API reference

Generated from the live docstrings on every build, so a signature here cannot drift from the one in the code.

Most of what a metric means is written on the metric itself — the field docstrings in metrics.stats state each figure's basis, and Reading the report is the narrative version of the same material.

• topstep_backtest — The public surface. Five names cover almost every script: Backtest, SymbolStrategy, AccountSize, Report and SummaryStats.
• harness — Backtest — the two-line runner that assembles the simulation stack correctly — and Report, which is what run() hands back.
• strategy.symbol — SymbolStrategy: the single-instrument chassis you subclass. Owns indicator updates, the warmup gate, and position bookkeeping.
• strategy.base — The lower-level Strategy protocol and StrategyContext — the seam that lets one class run against the simulator or the live gateway.
• protocols — The structural types both the simulator and the live SDK satisfy. Bar lives here, including what its two timestamps mean.
• indicators — Typed aliases for TA-Lib functions. No formula in this project is reimplemented, so none can drift from the reference.
• metrics.stats — SummaryStats and its three nested blocks. Every field states its basis, because most metrics admit two honest answers.
• tearsheet — report.to_html() / report.show(): the interactive tearsheet — one self-contained HTML file with the tape, every fill marked, the equity curve against the MLL floor, and the text render's stats.
• replay — Backtest(record=True)'s bar-by-bar recording: every decision, event, indicator value and running-stats snapshot — what the tearsheet's replay scrubber steps through.
• metrics.montecarlo — Block-bootstrap the run's own days through the real rule kernel for a pass probability and — more usefully — an autopsy of the failures.
• metrics.confidence — How much to trust the Monte-Carlo number: a double-bootstrap CI, block-length sensitivity, per-year strata, and the cross-check against real windows.
• metrics.windows — Replay a long tape as consecutive independent Combine attempts — the empirical counterpart to the Monte-Carlo, biased the opposite way.
• metrics.walkforward — optimize (which keeps every trial, not just the winner) and anchored walk_forward with an efficiency ratio.
• metrics.overfitting — The guards that make a search defensible: deflated Sharpe against a trial ledger, and PBO by CSCV.
• metrics.economics — Turns a pass probability plus your own prices into an expected value per attempt, and the pass rate at which it turns positive.
• engine.backtest — The deterministic event loop and the frozen BacktestResult it returns. One loop, strict time order, no look-ahead.
• execution.sim_broker — The simulated broker: full order lifecycle, FIFO lots, OCO brackets, and flat-to-flat RoundTrip records.
• rules.kernel — The prop-firm rule engine — trailing MLL, daily loss limit, consistency — enforced in real time, not scored afterwards.
• rules.params — AccountSize and the CombineParams behind it: the cited constants for each account tier.
• fills.bar_fill — The Tier-0 fill model and its one pessimistic intrabar path — the assumption every fill price in a report rests on.
• fills.fees — Per-half-turn commissions and exchange fees. Charged on entry AND exit, which is why so many statistics here are per half-turn.
• data.wrangler — Turn candles into validated Bar streams. Makes you declare whether your timestamps mean the bar's open or its close.
• data.loaders — Load a databento-data-playground Parquet export whose product, bar span and timestamp semantics ride in the file's own metadata.
• data.validator — What a bad bar stream looks like, and which defects are fatal versus merely worth knowing about.
• data.continuous — Stitching quarterly futures contracts into one continuous series without inventing P&L at the roll.
• data.synthetic — Deterministic seeded bars, for tests and for examples that must run without shipping a data file.
• core.instruments — Tick size, tick value and contract specifications — the numbers that convert a price move into dollars.
• core.time — Session boundaries and the trading-day definition. A Topstep day is not a calendar day, and this module is where that lives.
• core.sessions — Asia, London and New York — the regional windows that scope indicator data and tradable hours. A different axis from the Topstep day above.