Manual backtesting
Automated backtesting tells you whether a set of rules would have made money. Manual backtesting tells you whether you can trade them. Replay historical forex charts candle by candle, place trades without knowing what comes next, and have every one of them recorded and measured automatically.
These are usually framed as competing options, with manual testing cast as the primitive version. That misreads what each one measures.
"Would these rules have been profitable?"
Fast across thousands of trades, perfectly consistent, and completely unaffected by how you feel. It also requires your strategy to be fully expressible as code — every entry, filter and exit.
"Can I actually execute this?"
Slower and smaller-sample, but it tests the thing that actually happens in live trading: you, reading a chart in real time, making judgement calls under uncertainty. The only honest test of a discretionary edge.
If your strategy depends on context — structure, confluence, "does this look clean" — then a coded backtest is not testing your strategy. It is testing a simplified approximation of it, and the gap between those two is where live results disappear.
Manual testing has a bad reputation for good reasons. All four are fixable, and none of them are fixed by willpower.
Scrolling back, switching timeframe, or glancing at the right edge tells you what happens next. Your 'test' becomes a memory exercise.
In BacktestFlowReplay is clamped to the present edge, and switching timeframe rebuilds the final candle from finer bars so higher timeframes never reveal price that hasn't happened.
Fifty replayed trades later, all you have is an impression — usually a flattering one. The losers blur, the winners stay vivid.
In BacktestFlowReplay will not advance while a close is unjournaled. Every trade is recorded before you can move on, along with a screenshot showing your order lines.
Sizing by feel makes results meaningless — a few oversized winners can carry an edgeless strategy.
In BacktestFlowSize by risk percentage or by lots, priced through each instrument's contract size and quote currency, with position-sizing consistency reported back to you.
Setups you passed on never enter the record, so a strategy looks more selective than it was.
In BacktestFlowMark a setup as a missed opportunity and it stays in the data. The statistics report what untaken setups cost you.
Manual backtesting means replaying historical price data candle by candle and placing trades yourself, as though the market were live, without knowing what happens next. It contrasts with automated backtesting, where a coded set of rules is run over historical data.
Chart replay, bar replay and market replay all describe the same mechanic: hiding future price and stepping a historical chart forward candle by candle. Backtesting is what you do with it — placing trades, recording the results, and measuring whether the strategy has an edge. A trading simulator gives you the replay; a backtesting platform also keeps the record and computes the statistics. The distinction matters, because replay without measurement produces a feeling about your strategy rather than evidence.
They answer different questions. Automated backtesting tests whether a specification would have been profitable, and it is far faster over large samples. Manual backtesting tests whether you can execute a strategy in real time, including the judgement calls a discretionary approach depends on. If your strategy cannot be fully written as code, automated testing cannot validate it.
Yes. Manual replay requires no scripting, no Expert Advisors, and no programming. You step through historical candles and place trades on the chart, which is why it suits discretionary and price-action strategies that resist being expressed as rules.
Look-ahead bias creeps in whenever future price becomes visible — by scrolling forward, switching timeframe, or seeing a chart edge that extends past your replay point. The fix is a replay environment that hard-limits history to the current bar, including when you change timeframe, rather than relying on your own discipline not to look.
No installation. BacktestFlow runs entirely in the browser, so it works the same on macOS, Windows and Linux — there is no download, no plugin, no MetaTrader required, and no broker account needed to start. Mac users in particular tend to hit this problem, because several established backtesting tools are Windows-only and their documented workaround is to run Windows inside a virtual machine.
Estimates for replaying a single year commonly range from around 10 to 40 hours depending on the timeframe, since lower timeframes contain far more candles. Speed controls and keyboard shortcuts reduce that considerably, but manual backtesting is inherently slower than automated testing — the trade-off is that it tests execution, not just a rule set.