Liquidation & leverage · how it works

Mark Price vs Last Price: Which One Liquidates You, and Which Should Trigger Your Stop

Two traders, one candle. The first watches a wick spear through the level they worked out by hand and keeps the position; the second never sees the last price get there and is closed out anyway. The exchange was reading a different number. Liquidation fires when your maintenance margin ratio falls to 100% or below, and the unrealized PnL inside that ratio is calculated from the mark price; the last traded price is only one of three inputs to it. The stop is your own choice between last, mark and index price. If your stop sits well clear of liquidation and you don't want one wick to take it out, put it on mark. If it has to follow the traded price, put it on last, wicks included.

Illustration: two pointers, one black and one pale, a short distance apart on the same ruler
In a wick the last price goes deeper and the mark price follows only part of the way — the PnL behind a liquidation is calculated from the mark

Three prices on the screen, not two

The number that ticks fastest on a futures order page is the last price: whatever the most recent trade printed at. One large order into a thin book can drag it a long way.

The mark price isn't a trade at all. It is a reference price the exchange calculates. Gate's help center gives its purpose as steadier markets and fewer forced liquidations during abnormal swings, and adds that it is the price unrealized PnL is calculated from. The third number, the index price, is described in Gate's stop-loss guide as an average price from multiple exchanges.

Most of the time the three sit almost on top of each other. When we pulled a BTC_USDT snapshot from Gate's public market API on 10 November 2026, the mark price was 74,617.72, the index 74,652.74 and the last price 74,604.5. Mark and last were 13.22 USDT apart.

So which one liquidates you? Strictly speaking, the trigger is written as a ratio. Gate's liquidation page puts it at the maintenance margin ratio falling to 100% or below, and the unrealized PnL that feeds that ratio is calculated from the mark price. For a long, unrealized PnL = number of contracts × contract multiplier × (mark price − entry price). The mark price is the one pushing you toward liquidation. Check a hand-worked liquidation price against the last price and a mismatch is what you should expect. For how to read the ratio itself, see maintenance margin rate vs margin ratio.

One real 1-minute candle: how far apart they got

The figures below come from Gate's public market API: 1-minute BTC_USDT candles, pulled once for the ordinary contract and once for the mark-price series (we fetched them at 07:13 UTC on 10 November 2026, covering the previous five days). The minute that matters is 21:55 UTC on 8 November 2026.

Time (UTC)Last price lowMark price lowGap
21:5573,42073,703.96283.96 (about 0.39%)
21:5673,557.373,679.32122.02 (about 0.17%)
Both minutes73,42073,679.32259.32 (about 0.35%)

Figures in USDT; percentages use the mark-price low as the base. This is historical API data as returned in November 2026. For live values, go by what the platform shows.

In the 21:55 minute the last price opened at 74,891.3, hit 73,420 and closed at 73,847.8, a drop of about 1.96% from open to low. The mark price opened at 74,900.6 and bottomed at 73,703.96, about 1.6%.

Take a hypothetical long with a liquidation price of 73,600. Across those two minutes the mark price got no lower than 73,679.32, which left 79.32 USDT of room, so the liquidation condition was not met. Now give the same position a stop at 73,600 set to trigger on last price. It fired inside the 21:55 minute, and that minute closed at 73,847.8. Set to trigger on mark price, the same stop stayed put.

There is a second half to this. The mark price did not stand still: over the two minutes it went from 74,900.6 to 73,679.32, about 1.63%. A hypothetical long with its liquidation price at 73,700 was reached by the mark price in the 21:56 minute. What the mark price spared you was the deepest slice of the move and nothing more.

One sample, and no promise about the size of the next gap. The same five-day window holds 7,201 one-minute candles. Measured at each minute's close, the absolute gap between the two prices had a median of 0.0013%, a 95th percentile of 0.0196% and a maximum of 0.1952%. In nearly every minute you could not tell the two apart. The damage is done in the handful that are left.

Why a mark price exists at all

Suppose liquidation read the last traded price directly. When the book is thin, one large order can knock the last price down and let it snap back, and every leveraged position parked near that level gets triggered in a batch. Traders call it a wick. Gate's stop-loss guide describes the mark price as a calculated reference price meant to prevent manipulation.

