Warmup and readiness
Distinguish available history from a calculation that is ready to support a decision.
Readiness is a contract between your calculation and its inputs. A model object existing, a cache containing rows, or the first chart bar appearing does not establish that a signal is ready.
Strategy-local warmup
A rolling calculation needs enough accepted observations. The SJS crossover example warms its local window, establishes an initial relationship, and only then detects a later crossing. Repeated or invalid observations do not advance its sample count.
A gap can reset that window if the strategy defines such a rule. That reset belongs to the example; it is not a universal behavior of every indicator.
External session warmup
Memory and Polymarket-memory sessions resolve external warmup to zero. Their supplied historical window is the input to replay.
For live sessions, warmup depends on configuration. Pine sessions default to a one-hour external preload unless overridden. Generic warmup processes its observations with inWarmup set, suppressing execution and normal update publication.
That preload can have a different granularity from the later feed. Check the actual initial observations when comparing a live start with a historical run.
Include leading history deliberately
If a backtest needs earlier observations, include them in the requested data or use the calculation's supported history-loading behavior. Extending the start date does not automatically exclude that added period from trading or reported results.
When you need a later evaluation boundary, continue updating calculations during the leading period and explicitly gate entries until the boundary. Define whether positions may cross it.
Check each producer
| Evidence | Readiness question |
|---|---|
| Local rolling window | Are there enough accepted observations under the chosen sampling rule? |
| Configured algorithm | Does its documented readiness contract hold? |
| Historical calculator | Does the returned window have sufficient usable coverage? |
| Pine output | Is the required point available at the expected timestamp? |
Continue with missing and stale data. The backtesting warmup guide explains the effect on historical comparisons.