Historical data
What data a backtest uses, how deep it goes, and where the edges are.
The window
| Earliest available | 1 January 2024, 00:00 UTC |
| Latest | Now, floored to the top of the current hour |
| Same for | Every timeframe and every asset class |
Earlier docs stated "Equity: up to 2 years of daily OHLCV; Crypto: up to 1 year of 1-minute OHLCV." Both figures were wrong and the asymmetry did not exist.
Both asset classes use the same 1 January 2024 floor, giving the same depth, currently about two and a half years and growing, identical across asset classes and timeframes.
Resolution
The platform stores and fetches 1-minute OHLCV and the engine resamples upward. That is why arbitrary timeframes work, 2m, 7m, 45m, 4h are all valid, not just a curated preset list. Any well-formed interval from 1 minute to 1 day is accepted.
Two consequences:
Resampling boundaries are the engine's. A 45-minute candle's boundaries are computed by the resampler and may not align with how your broker's chart draws them. Comparing a signal timestamp against a chart on an unusual timeframe can show an apparent discrepancy that is a boundary difference, not an error.
Deep 1-minute series are expensive to fetch. A 2.5-year 1-minute series takes 30 to 50 seconds to fetch, which is why the fetch cache exists and why the window's end is floored to the hour. See Choosing a test period.
Fields
Open, high, low, close, volume, per bar, per instrument.
Volume is not always present or usable. The engine checks whether a series carries usable volume before evaluating a volume-dependent indicator, and there are volume-free fallbacks for specific studies. A strategy conditioned on volume for an instrument with no usable volume series is a strategy whose condition cannot be evaluated, worth knowing before you build one.
Warm-up
An indicator produces no value until it has enough bars. A 200-period moving average has nothing before the 200th bar.
With a stated window, the fetch start is moved earlier than your simulation start by a computed padding, so indicators are warm when the tested period begins. The padding depends on timeframe, the indicators used and the objective, and it is clamped to the 1 January 2024 floor.
With the default window, data is fetched from the configured start as-is; the window is long enough to absorb warm-up at the front.
The edge case: a stated window starting close to the floor cannot get full padding. A 200-period daily indicator on a window starting 15 January 2024 has 15 days to warm on. That does not error. It produces a strategy whose entry cannot be true early in the window.
Instrument coverage
| Asset class | Coverage |
|---|---|
| Indian equity cash | NSE and BSE instruments in the platform universe |
| Crypto spot | Pairs on supported exchanges |
| Indian F&O | Index and stock futures and options, with contract series stitched across expiries |
| US equity | Seeded, pending activation |
| Other equity venues | Venues and calendars exist; no instruments yet |
F&O series are stitched across contract months so a continuous history exists for a series whose individual contracts expire. That stitching is a modelling choice, and roll effects at the seams are a real characteristic of any continuous futures series.
Per-instrument detail: Asset Explorer and Instruments.
Alternative and fundamental data
Beyond price, the platform carries a custom indicator catalogue covering fundamentals,
valuation, corporate events, news, economic and on-chain data, addressed by GROUP:KEY codes.
Availability is per tenant and per entitlement, and history depth varies by feed and is generally shallower than the price series. A fundamental metric with quarterly frequency gives you a handful of observations over the backtest window, which bounds what a rule conditioned on it can demonstrate.
See Indicators.
Data quality
| Consideration | What to know |
|---|---|
| Corporate actions | A corporate-actions service handles splits, dividends, mergers, spin-offs, symbol changes, redemptions and coupons. Check surprising equity results around known action dates |
| Gaps and missing bars | Handled by forward-fill where appropriate. A series with substantial gaps produces a less reliable simulation |
| Symbol changes | Tracked as corporate-action events |
| Survivorship | The universe is current. Delisted instruments are absent |
| Freshness of the latest bar | An open-ended window ends at the top of the current hour, so the most recent partial candle is not included |