Backtests
Backtest mode in the Strategy tab runs the simulation engine over historical data using your current draft strategy configuration. It is how you evaluate a change before deploying it.
A backtest is a historical simulation and an approximation — it is not a prediction and not a guarantee of future results. The report is deliberately honest about its own limits (see below).
Running a backtest
- Pick a date window. It defaults to the trailing 30 days.
- The run uses your draft strategy configuration, so configure the strategy first in Tune.
- Submit the run. It moves through queued and running and then shows a report. The panel updates on its own as the run progresses.
Credits
When the credit system is enabled, each run costs credits and your current balance is shown above the form:
- If your balance is short, the form says how many credits the run needs. When top-ups are available, a Buy credits button starts a checkout; otherwise it explains that top-ups are not available right now.
- Credits may be replenished on a recurring basis when that is configured.
- A failed run is automatically refunded — you are not charged credits for a run that did not complete.
When the credit system is not enabled, runs do not cost credits.
The report
A completed report is laid out so nothing is hidden:
- Report warnings first — any honesty caveats about the run (for example, where settlement was approximated) are shown at the top, above every metric.
- Summary tiles — the run can surface figures such as EV per contract, win rate, max drawdown, and net P&L for the simulated window, along with the clip size those totals are based on. A metric the run does not have is simply omitted rather than shown as zero.
- Per-band and per-hour breakdowns — tables of the same figures split by probability band and by hour.
- Raw report — the full underlying report is always available to expand.
A failed run shows its error message and, when a credit was charged, a note that it has been refunded.
Evidence for deploying
Deploying a structural strategy change goes through the evidence gate. A completed backtest of that exact composition satisfies the gate, so the Deploy button shows a "Backtested ✓" note. If a run leaned mostly on approximation, the note says so.
Faster preview
For an instant, free preview of a threshold change — without spending a run — use What-if replay on the canvas.