See the risk your backtest report leaves out.

Yes. There is no account, no sign-up, no trial, no limit on how many reports you analyse and nothing held back behind a paywall. You are looking at the whole tool.

Roboquant, the platform this tool comes from, is paid. This analyzer is free and separate: it reads a file your platform already produced and does not connect to anything.

No. There is no upload step. The file is read and analysed by JavaScript running on this page — on your machine, in a background worker — so it is never sent to a server, never stored and never written to disk.

The honest cost of that: the results only exist in the open tab. Reloading the results page has nothing to read, and you have to drop the file again.

TradingView: the Strategy Tester “List of Trades” CSV. Comma, semicolon and tab-separated exports all parse, including European number formats such as 1.234,56.

MetaTrader 5 (MT5): a Strategy Tester report saved as XLSX or HTML. Profit, commission and swap are all counted, so the totals match the report's own balance column.

What does not work: a report exported in a language other than English, on either platform. Both parsers look for English column names — and the MetaTrader 5 one for English section names too — so they refuse the file rather than mis-reading it. Switch the platform's language to English and export again.

Legacy binary .xls workbooks do not work either — re-save those as XLSX or HTML.

Because the inputs differ, and every platform picks its own. We compute per-trade returns against a fixed capital basis, then annualise using the real elapsed span of the backtest — not a fixed 252 days and not a count of trades. The risk-free rate is zero unless you are told otherwise.

We also refuse to print the number when the sample is too small for it to mean anything: you get an em dash and a reason instead of a figure that looks authoritative and is not.

It takes your own trades and reorders them: 1,000 random shuffles of the exact same set, plus 1,000 resamples drawn with replacement. For each run it rebuilds the equity curve and records how deep the drawdown went.

The point is that trade ORDER alone changes how survivable a strategy is. It rearranges results that already happened. It does not forecast the market, it does not simulate new trades, and none of it is a prediction.

Every percentage — return, drawdown depth, risk-adjusted ratios — is a fraction of the capital behind the trades. Without a capital figure those numbers have no denominator.

If the report declares its initial deposit we use it. If it does not, we assume 10,000 in the report's currency and say so on screen, on every affected card, so you can mentally rescale. Absolute currency amounts and trade counts are unaffected.

Yes. The analyzer reads a list of closed trades, so the market and the instrument do not matter — if your platform exported it in one of the supported formats, it analyses.

Amounts stay in whatever currency the report used; nothing is converted. Sessions and hour-of-day buckets are shown in the report's own timestamps, which is broker or chart time, not your local time.

No, and there is nowhere to put it. The analysis is derived entirely from the trade list your report contains.

That is also the limit of what this tool can do: it can show how differently the same trades could have played out, but it cannot re-run the strategy on different data or different parameters. Re-testing is what Roboquant does.

They are ours. The sample-size, profit-concentration, decay, direction-bias and win-rate checks use thresholds we chose, and the page says so; they are not published standards and no regulator or exchange endorses them.

Each flag prints the value it measured next to what it would mean, so you can disagree with the threshold and still use the number.