Data quality

Backtesting is only as good as the data underneath it. StrategyTune's replay is built on tick-based historical data — real bid/ask quotes, in the order they were recorded — not synthesized intra-bar paths.

What tick-based means here

Most "backtesters" run on completed OHLC candles and fake the intra-bar path with a fixed sequence (open → high → low → close, for example). That's an interpretation, not the market. StrategyTune doesn't do that.

Instead, every momentum bar carries a real bid/ask quote. Between candle closes you see the quotes that actually printed, in the order they printed. Pending orders evaluate against those quotes, in that order. So:

• A stop-loss fills at the moment price first crosses your level — not at the candle's close.
• A limit order at a level inside a bar fills if and only if a real quote crossed that level during the bar.
• Two stops at very close prices fill in the order their levels were actually crossed.

Spreads are real

Bids and asks are independent. Buys fill at the ask, sells at the bid; the spread you'd have crossed in real life is the spread you cross in the simulation. There is no "midpoint fill" mode — the simulation always pays the realistic side.

What's not modeled

StrategyTune simulates execution against a single-source tape. Things that aren't part of the simulation today:

• Liquidity-driven slippage. Beyond the bid/ask spread, the model doesn't simulate market impact. Fills happen at the quoted side at the moment the level was crossed.
• Partial fills. Orders fill in full. There's no "you got 30% of your size" model.
• Multiple venues. Each instrument is replayed against one data source. Differences between venues for the same symbol aren't modeled.
• Commissions and financing. Not currently included in the simulation. Net them off mentally if you're stress-testing low-edge strategies.

Where the data comes from

Each security in the catalog has a named data provider. The /securities page shows the provider per instrument, with the date range of its archive. Providers update at their own cadence; the catalog reflects the latest available.

Because data quality and depth depend on the provider, two instruments with the same ticker on different exchanges or different providers may have different histories. Always confirm the date window on /securities before running a long backtest.

Determinism

Replay is deterministic. The same security, the same date range, the same indicators, and the same trades will always produce the same results — across runs, across devices, and across time. The engine doesn't introduce randomness anywhere; the only stochastic input would be your decisions.

For how this data feeds the simulation, see How replay works.