The execution specification defines what correct behavior is. This suite is what actually checks the engine against it, built directly from that spec as its source of truth — every fill-price rule, every TIF variant, every bracket-resolution case the spec defines has a corresponding assertion here.
The test suite's script files are available on request with the evaluation kit. It's the execution specification codified into code, so the engine can be verified against it directly.
What it runs
One script, tests/test_execution_engine.py, drives real OHLCV data through reamer_py.run_backtest() across eight configurations and checks the result against explicit invariants — not snapshot comparisons, not mocked engine internals, the actual compiled binary running actual bars:
- Single-asset — all order types, all TIF variants, bracket exits
- Multi-asset (2 tickers) — concurrent positions, shared equity, per-ticker independence
- Multi-asset (4 tickers) — four tickers simultaneously
- High leverage — 10x, position sizing and margin checks
- Commission modes — RoundTrip / OpenOnly / CloseOnly
- Spread / OhlcvType — Bid / Ask / Midpoint fill-price adjustments
- Overnight swap — equity reduced by swap charges across date boundaries
- Alignment (union) — mixed-schedule assets
What a "check" actually is
Each of the 267 is a real assertion against the result of an actual backtest run — things like: total slippage equals the sum of per-trade slippage, cancelled orders carry fill_price=0, IOC orders never remain Pending past the bar they were submitted on, margin rejections fire exactly when the spec says they should, roll events land only on roll-configured tickers. Every one of these ties back to a rule in the execution specification — the suite doesn't define behavior, it verifies conformance to behavior already defined there.
Running it yourself
Three files make up the suite: the runner and invariant checker, plus two strategy files that generate the real order flow being checked — real strategies submitting real orders, not fixture data. All OHLCV data is generated locally by the suite itself, from a fixed seed — no external files, no market data license, nothing beyond a licensed reamer_py build. Pass condition is printed at the end: N/267 checks passed — ALL PASS.
Get it
The suite's script files are available on request as part of an evaluation. A free, time-limited test license is available to evaluate before committing to a seat.