Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/backtest-kit/backtest-monorepo-parallel/llms.txt

Use this file to discover all available pages before exploring further.

A frame schema defines the temporal boundaries of a backtest — the start date, end date, and candle interval that the engine uses when replaying historical market data. It is registered via addFrameSchema alongside the exchange and strategy schemas, and is resolved by name when Backtest.background dispatches a parallel runner. Without a registered frame, backtest.ts throws "Frame not specified" before any runners are launched.

​
addFrameSchema

Registers a named backtest time window. 
addFrameSchema({
  frameName: string;
  interval: string;
  startDate: Date;
  endDate: Date;
})

​
Parameters

​
frameName
string
required
Unique identifier for this frame. Must match the frameName passed to Backtest.background(symbol, { frameName }). Also used by listFrameSchema() in backtest.ts to retrieve the registered frame for the --cache pre-warm step.
// Referenced in backtest.ts
const [frameSchema] = await listFrameSchema();
Backtest.background(symbol, {
  exchangeName: exchangeSchema.exchangeName,
  strategyName: strategySchema.strategyName,
  frameName: frameSchema.frameName,
});
​
interval
string
required
Candle interval string. Must be one of the values supported by backtest-kit’s internal Candle.schema.ts interval enum:
ValueDescription
'1m'1 minute
'3m'3 minutes
'5m'5 minutes
'15m'15 minutes
'30m'30 minutes
'1h'1 hour
'2h'2 hours
'4h'4 hours
'6h'6 hours
'8h'8 hours
'1d'1 day
The apr_2026 strategy uses '1m' to maximise signal resolution across the 27-day window. Coarser intervals reduce total tick count at the cost of intra-candle precision.
​
startDate
Date
required
Inclusive start of the backtest window. The engine fetches candles from this timestamp forward. Must be an exact UTC boundary aligned to the chosen interval — e.g. midnight UTC for '1m' or '1d' frames.
startDate: new Date("2026-04-01T00:00:00Z")
​
endDate
Date
required
Inclusive end of the backtest window. Candles at exactly endDate are included in the replay. The --cache pre-warm in backtest.ts passes this value directly to cacheCandles as the to parameter.
endDate: new Date("2026-04-27T00:00:00Z")

​
listFrameSchema

Companion function that retrieves all registered frame schemas at runtime. 
const [frameSchema] = await listFrameSchema();
// frameSchema.frameName  → "apr_2026_frame"
// frameSchema.startDate  → Date("2026-04-01T00:00:00Z")
// frameSchema.endDate    → Date("2026-04-27T00:00:00Z")
// frameSchema.interval   → "1m"
Used in two places inside backtest.ts:
  1. Schema validation — if listFrameSchema() returns an empty array, the module throws before any runners are launched.
  2. Cache pre-warm — CACHE_CANDLES_FN destructures startDate and endDate from the first registered frame schema to call cacheCandles for every symbol.

​
Reference frame: apr_2026_frame

// content/apr_2026.strategy/modules/backtest.module.ts
addFrameSchema({
  frameName: "apr_2026_frame",
  interval: "1m",
  startDate: new Date("2026-04-01T00:00:00Z"),
  endDate: new Date("2026-04-27T00:00:00Z"),
});
The startDate and endDate values must cover the same range you pass to cacheCandles when using --cache. If the frame and the cache window are misaligned, the engine will fall back to live ccxt HTTP fetches for the uncached portion, significantly slowing the hot loop. Always run --cache with the same strategy file that registers your frame before starting the parallel backtest.

​
Replay performance

The 2026-04-01 → 2026-04-27 frame at 1m granularity produces 38 880 candles per symbol (27 days × 1 440 minutes/day). Across all nine default symbols that is approximately 350 000 candle ticks total.
MetricValue
Frame window27 days (2026-04-01 → 2026-04-27)
Candles per symbol38 880 (27 × 1 440 @ 1m)
Total candle ticks (9 symbols)~350 000
Aggregate replay speed~6 326× real-time
Event throughput (single Node process)~103 events/sec
Wall-clock time for full frameOn the order of a few minutes with pre-warmed cache
These numbers were measured on a HP Victus 15-FA1022CI laptop (i5-13420H, 16 GB DDR4-3200, NVMe SSD) with Mongo + Redis running on the same machine via docker-compose. The per-symbol replay speed is approximately 703× real-time; with nine symbols running concurrently in the same Node process the aggregate speed reaches ~6 326× real-time.
The frame window is entirely determined by startDate and endDate — there is no separate “duration” field. To extend or shorten the replay, edit these two values and re-run with --cache to pre-warm the new date range before starting the parallel runners.

Build docs developers (and LLMs) love

Get started for freeTalk to us