Mark Price vs Index Price vs Last Price in Crypto Futures

Instead of comparing screenshots taken minutes apart, we collected all three fields in one public Bybit BTCUSDT linear ticker response. The response timestamp was 7 October 2026 at 21:04:51.427 UTC. It returned a last price of 83,345.80, a mark of 83,345.90 and an index of 83,381.65 USDT per BTC. This is a dated observation for teaching, not a live quote.

In this observation, mark and last differ by 0.10, while mark is 35.75 below index—about 0.0429%. Even this small gap gives each trigger a different reference; the snapshot is not an execution quote.

The ticker also reported size at the best bid and ask, but it is a snapshot rather than a promise of available liquidity later. Someone else can trade or cancel before your order arrives. Last tells you where a trade happened; the order book describes offers that existed when observed. Neither is your personal average fill.

In a quote comparison, keep the settlement currency visible. A 35.75 USDT difference per BTC is not a 35.75% spread, and it is not automatically the profit available on a trade. The percentage here uses the index price as its denominator.

A futures interface can show different prices in different panels without any of them being a typo. A trade-history chart commonly follows last trades. The risk engine values the position using its prescribed reference. A conditional order watches whichever reference was saved in that order. Changing the chart view does not necessarily change the order or the margin calculation.

Bybit documents mark-based valuation for risk and a choice of last, index or mark for TP/SL triggers. A position panel’s headline P&L may use a display convention different from an account-risk panel, so read its label or explanation. Realized trading P&L ultimately depends on the executed entry and exit, plus the relevant fees and funding—not on which chart series you preferred.

In cross or portfolio margin, a displayed liquidation price can be an estimate while the account maintenance ratio is the actual trigger condition. Switching the chart to mark helps inspect the relevant valuation, but does not turn an account containing several positions into an isolated one-position calculation.

Consider a separate invented sequence for a long position. Its last-price stop is set at 50,000 and its isolated mark-based liquidation boundary is 49,700. A burst of selling briefly prints at 49,920 while the index remains 50,090 and mark is 50,050. Only the local traded-price series has crossed the stop at that moment.

The last-triggered condition is met at the local wick. A mark-triggered stop at the same 50,000 has not met its condition yet, nor has an index-triggered one. The position has not reached the stated 49,700 mark boundary. This difference does not require the exchange to ignore an order: the orders are asking different price streams different questions.

The same trigger value can activate at a different moment

At the final point in the illustration, mark is 49,980 and index is 49,960, so either reference has now crossed 50,000. A mark- or index-triggered order could activate then if still active. The earlier last-triggered order would already have been activated; it does not wait for all three prices to agree. The graph is a sequence of reference values, not three simultaneous executions.

A stop triggered by mark still trades against the local order book. If available bids are below mark when the sell order arrives, its fills can be lower than the trigger. Conversely, a brief wick may leave better prices by the time the order executes. Selecting a reference controls when the instruction becomes eligible; it does not select a guaranteed execution price.

Mark can also cross the liquidation boundary before last reaches a nearby stop. Neither reference always leads or lags the other; the wick example illustrates just one possible ordering.

An index combines specified market observations rather than choosing whichever website displays the most attractive Bitcoin price. Those inputs need compatible units and a weighting method. A BTC/USD quote and a BTC/USDT quote are not automatically identical measurements if the stablecoin’s dollar value moves. The index provider’s conversion rules decide how they are made comparable.

For an invented three-market illustration, take equivalent quotes of 50,000, 50,020 and 49,980 with weights of 50%, 30% and 20%. Their weighted result is 50,002. The average is close to all three but is not a trade that necessarily happened on any one venue. It is a reference assembled from those inputs.

Bad or stale inputs need a rule, not blind averaging

If the third feed suddenly reports 45,000 while the other two stay near 50,000, keeping its 20% weight pulls the naive average down to 49,006. That result may reflect a broken feed or an isolated market event rather than a broad repricing. Excluding a feed also changes the remaining weights, so robust construction needs explicit handling of both the abnormal price and the gap it leaves.

Bybit’s current index documentation describes weighted market inputs, abnormal-price filtering and special treatment when data or liquidity is insufficient. It also allows alternative inputs and fallback procedures in exceptional conditions. Thus “index” does not mean an immutable list of spot exchanges that can never change. The relevant question is how the selected contract’s index behaves when a constituent becomes unreliable.

For a trader, the practical consequence is clearest during a price discrepancy. A price on one local venue can move while the broader reference barely changes. That venue’s contracts may execute at the local price even though a risk valuation stays closer to the wider market. Seeing both is more useful than deciding in advance that one is the only “real Bitcoin price.”

Neither definition works everywhere. A mark is a prescribed estimate for valuing the contract, usually with safeguards against isolated prints. Its method can combine an underlying index with information about how the futures market trades relative to that index. The contract can trade at a premium or discount, so copying the index unchanged is not always the chosen valuation.

Bybit’s current explanation uses a median of candidate prices that incorporate the index, funding-related basis, a smoothed futures basis and the last trade. Deribit describes its own bounds and smoothing process and explicitly warns that mark can fall outside the current bid–ask spread. These are exchange-specific methods, not a single universal formula shared by all perpetuals.

Smoothing can reduce a spike without making execution easier

Suppose the order book briefly becomes thin and one small trade prints far away from its neighbors. A valuation method that simply copies that print would instantly mark every open position to an exceptional execution. Using broader inputs or smoothing can reduce that sensitivity. But it does not refill the missing bids or make a large exit possible at the reference price.

A persistent market move is different from one isolated print. If the wider market genuinely reprices, reference valuations must respond as their rules dictate. A mark designed to resist brief distortions is not a floor under losses. Assuming it will keep rejecting adverse movement confuses a valuation safeguard with a guarantee of position survival.

This is also why two venues can show slightly different marks for apparently similar BTC perpetuals. Their underlying indexes, quote currencies, contract demand and valuation methods may differ. For an exchange-to-exchange comparison, match the underlying asset, quote and settlement currency before comparing the two marks.

Choose according to the condition the order is meant to detect. If a trading setup is defined by actual prints on that exchange, last refers to that traded market. If the exit is designed around the valuation used by the margin system, mark is more directly related. An index trigger can express a condition on the broader underlying reference instead of one venue’s latest trade.

These choices create trade-offs rather than a universal safest setting. Last may react to a local wick that broader references do not follow. Mark or index may activate while the local execution market is already elsewhere. Keeping liquidation comfortably separated from the planned exit still matters; changing a trigger selector is not a substitute for that room.

Bybit lets TP and SL reference choices be configured independently. Its price-protection feature also has a specified scope for last-price TP/SL orders. A reader investigating an apparent missed trigger should therefore check the saved order reference and applicable protection, rather than assuming the currently displayed default matches the historical order.

On mobile, review the reference in the saved order after submission, especially if the interface hides advanced fields. A preference used for a previous contract need not describe the next order. Keep the confirmation or order-history record so a later chart comparison uses the setting that was actually submitted.

Compare prices from the same contract and moment. For a missing trigger, identify whether the order watched last, mark or index. For a margin change, start with the mark and eligible account equity. A chart’s last trade is not a substitute for either field.

Index constituents explain a difference between the benchmark and one exchange. Mark construction explains a difference between the index and the margin reference. A stale or missing feed deserves a separate status check rather than a forced comparison with a newer chart.