Question 1

What is walk-forward testing?

Accepted Answer

Walk-forward testing trains a strategy on one stretch of history, then tests it on the next stretch it has never seen, and rolls that window forward across the whole record. Every number you see is measured on out-of-sample data, so a rule that only worked because it was curve-fit to the past has nowhere to hide. A plain in-sample backtest, tuned and scored on the same data, almost always looks better than the strategy really is.

Question 2

What is the 100-run random-entry gauntlet?

Accepted Answer

The gauntlet keeps a setup's exit rules but replaces its entry timing with random entries, 100 times over, to build a distribution of what luck alone would have returned. A real edge has to beat that random distribution, not just make money. Plenty of setups look profitable only because the market drifted up while they were in, and random timing would have done just as well.

Question 3

Why is a backtest not the same as live returns?

Accepted Answer

A backtest is a model of the past, and live trading adds costs a model rounds away: slippage, partial fills, spreads, and regimes that change after the test window ends. We include fees and slippage in every result, but a backtested return is still an estimate of edge, not a promise of profit. Treat it as evidence a setup is worth running on your own account, not as a forecast.

Question 4

How do you decide whether a setup survived or got busted?

Accepted Answer

A setup survives when it stays positive across the walk-forward periods and beats the random-entry gauntlet on enough real signals to trust, on at least one ticker we ran it on. It gets busted when it cannot clear that bar anywhere, most often because it fires so rarely that even a high score is just a handful of trades, which is a coin flip and not an edge.