanges set the maintenance line differently. Binance requires 0.4 percent of the position for Bitcoin positions up to 300,000 dollars. A 20× long there is thrown out only after a 4.6 percent move against it, but by then at most 8 percent of the stake is left. Kalshi, the US venue, goes up to about 6× on Bitcoin and gives its own example: at 6×, a 17 percent move against you can wipe out the whole stake.
Pick the leverage by the holding period
The question before a leveraged trade is not how sure you are of the direction. It is how long you want to hold. Under Polymarket's rule, a Bitcoin long since 2024 survived a week at 5× in 97.8 % of cases, at 10× in four out of five, and at 20× in a little over a third.
How leverage behaves inside a whole strategy, with position size and funding, you can run in the leverage backtest. Why even 2× never doubles your coins but can wipe them out is in How does leverage trading work.
Still open is what the same table looks like for Ether and Solana, which also go up to 20× on Polymarket and swing far more than Bitcoin.
FAQ
What is a perp? A perpetual future is a futures contract with no expiry. You hold a position that moves with the price without owning the coin. To keep the contract price from drifting, the long and short side regularly pay each other funding.
Why exactly 2.5 percent? Polymarket sets the maintenance line at half the collateral a market's maximum leverage requires: 0.5 divided by 20, or 2.5 percent of the position. At 20× you post 5 percent. Once you have lost half of that, the position is closed.
Isn't that really 40× leverage? No. Leverage says how strongly your stake reacts to the price, and that is set by the collateral at entry. With 1,000 dollars at 20×, you hold 20,000 dollars of Bitcoin. A 1 percent drop costs you 200 dollars; at real 40× it would be 400. The maintenance line only decides when it ends: after a 500 dollar loss. At that moment the position is 40 times your remaining account, and that is exactly why it gets closed. A lower line would give you more room, but would also take more of your stake in the end.
Do I get nothing back after a liquidation? In a normal liquidation, Polymarket sells the position at market and the rest stays in your account. That is at most half the collateral, less if the forced sale runs into a thin market. In severe crashes an insurance fund takes over the position together with the posted collateral.
How was it calculated? Binance BTCUSDT, daily candles 17 Aug 2017 to 19 Oct 2026. Entry at the daily close, liquidation as soon as a later daily low reaches the entry minus the distance. Fees and funding are not included. The scripts are on file; a second run gave the same numbers.
Not investment advice, not a recommendation, not a forecast — historical patterns are no guarantee.
Study the Past — Improve your Future 🥋