Backtesting answers one question honestly: “would these rules have made money in the past?” It cannot see the future — but skipping it guarantees you pay tuition in real money instead. Here is the beginner workflow in MT5, the numbers that matter, and the traps that fake results.
The three ways to test (use them in this order)
- Manual backtest — scroll bar-by-bar through history, write every trade on paper/spreadsheet. Slow, but teaches you the strategy's personality. Best first step for rule-based setups.
- MT5 Strategy Tester — automatic replay on years of data in minutes. Requires rules precise enough to code (or use the visual tester). See our EA install guide for the mechanics.
- Forward test on demo — weeks of live-market paper trading. The final exam: spreads, slippage and news gaps are real here, and this is where curve-fit strategies visibly fall apart.
The numbers that matter (ignore everything else)
- Net profit after costs — the only profit line worth reading.
- Max drawdown — the deepest peak-to-trough loss (drawdown trading guide). If you could not stomach it live, the strategy is unusable — no matter the profit.
- Trade count — under ~100 trades, results are mostly luck. Want statistical meaning? Get 300+ — or draw the whole distribution of outcomes with the Monte Carlo strategy simulator.
- Profit factor — gross wins ÷ gross losses. Above 1.3 after costs is interesting; above 2.0 on trendy data usually means curve-fit.
- Consecutive losses — your emotional stress test. Could you take 11 losses in a row (they happen at 90% win rates too) without abandoning the plan?
The five traps that fake a backtest
- No costs. Always include your real spread + commission. Twice.
- Over-optimization. If you test 500 parameter sets, one will look brilliant by pure chance — that is curve-fitting, not an edge. Optimize on one period, verify on untouched data (“walk-forward”).
- Look-ahead bias. Using today's information inside yesterday's trade (like an indicator that repaints). Fixed-time-bar rules only.
- Tiny samples. 20 trades telling a story is a coincidence with a plot.
- Survivorship in symbols. Testing only on pairs/periods you already know trended is picking the answer first.
Frequently asked questions — Backtesting Basics
How much history do I need?
Ideally 5+ years so the sample includes different regimes (trends, ranges, crises). More history, fewer surprises — but verify the data quality first: free downloads often miss spreads.
My backtest is profitable — can I go live?
Not yet. Walk it forward on untouched data, then 4+ weeks demo (see 🔄 Code conversion guides🛡️ Risk Guard EA setupDo Forex Bots Work?). Live is the last exam, not the first.
Can I backtest manually without coding?
Yes — bar-replay on the chart (MT5's visual mode or any replay tool), a spreadsheet, and brutal honesty about entries/exits. Slower, and you learn far more.
Next steps — Backtesting Basics
📊 See it on a live chart
Theory meets reality: below is the live EUR/USD 1H chart, drawn from the same feeds our tools use. Watch how candles, ranges and momentum look in practice — then try marking structure on the full charts page.
Sources & further reading — Backtesting Basics
Independent references consulted while writing this guide. PipAtlas is educational and not investment advice - verify anything trade-critical with your broker and the source documents.
- Investopedia - Backtesting (definition) - the standard workflow and its pitfalls.
- CFTC - Learn & Protect - why unverifiable performance claims are a red flag.