Platform

Four instruments of review

What each part checks, which data it uses and where it stops. Nothing here trades.

Strategy Review

SimulationSaved, versioned

A report for one strategy: parameters, results under two settings, rule checks and your assumptions, recomputed on the server every time you save.

Inputs: Template (MA crossover, breakout, mean reversion), windows, position size, costs per side.
Output: Eleven metrics per side with formulas; rule checks with first breach and worst value.
Storage: Up to 200 reports; each save creates a new version.
Limits: Not a full backtest: synthetic daily bars, fixed costs, no order book.

Deterministic checks

Four limits written as numbers, checked bar by bar during the simulation, with every breach timestamped by bar.

drawdown_t = (peak_t − equity_t) / peak_t breach if drawdown_t ≥ limit → flatten at close_t, halt

Max position: Static check of position size against the limit.
Drawdown stop: When drawdown from peak reaches the limit: flatten, halt, log the bar.
Single-bar loss: Every bar with an equity change worse than −limit is logged.
Entry frequency: Entries beyond the limit in any 20-bar window are blocked and logged.

Sample data

The same strategy under two settings — regime, costs, or both — with the difference shown for every metric.

Regimes: Steady trend, range-bound, volatility shock, drawdown & recovery.
Costs: Fee and slippage in basis points per fill, set independently for A and B.
Output: A, B and Δ (B − A) for each metric.
Limits: Four synthetic paths are a teaching set, not a sample of real markets.

Append-only

Every save, change and export of a report, in order, with the exact fields that changed.

Recorded: Creation, revisions (field: old → new), recomputations, exports, AI explanations.
Snapshot: Each version stores the full inputs and computed results.
Export: Included in JSON and Markdown exports.
Limits: Per-user trail; there is no multi-user sign-off workflow yet.

Confirm your email, set two scenarios and four limits, save the report. Then change one assumption and read the trail.