Crypto backtest intrabar data uses smaller time intervals to inspect the sequence inside a larger candle. It can establish that a stop level was reached in an earlier interval than a target, provided the orders were already eligible. Smaller candles still omit the trade-by-trade path within each interval, and neither a candle touch nor a trade print proves that your order would have filled.
The useful question is how much uncertainty disappeared when the data became finer. A test described only as “one-minute precision” leaves out whether those minutes cover every trade, whether both exits occur inside the same minute, and when the strategy was allowed to submit them.
Separate the signal clock from the fill clock
A strategy can make decisions on five-minute closes while a simulator checks its existing orders against one-minute data. Those are different jobs. The smaller candles refine the evidence available to the fill model; they do not automatically authorize the strategy to calculate a fresh signal every minute.
Record three choices separately: the interval used to calculate the signal, the moments when the strategy recalculates, and the data used to simulate eligible orders. If a rule requires the completed 10:00–10:05 candle, it cannot use that final close at 10:02 merely because the replay contains a 10:02 candle. The look-ahead bias audit covers that availability problem in detail.
TradingView provides a concrete engine example. Its broker emulator ordinarily infers a historical path from chart OHLC values; its strategy settings separately control recalculation and order simulation. Its documentation distinguishes realtime tick recalculation from historical behavior, so a setting containing “tick” is not evidence that a full historical trade stream was recovered. TradingView's strategy execution documentation
Ask what changed between two runs. If the finer run also changes signal timing, order activation or strategy parameters, a difference in results cannot be attributed to intrabar data alone.
What five smaller candles can establish
This original hypothetical fixture uses an invented asset quoted in USDT. A long position already exists before 10:00 UTC. A protective stop at 98 and a sell target at 103 are both eligible throughout the interval. Assume the simulation exits the entire position when the first level is reached and immediately disables the other exit. This convention isolates price ordering; actual fill quality remains a separate question.
The five-minute parent candle has open 100, high 104, low 97.5 and close 102. Both exit levels lie within its range. The one-minute children reveal more:
| Minute opening, UTC | Open | High | Low | Close | Information added for the existing exits |
|---|---|---|---|---|---|
| 10:00 | 100 | 101 | 99 | 100.5 | Neither level reached |
| 10:01 | 100.5 | 101 | 97.5 | 98.5 | Stop level reached; target absent from this minute |
| 10:02 | 98.5 | 102 | 98 | 101 | Earlier stop event already governs this illustration |
| 10:03 | 101 | 104 | 100.5 | 103 | Target reached later, after the modeled position closed |
| 10:04 | 103 | 103.5 | 101 | 102 | Does not restore the exited position |
The parent values reconcile: take the first open, highest high, lowest low and last close. The children establish that the stop level was reached in an earlier minute than the target. The parent alone could not establish that ordering.
Notice what the table does not establish. It does not give the exact second of the stop event, the executable bid after the trigger, available quantity or exchange acknowledgment time. Calling this a “stop-first price sequence under the stated activation rule” is more precise than calling it a verified stop fill.
TradingView's Bar Magnifier explanation illustrates how lower-interval observations can change entry and exit ordering within a chart bar. That describes its emulator, not a claim that KLINEO or every other engine uses the same implementation.
The uncertainty moves into the smallest candle
Now change just the 10:01 high from 101 to 104. That minute's range contains both 98 and 103. The lower timeframe no longer establishes which exit level came first. It has located the unresolved event more precisely, without resolving it.
Do not silently assign this trade the same confidence as the previous fixture. Use a declared tie policy, seek finer observations, or retain alternative outcomes. A stop-first assumption is a stress case, not historical proof. The broader backtest evidence guide explains how ambiguous exit conventions can change a report.
Trade records can narrow this remaining price-sequence uncertainty, but inspect what the provider means by a record. Binance Spot exposes separate candle, recent-trade and aggregate-trade endpoints. Its aggregate-trade documentation says qualifying executions from the same taker order, time and price are compressed together. Aggregated trades should therefore not be described as an untouched record of every individual execution. Binance Spot market-data formats
Even a complete public trade sequence omits your hypothetical order's queue position and arrival effects. A print at the limit price does not show how much resting quantity ahead of you consumed the available flow. Keep liquidity and slippage assumptions explicit after improving price chronology.
Measure coverage before comparing returns
Smaller-interval history may cover only part of a test. TradingView currently documents a maximum of 200,000 requested lower-timeframe bars for Bar Magnifier and warns that early chart trades may fall outside that coverage. The relevant lesson for any engine is to report the actual covered interval and fallback behavior; do not transfer that numerical limit to another platform.
Maintain an ambiguity ledger. Start with parent candles containing competing eligible events, then classify each event after inspecting finer data. Here is a second, independent invented example:
| Classification of 20 parent-candle collisions | Count | Interpretation |
|---|---|---|
| Ordered by available child candles | 14 | Competing levels occur in distinct children with valid order activation |
| Both events still inside one child | 3 | Finer data exists, but the price order remains unresolved |
| Child data absent or invalid | 3 | No defensible refinement for that parent interval |
Child-data coverage is 17 / 20 = 85%. Resolved ordering is 14 / 20 = 70% of the original collisions. These denominators answer different questions. Advertising 85% coverage as 85% resolution would overstate the evidence.
Validate child timestamps, interval alignment and OHLC reconciliation before assigning either status. A missing minute should not become an invented flat path. The candle-data quality audit addresses gaps and conflicting records; apply those checks to the child series as well as the parent series.
A replay record another reader can inspect
Keep the same saved strategy and compare parent-only and finer-data runs over the same dates. Attach the source identity, instrument, interval definitions, file checksums, coverage range, order-activation rule and fallback convention. For each changed trade, retain the first child interval that changes its interpretation.
First vary only data granularity. Then test parameter changes separately using the parameter-sensitivity worksheet. Otherwise a new stop distance may remove an ambiguous collision and disguise itself as a data improvement.
Also compare trade counts and exposure time. A finer replay can close a position earlier, free capacity and permit a later entry that the coarse run never takes. The resulting trades are not automatically additional independent evidence. The trade-count and sample-size guide explains why counting trades alone cannot settle that question.
A useful disposition names what remains unresolved: “Fourteen collisions now have ordered price evidence; three remain within-minute ties and three lack valid children.” That sentence supports a concrete next step. Obtain the missing detail or test the unresolved cases under explicit alternatives before interpreting the aggregate performance.
Method and limits
This AI-assisted article synthesizes the linked primary software documentation. The candle fixture and ambiguity ledger are original educational examples; they are not historical market observations, KLINEO results or trade recommendations. Parent aggregation and coverage arithmetic were checked. No exchange replay or live fill experiment was performed, and finer timestamps cannot certify a strategy's profitability or execution quality.