Crypto futures trigger price types choose which reference activates a conditional order: the latest contract trade, the venue's index, or its mark price. Those references can cross the same threshold at different times. The selected trigger does not set the eventual fill price or guarantee an exit.
A complete stop instruction therefore needs more than a number. “Sell if price falls to 99” leaves the source unnamed. A trade-price wick, an index move and a mark-price move can produce different answers to that condition, even for the same contract.
This AI-assisted article uses official Bybit documentation and an original hypothetical replay. The venue examples explain conditional-order semantics; they do not establish that every exchange or KLINEO connection exposes the same choices.
Give each reference its own job
Last traded price records the most recent completed trade in the contract's market. It tells you where a transaction occurred. It does not promise another order can execute there, especially after the book changes.
An index is a calculated reference, typically assembled from underlying spot-market inputs. Its constituents, weights, conversions and fallback rules matter. Bybit's index-price methodology describes both spot inputs and alternatives when normal data is insufficient. “Index” is therefore a methodology to inspect, rather than a universal, immutable price.
Mark price is another calculated reference with a risk-accounting role. Bybit publishes different rules for perpetual and expiry products, including fallback conditions. Its mark-price methodology also describes temporary divergence from the last trade. Do not assume mark always sits between last and index, or that their separation has a fixed maximum.
For Bybit linear and inverse derivatives, the V5 order API lists LastPrice, IndexPrice and MarkPrice for a conditional order's triggerBy field. It separately specifies direction and threshold. That is a useful model for reading an instruction: source, comparison, number, then the order to activate.
Replay one threshold against three sources
Consider a fictional XYZ/USDT perpetual with a falling conditional threshold of 99.00 USDT per XYZ. Run three independent paper scenarios. Each starts untriggered with the same threshold and resulting order; only the selected reference differs. These are alternative tests, not three live exits submitted against one position.
The following invented observations are the complete source updates in this simplified replay. Assume no extra protection condition blocks activation and all other trigger requirements are satisfied. Real feeds need their actual event sequence.
| Observation | Last trade | Index | Mark | First activation in the three alternative tests |
|---|---|---|---|---|
| A: starting state | 100.00 | 100.00 | 100.00 | None |
| B: local trade drops | 98.80 | 99.50 | 99.40 | Last-price test |
| C: last rebounds, mark declines | 99.60 | 99.10 | 98.95 | Mark-price test |
| D: broader reference declines | 98.70 | 98.90 | 98.80 | Index-price test |
At B, 98.80 ≤ 99.00 is true for last price. The index remains 0.50 above the threshold and the mark remains 0.40 above it. A last-price trigger activates while the other two continue waiting.
At C, the mark condition becomes true because 98.95 ≤ 99.00. The last-price test has already activated at B; its source rebounding to 99.60 does not rewind that event. The index test remains untriggered because 99.10 > 99.00.
At D, the index reaches the condition for its first activation. The three tests used identical numbers and comparisons, yet produced three activation observations. None of those results reveals the fill price, filled quantity or remaining exposure.
This ordering is illustrative. In another path, mark or index can cross first, all three can cross together, or one can remain untouched. A transient local trade can activate a last-price instruction while broader references stay above the boundary. A sustained move can carry every reference through it.
Choose the source that describes the condition
Write the intended event before choosing a field. A rule responding to a local contract trade should say so. A rule responding to a broader index should identify that index. A risk-related exit should be evaluated alongside the account's actual risk mechanism.
These are different specifications, rather than a universal ranking from best to worst. Filtering a local wick may avoid one unwanted activation, but it can also leave an instruction waiting while the tradable market moves. Responding to a local print may activate sooner in one replay and later in another.
The liquidation comparison deserves particular care. Bybit's execution and liquidation FAQ distinguishes isolated-position mark-price thresholds from cross and portfolio account maintenance-margin ratios. The mark-price and liquidation guide explains that risk comparison. Selecting a mark trigger does not guarantee an exit happens before liquidation.
The risk state can also change without changing the trigger source. Position size and tier-dependent maintenance requirements belong in the futures risk-limit-tier review. A source selection cannot substitute for that separate account check.
Keep the trigger source separate from the resulting order
An index-based instruction does not trade on the constituent spot exchanges merely because their inputs contribute to the index. A mark-based instruction does not reserve an execution at mark. Once activated, the resulting order interacts with its own market and terms.
A market instruction can encounter changing liquidity and venue restrictions. A limit instruction can remain unfilled. The stop-market versus stop-limit comparison covers that execution choice; changing the reference does not remove it.
Order configuration can drift during maintenance. A replacement may accidentally retain the threshold while reverting the source, direction or closing semantics. If the resulting instruction is an active limit and its price or quantity later changes, the order-amendment and queue-priority guide explains the separate matching concern. Queue position cannot answer which reference activated the instruction.
Use a source-and-event worksheet
Keep this record with the order configuration and use it again when reviewing an incident. The values in the middle column continue the fictional replay.
| Check | Example record | Evidence to preserve |
|---|---|---|
| Product identity | XYZ/USDT perpetual, named venue and account | Contract and account identifiers |
| Trigger source | Mark price | Saved source field, not chart preference |
| Direction and threshold | Falling to 99.00 | Actual accepted parameters |
| Relevant observation | C: mark 98.95 | Source timestamp and receipt time |
| Other references | Last 99.60; index 99.10 | Separately named observations |
| Resulting instruction | Specified sell order and quantity | Child order identity and terms |
| Final exposure | Unknown until reconciled | Executions and remaining position |
Bybit's ticker response exposes separate last, index and mark fields for linear and inverse products. That makes source naming possible, but a later snapshot cannot prove which update activated an earlier order. Preserve event times, and distinguish an exchange timestamp from the time your application received it.
A chart mismatch is an evidence problem before it is an exchange-error conclusion. Retrieve the matching series: Bybit offers separate index-price candles and mark-price candles. Candle ranges can establish that a level appeared within an interval; they cannot reconstruct every trigger and execution inside it.
Test the awkward cases explicitly
Include a local trade wick that leaves index and mark untouched, a mark crossing while last remains above the threshold, and a rebound after activation. Verify that the replay does not activate twice or infer a fill from the reference value.
Also test missing or stale source data. A frozen local display is not evidence that the exchange's reference stopped updating. Conversely, replacing an unavailable index with the last trade in a local simulation changes the strategy unless the documented methodology and test policy justify it. Keep fallback assumptions visible.
Finally, reconcile what actually executed. The partial-fill guide explains why an activated exit can leave exposure. The operational record should finish with the executed quantity and residual position, alongside the source and event that started the attempt.