Every profitable strategy has an edge: a reason it makes money that's more than coincidence. The problem is that random noise, given enough tries, produces gorgeous equity curves. Finding a real edge is less about discovering winners and more about ruthlessly disproving fakes.
The mindset: try to kill it
Treat every promising backtest as guilty until proven innocent. Your job is to attack the strategy from every angle you can think of; whatever survives is worth trading. This is the opposite of the usual retail loop — tweak settings until the curve looks good, then trade it and find out live that you curve-fit.
The battery that separates edge from luck
- Parameter sweeps — test thousands of setting combinations and keep the stable plateau, not the lucky spike. An edge that only works at exactly one setting is noise.
- Walk-forward — optimise on the past, test on the future, roll forward. If it only works in-sample, it isn't an edge.
- Monte-Carlo — reshuffle the trade order into thousands of alternate histories and read the ruin risk. Most "edges" die here.
- Untouched holdout — set aside data the optimiser never sees and replay it once at the end. Pass it and you have evidence; fail it and you dodged a blown account.
- Overfitting gates — deflated Sharpe and probability-of-backtest-overfitting flag strategies that look good only because you tried so many.
Honesty includes saying "no edge"
The most valuable output of real research is sometimes a clean rejection. A strategy that says "momentum intraday: no edge, and here's the proof" just saved you months of losses. One honest no beats thirty polite maybes.
TapeScript runs this entire battery automatically on every version you build, then hands you a verdict in plain numbers. Put your idea through the gauntlet →