BACKTESTING & MODEL DEVELOPMENT

Here is an uncomfortable truth: the majority of traders who learn ICT concepts never systematically backtest them. And the majority of those traders never achieve consistent profitability. This is not a coincidence. REASONS TRADERS SKIP BACKTNew York timeING: 1. Impatience -- they want to trade now, not study historical data 2. Overconfidence -- "I understand the concept, that's enough" 3. Avoidance -- backtesting reveals that their current approach doesn't actually work 4. False belief that demo trading IS backtesting (it is not) WHY SKIPPING BACKTNew York timeING FAILS: • Without backtesting, you do not know your actual win rate • Without knowing your win rate, you cannot size positions correctly • Without correct sizing, even a positive expectancy system destroys accounts through luck-based drawdowns • Without statistical evidence, you abandon valid setups after 3 losses and keep invalid setups because they feel right BACKTNew York timeING IS NOT OPTIONAL: For serious ICT practitioners, backtesting is the equivalent of a pilot's flight simulator hours -- you prove you can navigate the conditions before risking lives. Backtesting proves you can execute the strategy before risking capital.

TOOLS FOR ICT BACKTNew York timeING: OPTION 1 -- TRADINGVIEW REPLAY (FREE, RECOMMENDED): TradingView's Bar Replay feature allows you to go back in time on any instrument and move forward candle by candle, making trading decisions as if you are in real time. This is the gold standard for manual ICT backtesting. Setup: Open TradingView → select your instrument → click the Bar Replay button → go back 6-12 months → replay day by day, marking your analysis before you advance. OPTION 2 -- FOREX TNew York timeER: A dedicated backtesting software that allows tick-by-tick replay with order simulation. More realistic but requires a paid subscription. Excellent for traders who want simulated execution with real spread and slippage data. OPTION 3 -- CHART SCROLLING METHOD: Scroll your chart back in time → cover future price with a box → advance forward candle by candle manually. Cruder than replay tools but accessible on any platform. WHAT YOU NEED TO RECORD: • Date and instrument • Entry price and entry reason (which PD Array, which session, which AMD phase) • Stop loss level and reason • Take profit level and reason (which DOL) • Exit price and exit reason • R:R planned vs R:R achieved • All criteria from your entry checklist (which were present, which were missing) • Outcome: win / loss / breakeven • Notes on execution quality

A backtest record that does not capture sufficient detail is nearly useless for analysis. Here is the minimum required data for each backtested trade: TRADE IDENTIFICATION: • Trade number (sequential) • Date and time of entry • Instrument traded (EURUSD, NAS100, XAUUSD) • Session (London, NY AM) • Direction (Long / Short) SETUP QUALITY: • Setup type (FVG entry, OB entry, BPR entry, Turtle Soup, etc.) • HTF alignment: Monthly direction, Weekly direction, Daily bias • Was price in discount (for longs) or premium (for shorts)? • Killzone: Yes / No • Liquidity sweep present: Yes / No • Entry criteria checklist: which items were confirmed? TRADE MECHANICS: • Entry price • Stop loss price and distance in pips • Planned TP1, TP2 • Initial planned R:R OUTCOME: • Exit price • Actual R:R achieved • Result: Win / Loss / Breakeven • Did price reach TP1 before TP2? Before stop? • Were there any execution errors (entered late, moved stop, exited early)? POST-TRADE NOTES: • What worked about this trade? • What would you do differently? • Confidence level in setup (1-10)

There is no universal number of trades that guarantees statistical significance or proves a trading model is ready for live use. The amount of data required depends on the effect size being tested, variability, the number of filters, and how independent the observations are. Use a documented sample plan, avoid changing rules after reviewing outcomes, and separate rule development from out-of-sample validation. Report trade count, expectancy, uncertainty, drawdown, market coverage and costs; then forward-test before risking material capital.

After completing 100+ backtested trades, analyze the data systematically to build your defined trading model: KEY METRICS TO CALCULATE: WIN RATE: (Winning trades / Total trades) × 100 No universal win-rate target applies across instruments, rules, and costs. AVERAGE R ON WINNERS: Sum of all winner R multiples / Number of winners No universal average-R target applies across strategies. AVERAGE R ON LOSERS: Sum of all loser R multiples / Number of losers Target: should be -1R (full stop hit) EXPECTANCY: (Win rate × Average R winner) - (Loss rate × Average R loser) Estimate expectancy on a representative sample and account for uncertainty and costs; do not assume a universal 0.5R threshold. MAXIMUM CONSECUTIVE LOSSES: The longest losing streak in your 100-trade sample This determines how much drawdown to expect during the inevitable bad runs BNew York time PERFORMING FILTERS: Analyze which criteria, when present, correlated with the highest win rates: • Setup type: which PD Array performed best? • Session: London or NY AM? • HTF alignment: did 3-timeframe alignment outperform 2-timeframe? • AMD phase: did distribution phase entries outperform manipulation entries? MODIFICATIONS BASED ON DATA: If certain criteria consistently appear in losers but rarely in winners -- add them as EXCLUSION filters. If certain criteria consistently appear in winners -- make them REQUIRED. Build the final model from what the data tells you, not what feels right.

After defining your model from backtesting, the next step is forward testing on a demo account -- trading the model in real time on paper money to validate that it performs similarly in live market conditions: WHY FORWARD TNew York timeING MATTERS: Backtesting, even done perfectly, has limitations: • You know roughly what price does in the data you are reviewing • Emotional pressure is absent • Execution is hypothetical Forward testing on demo introduces: • Real-time uncertainty -- you do not know what price will do • Real-time emotional pressure -- even on paper money, decisions feel more real • Real execution practice -- actually placing orders, managing positions FORWARD TNew York time REQUIREMENTS: • Trade the EXACT same rules from your backtest model -- no modifications • Record every triggered trade with the same detail as backtesting • Minimum 50 forward-tested trades before considering live trading • Target: 50 consecutive trades following ALL rules of your model (including staying out when conditions are not met) WHEN FORWARD TNew York time RESULTS MATCH BACKTNew York time: Similar backtest and forward-test win rates alone do not prove robustness or readiness for live trading. Compare expectancy, drawdown, trade frequency, costs, and rule adherence on an independent sample. WHEN THEY DON'T MATCH: If forward test results are significantly worse than backtest -- either your backtest had replay bias, or there are real-time conditions that affect the model that were not present in historical data. Identify the discrepancies and fix them before going live.