Published
My liquidation price at 100x was $99,400. My chart was still showing $99,899 -- half a percent above it -- when the position got liquidated anyway. Nothing shady happened. No exchange glitch, no manipulation. I just didn't understand that the number deciding my liquidation was never the number I was staring at.
Two prices, one account
Every perpetual exchange runs two separate prices for the same contract. Last price is whatever the most recent trade printed on that exchange's own order book -- it's the number your candles are built from. Mark price is a smoothed reference price, usually the index price (an average of spot prices pulled from several exchanges) adjusted for the funding basis. Liquidations trigger off mark price, not last price. That's not a quirk, it's deliberate: if liquidations fired off last price, one thin, manipulative, or just unlucky print on a single order book could cascade-liquidate every leveraged position sitting near that level. Mark price is built to be harder to move with one order. Good for market stability. Bad for anyone who assumed the price on their chart was the number that mattered.
The buffer shrinks faster than people expect
Ignoring fees, a long position's liquidation buffer -- how far price can fall before you're liquidated -- is roughly 1/leverage β maintenance margin rate. At a typical 0.4% maintenance margin rate on a major pair:
| Leverage | Liquidation buffer | On $100,000 entry |
|---|---|---|
| 5x | 19.60% | $19,600 |
| 10x | 9.60% | $9,600 |
| 20x | 4.60% | $4,600 |
| 50x | 1.60% | $1,600 |
| 100x | 0.60% | $600 |
That part isn't news to anyone who's used a liquidation calculator. What I actually wanted to know: how much of that shrinking buffer does a completely ordinary mark-vs-last gap eat, before price has even moved the "expected" amount?
What a normal gap does to each buffer
Mark-index gaps of 0.1%-0.5% aren't exotic -- they show up during funding rate spikes, thin overnight liquidity, or just normal basis drift between the index constituents and one exchange's order book. I ran the same gap sizes against every leverage tier above:
| Leverage | Buffer | 0.1% gap eats | 0.3% gap eats | 0.5% gap eats |
|---|---|---|---|---|
| 10x | 9.60% | 1.0% | 3.1% | 5.2% |
| 20x | 4.60% | 2.2% | 6.5% | 10.9% |
| 50x | 1.60% | 6.2% | 18.8% | 31.2% |
| 100x | 0.60% | 16.7% | 50.0% | 83.3% |
At 10x, even a fat 0.5% gap only eats about 5% of the buffer -- background noise. At 100x, that exact same 0.5% gap eats 83% of the entire buffer. The gap size doesn't grow with leverage. The buffer shrinks with it. So the same real-world noise becomes a completely different fraction of what's protecting you, and nobody adjusts their mental model for that when they slide the leverage toggle up.
What the chart shows while it's happening
On a $100,000 BTC entry, here's what your candles still display at the exact moment mark price has already crossed your liquidation level:
| Leverage | True liquidation price | Chart (last) price at that moment, 0.5% gap |
|---|---|---|
| 10x | $90,400 | $90,854 |
| 50x | $98,400 | $98,894 |
| 100x | $99,400 | $99,899 |
At 100x, your chart is showing a price a full $499 above where you'd naively expect to get liquidated -- and you're already gone. That's not slippage, not a stop-loss missing its fill, not an exchange being slow. It's a different price feed doing exactly what it's designed to do.
None of this means mark price is unfair -- it exists specifically to stop the last-price-manipulation version of this problem, which is worse. It means leverage doesn't just compress how much room price has to move against you. It compresses how much room your liquidation trigger has to tolerate completely normal, unavoidable noise between two legitimate price feeds. Above roughly 20x-50x on a major pair, that noise floor stops being a rounding error and starts being a real chunk of your entire safety margin. I don't run size above 20x anymore for exactly this reason -- not because I'm afraid of the price moving, but because at higher leverage I'm no longer just betting against the market, I'm betting against basis noise I can't see on any chart. Model your own leverage, maintenance tier, and gap assumptions on the mark price vs last price liquidation gap calculator.