The backtester — what it is and how it's laid out

The backtester — what it is and how it’s laid out

A backtest replays your strategy over market history that already happened. You describe the rules, the engine walks the history tick by tick and trades virtual money. In minutes you learn what live trading would take weeks to show — without risking a cent.

Key property: the backtester runs the very same trading engine that powers your paper and live bots — it is simply fed history instead of the live feed. A strategy you backtested will be executed by a bot with identical logic, down to fill prices and fees.

We trade Polymarket Up/Down markets: a contract’s price is a probability from 0 to 1 (UP at 0.55 = “the market gives a 55% chance of a rise”); at resolution it becomes 1.00 or 0.00.

A backtest doesn’t answer “will I make money?” — it answers “does the strategy have a statistical edge, and does that edge survive fees?” Tuning is a loop: change one parameter → run → compare → conclude.

Section map

“Run / Results / Trades / Validation / Analytics” are five tabs of one run; each has a run switcher at the top (name, status, coin, dates) — the section is preserved when you switch runs.

Statuses and limits

Run lifecycle: Queued → Running → Done / Failed / Cancelled. On completion you get a bell notification (“Backtest finished: N trades, P&L X%, Sharpe Y”) and an auto-redirect to Results.

The “In flight: N / M” badge in the Registry is your limit on concurrent backtests (queued or running) — not a total allowance: a slot frees up when a run finishes. The maximum window per run is 365 days. Backtests also work in demo mode — pure computation, no money involved.

Next: create your first backtest.