A strategy has an edge when it shows positive expectancy - it makes money per trade on average after costs - across a large, representative sample rather than a lucky streak. That is the whole definition. Everything else, from win rate to reward-to-risk, is just an input into that one number.
The number that defines an edge
Expectancy combines your win rate and your average reward-to-risk into the average result per trade. A strategy that wins 40% of the time with 2.5R winners has a strong positive expectancy; one that wins 70% with 0.5R winners may be negative. This is why win rate alone is meaningless - a high win rate can hide a losing system, and a low one can hide a great one.
Why sample size is non-negotiable
A brilliant 15-trade run proves nothing - luck dominates small samples, and even a random strategy will occasionally string wins together. You need at least 100 trades, and more for low win rate strategies whose returns depend on a few large winners. Statistical significance is the difference between "this worked" and "this works."
The three ways a fake edge fools you
- Hindsight in testing: skipping the losers that "obviously" would not have triggered. Take every signal or the number lies.
- Curve fitting: tuning the rules until one stretch of history looks perfect. A fitted edge vanishes on new data.
- Ignoring costs: a positive raw result can turn negative once spread and commission are included, especially for frequent strategies.
Important: the signature of a real edge is not a perfect equity curve - it is robustness. An edge that holds up across market phases, across pairs, and on out-of-sample data it never saw is real. One that collapses when you change a rule or a time period was fitted to the past. Test across conditions on purpose.
Read the whole report, not just the profit
Net profit alone can flatter a fragile strategy. Confirm the edge with the supporting metrics: profit factor above 1 with room to spare, a maximum drawdown you could actually stomach, and a steadily rising equity curve rather than one carried by a single lucky trade. A modest edge you can trust beats a spectacular one you cannot.
Prove it, then forward test
Run your strategy through 100+ trades in the simulator with the future hidden, read the full report, and if it clears the three tests, the last step is to forward test on data it has never seen. An edge that survives backtest and forward test - across conditions - is as close to proof as trading offers.
Trading edge FAQ
What does it mean to have an edge?
A positive expectancy - making money per trade on average after costs - that holds across a large, representative sample rather than a lucky streak.
How many trades prove an edge?
At least 100, and several hundred is better for low win rate strategies. Small samples are dominated by luck, so a great 20-trade run proves little.
How do I know it is not curve fitting?
A real edge survives conditions it was not tuned for - other market phases, other pairs, and out-of-sample data. If small changes collapse it, it was fitted.