You backtested a strategy. It looked great. Win rate was solid, profit factor above 2, the equity curve looked beautiful. You went live with real money.
Then you got crushed.
This happens to thousands of new traders every year. And it's almost always because their backtest was contaminated by one of these 5 mistakes. Here's how to spot them — and fix them.
Mistake 1: Hindsight bias (the silent killer)
What it is: Looking at a chart where you can already see what happened, then convincing yourself you would have entered at the perfect spot.
Why it's deadly: Your brain rewards you for being "right" about past moves. So you generously assume you'd have entered, held, and exited perfectly — when in reality you'd have hesitated, taken profits early, or missed the setup entirely.
The example: You see EURUSD pumped 200 pips after a small bullish engulfing pattern in February. Now you scroll back through and find every bullish engulfing pattern that worked. You ignore the 15 that didn't. Result: you "backtest" a 90% win rate strategy that's really 35%.
The fix:
- Use TradingView's bar replay feature — start from a past date and replay bar by bar
- Cover the right side of your screen with paper or your hand
- Mark your entry on the LIVE bar, before knowing what happens next
- Force yourself to take EVERY setup that matches your rules, not just the ones that look good
Mistake 2: Cherry-picking trades
What it is: Only counting trades you "would have taken" — typically the ones that won.
Why it's deadly: Live trading doesn't let you cherry-pick. You'll see borderline setups and have to decide in real time. If you skipped those in your backtest, your live numbers will be brutally worse.
The example: Your rule is "buy when RSI crosses 30 on the daily." You find 50 such setups in your data. 30 of them lead to clean reversals. 20 are messy or fail. You log only the 30 clean ones because "the others didn't really fit." Win rate looks like 70%. Live? You'll take all 50. Real win rate: 40%.
The fix:
- Pre-commit: log EVERY trade that matches your written rules, no exceptions
- If you find yourself thinking "well, that one doesn't really count," your rules aren't specific enough — rewrite them
- Treat ambiguity as a "no trade" — but log that you skipped it
Mistake 3: Ignoring spreads, slippage, and commissions
What it is: Backtesting at the perfect price shown on the chart, without accounting for the costs of actually placing trades.
Why it's deadly: Real spreads can eat 5-15% of your annual return. Slippage during fast moves can turn a 1:2 R:R trade into a 1:1.5 trade. Commissions add up. Backtest profit factor of 1.4 might actually be 1.0 live — break-even, then losing after taxes.
The example: You backtest gold with a 5-pip stop and 10-pip target. Looks like a great 1:2. But gold spreads can be 3-5 pips on volatile days. Real R:R becomes 1:1.4. Your "edge" disappears.
The fix:
- Look up your broker's average spread for the pairs you trade
- Subtract that from every win (or add it to every loss)
- Add 1-3 pips for slippage on stop-outs
- Calculate net profit factor AFTER costs. If it's below 1.2, your edge is too thin to survive live
Mistake 4: Curve-fitting (over-optimization)
What it is: Tweaking your strategy parameters until the past data shows perfect results — but the strategy has zero predictive power on future data.
Why it's deadly: Markets change. A strategy optimized for "RSI 14 crossing 31.5 with MACD signal between -0.0023 and 0.0014" is just a description of what happened before, not a real edge.
The example: You build a strategy with 12 parameters. You test thousands of combinations until you find the magic numbers that produce 200% returns on EURUSD from 2024. Go live. Strategy makes 0% in 2026 because those exact conditions don't repeat.
The fix:
- Keep your strategy SIMPLE — 2-4 conditions max
- Use round numbers (RSI 30, not RSI 31.5)
- Test on TWO separate time periods (e.g., 2024 and 2025). Strategy must work on both
- If small changes to your parameters massively change your results, you've curve-fit
Mistake 5: Sample size too small
What it is: Drawing conclusions from 10-20 trades and assuming the pattern continues.
Why it's deadly: Random chance produces all kinds of streaks in small samples. A losing strategy can win 9 of 10 trades by pure luck. A great strategy can lose 7 of 10 in a slump. You need enough data for the statistics to mean anything.
The example: You backtest a strategy over 15 trades. 11 wins, 4 losses. Looks amazing. You go live. Next 15 trades: 4 wins, 11 losses. The "edge" you thought you found was just variance.
The fix:
- Minimum 30 trades before drawing ANY conclusions
- 50-100 trades for meaningful confidence
- If you can't find that many setups in a year, your strategy is too rare — find one that triggers more often
- Watch for "winning streaks early, losses later" pattern — that's reversion to the mean
Bonus mistake 6: Not journaling DURING the backtest
People backtest 50 trades, see a 60% win rate, and stop there. But the gold isn't in the average — it's in the patterns. Were 80% of the losses on Mondays? Did you lose every time RSI was below 40 at entry? Did wins come from breakouts and losses from pullbacks?
You'll never know unless you write notes during the backtest.
TradeLens lets you add notes to every backtest trade. After 30-50 entries, the AI Journal (Pro) can scan everything and surface patterns — "your losses tend to come from late entries" — that would take you hours to spot manually.
The honest backtest checklist
Before you trust any backtest, can you say yes to all of these?
- ✅ I used bar-by-bar replay (no future visibility)
- ✅ I logged every single setup matching my rules
- ✅ I subtracted realistic spreads and slippage
- ✅ My strategy has 4 or fewer conditions
- ✅ I have at least 30 trades in my sample
- ✅ I tested across two different time periods
- ✅ My strategy still works in different market conditions
If yes to all 7 — you've done it right. Time to demo trade.