R
Reproducible rules
Another reviewer should be able to identify the same entries, stops, targets, and outcomes.
Lesson · Trade Setup and Execution
Backtesting
Test explicit setup and management rules across chronological historical samples, include realistic costs, and compare in-sample with out-of-sample performance.
Synthetic educational examples
Core concepts
Use these concepts as explicit rules rather than changing them after seeing the outcome.
B
Bias control
Chronological replay, complete samples, and unseen validation reduce look-ahead and selection bias.
S
Statistical distribution
Evaluate expectancy, drawdown, streaks, and regime sensitivity—not only net profit.
Interactive lesson
Freeze the assumptions, run the sample, read the statistics — then change one thing and run it again.
Positive resultNegative resultOut-of-sample segment
Net result—
Expectancy—
Win rate—
Profit factor—
Max drawdown—
Longest loss streak—
In-sample expectancy—
Out-of-sample expectancy—
Backtesting process
Use historical data to evaluate a predefined rule set, not to construct a perfect explanation of the past.
1
Freeze the rules
Document setup, entry, invalidation, target, management, and exclusions before testing.
2
Replay chronologically
Use only information available at each decision point and include every valid setup.
3
Record realistic outcomes
Include costs, gaps, partial fills, missed entries, and management rules.
4
Validate outside the development sample
Test unseen periods, instruments, or regimes before drawing conclusions.
Backtesting quality drill
Score 0 / 0Identify the test design that best limits bias.
Backtest requirements
✓
Rules defined before reviewEntry, invalidation, target, and management must be objective enough to reproduce.
✓
Chronological replayUse only information available at each historical candle.
✓
Costs and missed trades includedSpread, slippage, commissions, gaps, and valid setups not filled affect results.
!
Out-of-sample validationReserve unseen data or a later period to evaluate whether rules generalize.
Common mistakes
1
Selecting only obvious winning examplesExcluding valid losing setups inflates results and destroys representativeness.
2
Using future data to confirm past entriesLook-ahead bias can make impossible execution appear successful.
3
Changing many parameters on one sampleThe best historical combination may be noise rather than a stable relationship.
4
Evaluating only net profitDrawdown, loss streaks, expectancy, costs, and regime dependence affect practical viability.
Completion standard
✓
Finish when you can read equity, drawdown, streak and both expectancies from a run and say which one would make you reject the rules.
Then change exactly one assumption, run again, and say what moved and why — that is the whole method, on synthetic data.
Scope: This lesson teaches process and measurement. It does not provide live market data, personalized recommendations, or trade signals.