The cost is an extra conversion: the price you watch and the price your account is valued at are two different numbers.

How Gate calculates the mark price

Gate's help page on mark price calculation takes the median of three prices (we checked it in November 2026):

  • Mark Price = Median(Price 1, Price 2, Latest Traded Price)
  • Price 1 = Index Price × (1 + Funding Rate Basis Rate)
  • Price 2 = Spot Index + Moving Average Basis, where Basis = (Best Bid + Best Ask) / 2 - Index Price

The basis is normally sampled over the past 5 minutes, and the page says the window can shrink to as little as 1 minute when the basis is being pulled apart quickly.

For anyone holding a position, the word that matters is "median". Run it once with round numbers that are purely illustrative, not market data: Price 1 = 74,650, Price 2 = 74,620, and a large order smashes the last price to 73,420. In order, that is 73,420, 74,620, 74,650, and the middle one is 74,620. Push the last price another thousand lower and the median is still 74,620. Moving the last price alone does not move the mark price.

The reverse also follows. For the mark price to come down, at least two of the three have to come down. In the candle above the mark price lost about 1.6% in a minute, so the last price was not the only input falling.

Two limits sit on top of the formula.

The first is a cap on deviation from the spot index. Gate publishes the figure for TradFi contracts, where the mark price is held to within 3% of the spot index; for crypto contracts the page says only that the range is wider and gives no number.

The second is protection against sudden moves: when the mark price swings sharply compared with the past few minutes, it stops updating. The page spells out the rest: order placement carries on during the pause, unrealized PnL stays unchanged for the time being so positions aren't pushed into liquidation, and the mark price resumes once the price comes back on its own, or is eased back gradually if it doesn't. How large a move counts and how long the pause lasts are not stated.

The exchange can change these parameters, so treat the current help page as the reference.

Want to see how far the mark price has to travel before it touches your liquidation point at different leverage? Plug your entry price, leverage and direction into the liquidation price estimator — what it gives you is an estimated level measured against the mark price. The result is for reference only; go by your order page.
Open the liquidation price estimator →

Wicks: why you sometimes get liquidated and sometimes don't

The same wick can end two ways that look contradictory:

  • The last price went through your hand-worked liquidation price and you survived. The wick was in the last price, the median filtered it, and the mark price never reached your level. The 73,600 long above is this case.
  • The last price never seemed to get there and you were liquidated. The mark price falls too; in that candle it dropped about 1.6% in a minute. When the index and the order book are heading down together, none of the inputs is holding the line for you.

A third mismatch shows up in the PnL figure. Gate's page says it plainly: because unrealized PnL is based on the mark price, the figure displayed can differ from what you settle at when you close, and the final PnL follows the actual fill price. The floating number on your position page is at mark; the moment you close by hand is at the trade price.

What the system does once the liquidation condition is met, and where the money goes, is a separate story: liquidation, step by step.

Don't bet on the wickKnowing the mark price filters wicks is no license to open right next to your liquidation price and hope the wick stays out of it. Over those two minutes the last price fell about 1.96% at its deepest and the mark price about 1.63%, a difference of roughly a third of a percentage point. For how much cushion to leave, see futures risk management.

Setting a stop: which price should trigger it

On Gate a take-profit or stop-loss order has three trigger price types: last price, mark price and index price. The help page reminds you that the three are not always identical, so check that the order will trigger on the one you planned around.

TriggerHow Gate's help page describes itWhat it costs you
Last priceThe most recent fill price; suited to standard market conditionsOne wick in a thin book fires it, possibly near the very bottom
Mark priceA calculated reference price to prevent manipulation; suited to volatile or low-liquidity marketsReacts later than the tape, and the book may have moved on by the time it triggers
Index priceAn average price from multiple exchanges; suited to cross-exchange arbitrage strategiesIgnores this platform's own order book, so the trigger won't line up with the chart you're watching

The middle column paraphrases Gate's help page as checked in November 2026; the platform's current page takes precedence. The right-hand column is our own judgment.

Which is better comes down to two things: how far the stop sits from your liquidation price, and which loss bothers you more.

