Rewind plays history forward one bar at a time, so you make decisions with only what was knowable then — the single discipline that separates a backtest from a story told about a chart.
Most retail "backtesting" is scrolling a chart with the outcome visible. Replay hides the future and advances one bar at a time, which is the only way to test a discretionary rule set honestly. For coded strategies the same history runs in a dedicated worker, so a long run does not freeze the interface.
Fills are charged spread, venue fees and price impact — the identical model used by the live paper terminal and written out in the methodology. That matters because the usual reason a backtest disagrees with reality is not the signal; it is that the test filled everything at mid price for free.
Look-ahead bias, survivorship in the asset list, over-optimisation, and cost fiction. The long version is in backtesting vs paper trading, including the sequence that keeps a test honest: fix the rules first, hold out data, then confirm on live paper before risking anything.
A single in-sample period is a hypothesis with no test. Split the history: fit the rules on one window, confirm on the next window the rules have never seen, and repeat forward. A strategy whose parameters must be re-tuned for every window has not found an edge — it has found a description of each window. Replay makes this practical because you can move the boundary and re-run rather than rebuilding a harness.
Twenty trades is noise. A strategy with a genuine 55% hit rate will still produce long losing runs, and a strategy with no edge at all will produce a beautiful stretch often enough to be convincing. Judge on enough trades that one outlier cannot carry the result, and look at the underwater curve before the return — the drawdown is the part you have to survive to collect anything.
Historical testing sees years in minutes and cannot test you. Live paper trading tests execution at real speed and collects data slowly. Doing both, in that order, is what turns an idea into a measured strategy instead of a hopeful one.
Yes, and it needs no account. It runs in your browser on your own device.
Yes — the same fee, spread and price-impact model as the live paper desk, so the two are directly comparable.
Sharpe, Sortino, maximum drawdown with an underwater curve, realised slippage and the equity curve.
Open the desk — free, no account — no account, nothing to install