Level 6 · Lesson 9
Backtesting
Fundamentals
Prove your edge before risking capital. 100 trades. Bar-by-bar. No shortcuts.
First — Why This Matters
🔍 The Crash Test Before the Road
No car manufacturer puts a vehicle on the road without crash testing it first. They slam it into walls, roll it, test the brakes at speed, and simulate every possible failure. Only AFTER it survives every test does it get approved for real drivers.
Your trading strategy is the car. Backtesting is the crash test. Going live without backtesting is like driving an untested prototype at 100mph — you might survive, but the odds are terrible and you will not know why you crashed.
🔎 REAL SCENARIO
Two traders built identical strategies. Trader A went live immediately: blew 3 accounts in 4 months. Trader B backtested 150 trades first, discovered the strategy failed in ranges, added a range filter, re-tested, then went live: profitable for 14 consecutive months. Same strategy. The difference was testing before deploying.
01 — The 100-Trade Rule
Sample Size = Confidence
Watch how confidence grows as trade count increases. At 10 trades, a 70% win rate means nothing. At 100 trades, a 48% win rate is meaningful.
💡 Why 100? At 100 trades, the statistical margin of error drops to roughly ±5%. A measured 48% WR at 100 trades means your true WR is between 43-53% with high confidence. At 30 trades, that range is ±10% — your "55% WR" could actually be 45% or 65%. The gap between meaningful data and noise is the 100-trade threshold.
02 — The Hindsight Trap
What You See vs What Was Real
Left: a static chart where every winning OB looks "obvious." Right: the same chart revealed bar-by-bar — how YOU would have actually experienced it.
03 — The 6-Step Process
How to Backtest Properly
04 — The 5 Biases
Know Your Enemy
Five cognitive biases that corrupt backtests. Know them so you can defeat them.
05 — Backtest Grader
Grade Your Last Backtest
Answer honestly about your most recent backtest. See if it would pass the ATLAS quality standard.
Did you use replay mode (bar-by-bar)?
How many trades?
Did you include ranging conditions?
Did you record every trade?
Rules written before looking at charts?
Did you adjust rules to fit the data?
06 — The Trade Record
What to Record for Every Trade
| Field | Why It Matters |
|---|---|
| Date / Time | Reveals session patterns — do you win more in London vs NY? |
| Setup Type | Model 1 or Model 2 — which model performs better for you? |
| Trigger Used | Which trigger (engulfing, LTF BOS, etc.) has the highest hit rate? |
| Entry Price | Needed for R:R calculation |
| Stop Price | Needed for risk calculation |
| TP1 / TP2 Price | Were your targets realistic? Did price reach them? |
| Actual Exit Price | Where you actually closed — matches your management plan? |
| R:R Achieved | The real outcome — not the planned one |
| Win / Loss / BE | The binary result for win rate calculation |
| Notes | What you learned. What you would change. What the market showed you. |
07 — The 5 Key Numbers
Your Strategy's Report Card
Win Rate — What percentage of trades are winners. Means nothing without R:R context. A 40% WR with 1:3 R:R beats a 60% WR with 1:0.8 R:R.
Average R:R — Average winner size ÷ average loser size. This is the OTHER half of the profitability equation. Higher R:R = fewer wins needed.
Expected Value — EV = (WR × Avg Win) − (LR × Avg Loss). THE number. Positive = edge. Negative = no edge. Zero = breakeven (losing after commissions).
Max Drawdown — Longest losing streak × risk per trade. If 8 losses × 1% = 8% drawdown. Expect 1.5× this in live. Can you survive it?
Trades per Week — How often setups appear. 2/week at £50 EV = £100/week. 10/week at £10 EV = £100/week. Same income, different workload.
08 — Common Mistakes
4 Backtesting Killers
09 — Cheat Sheet
Backtesting Quick Reference
RULES FIRST = Write every rule before touching a chart. If it is not written, it does not exist.
REPLAY MODE = Bar-by-bar. No static charts. Hide the future. Non-negotiable.
100+ TRADES = Minimum sample size. Include trending, ranging, volatile, and quiet conditions.
RECORD ALL = Every trade. No exceptions. Missing trades = biased data.
NO CURVE-FIT = Do not adjust rules to fit historical data. Use standard values. Trust structure.
TRANSITION = After passing: demo or 0.25% risk for 2-4 weeks before full live.
10 — Test Your Understanding
Backtesting Decision Game
5 scenarios. Evaluate backtests and make the right call.
A trader backtests his Model 1 strategy on Gold 15M. He scrolls through 3 months of chart history on a static chart (no replay mode). He finds 45 trades and reports a 62% win rate. Should you trust this result?
11 — Knowledge Check
Final Quiz — 8 Questions
Question 1 of 8
What is the minimum number of trades needed for a statistically reliable backtest?
Question 2 of 8
What is hindsight bias in backtesting?
Question 3 of 8
Why is curve-fitting (optimisation bias) dangerous?
Question 4 of 8
A backtest shows 53% WR with 1:1.5 R:R over 110 trades. What is the EV per £100 risked?
Question 5 of 8
Your backtest shows a maximum losing streak of 8 trades. You risk 1% per trade. What should you prepare for in live trading?
Question 6 of 8
What is selection bias in backtesting?
Question 7 of 8
After 100 backtested trades, what five numbers should you calculate?
Question 8 of 8
What is the best transition from a successful backtest to live trading?