If the stop is well clear of liquidation and the thing you dread is being wicked out and then watching price come straight back, use mark. In the sample above, a last-price stop at 73,600 fired and a mark-price stop at the same level did not. The bill arrives on the other side: in a real decline it triggers a little later, and if it then sends a market order, that order takes whatever the book offers at that moment.

If the stop has to track the price you actually see trading, use last, and accept that wicks count.

If the stop is close to your liquidation price, neither choice rescues much. Gate's help page lists four reasons a stop can fail to place after it triggers: in cross margin, the calculated closing price would fall below the bankruptcy price; the market moves through your trigger so fast that liquidation happens at the same moment; the stop's size is larger than the position; or you reduced the position by hand and left the preset stop bigger than what remains. Its advice is to set the stop trigger far enough from the liquidation price, which it says improves the odds of the order executing.

My own preference comes from the candle above, and Gate does not say this anywhere. A stop on last price next to a liquidation that follows mark price means measuring one distance with two rulers. Within a single minute the two lows were more than 280 USDT apart, so the real gap between stop and liquidation wobbles with them. Put the stop on mark and both ends are on the same ruler. For trades with a wide stop I lean toward mark. As for index price, I can't see a reason to pick it for a plain one-directional position.

One more thing: triggering is not the same as filling at the trigger price. If the stop goes out as a market order, Gate's page explains, a large order may have to be matched across several price levels in the book. That is slippage, and it leaves the average fill away from the last price on the chart. The page adds that the execution price can differ from the trigger price, especially in high volatility. These orders are reduce-only, so they won't flip you into the opposite position.

If your fill came in well away from the trigger, stop-loss didn't trigger, or filled far off walks through the checks, and how to set a stop-loss covers placing the order itself. An entry order that triggered but never filled is a different problem: why a triggered conditional order didn't fill.

Four things to check before and after the order

In practice

1. Measure distance with the mark price. "How close am I to liquidation" is the distance from the mark price to your liquidation price. The last price jumps around enough to scare you on some days and reassure you on others.

2. Leave space between the stop and the liquidation price. Gate's reasoning is about the order having room to execute. Using liquidation as your exit plan gives that room away.

3. Look at which trigger price type is selected before you confirm. The help page doesn't say which one is pre-selected and we won't guess; go by what your order form shows at the time.

4. Review with the mark-price candles next to the ordinary ones. In the public market API the mark series is a separate contract name with a mark_ prefix (mark_BTC_USDT for BTC_USDT). Side by side, it is obvious which price got to your level first.

Risk warningFutures are a high-risk derivative; leverage amplifies losses, and you can lose your entire principal. This article explains a pricing mechanism, not investment advice — it predicts no prices and recommends no actions. Whether to take part is your call based on your own risk tolerance, and you should follow the laws and regulations in your jurisdiction.

FAQ

Which price actually triggers liquidation?

Strictly, no single price does. Gate's liquidation page sets the trigger at the maintenance margin ratio falling to 100% or below. The unrealized PnL inside that ratio is calculated from the mark price, so the mark price is what moves you toward liquidation; the last traded price is only one of the three inputs to the mark price.

Which is better for a stop-loss, mark price or last price?

It depends on how far the stop sits from your liquidation price and which loss you mind more. When there is plenty of room and a single wick is the thing you want to sit out, a mark-price trigger fits; the cost is that it reacts later than the last price and then fills at market, where the fill can differ from the trigger price. A stop that has to follow the traded price fits a last-price trigger, and a wick in a thin book will fire it. Gate also offers an index-price trigger, which its help page associates with cross-exchange arbitrage strategies.

Why didn't a deep wick liquidate me?

The wick was in the last traded price, and the mark price is the median of three prices, so a single outlier gets filtered out. In a BTC_USDT 1-minute candle we pulled from Gate's public market API, the last price fell to 73,420 while the mark price bottomed at 73,703.96; a long with a liquidation price between the two was not closed in that minute. The mark price fell as well, just not as far.

Is it normal for the mark price and last price to differ a lot?

Normally they are almost identical. Across five days of BTC_USDT 1-minute data we pulled in November 2026, the gap at each minute's close had a median of 0.0013%, a 95th percentile of 0.0196% and a maximum of 0.1952%. Inside a minute the lows can be further apart, and we saw about 0.39% in the same window. That is one contract over one window; other coins and other conditions will differ.