Ask a room of experienced traders why most strategies fail and almost none of them will say 'because the entry signal was wrong'. Entries are the most over-studied, least decisive part of trading. Strategies die for structural reasons — and the same four appear again and again.
1. They were curve-fit to the past
Give an optimizer enough parameters and it will find a combination that performed beautifully on historical data — because it is describing that data, not discovering a repeatable edge. The tell-tale signs: many finely tuned parameters, performance that collapses when you shift the test window, and an equity curve that looks suspiciously smooth. A real edge is usually robust to small parameter changes; a curve-fit one shatters.
2. They ignored costs and friction
A strategy that trades often lives or dies by transaction costs. Spreads, commissions, slippage on entry and exit — each may look tiny per trade, but they scale with frequency while your edge doesn't. Plenty of published 'profitable' systems turn negative the moment realistic costs are applied. If your backtest wins by a thin margin, costs aren't a detail; they are the whole story.
3. They were sized to maximize the backtest
Position sizing converts a statistical edge into an experienced outcome. The same rules that compound nicely at 10% of equity per trade can be ruinous at 40%, because losing streaks that are survivable at small size become terminal at large size. Failure usually arrives as a sequence — five, eight, twelve consecutive losses that were always statistically inevitable — hitting an account sized as if they were impossible.
4. The trader could not hold on
Every strategy spends most of its life below its previous equity peak. Drawdown is not a malfunction; it is the operating condition. But humans experience drawdown as evidence the strategy is broken, and abandon systems at their lowest point — often just before recovery, usually to chase whatever performed best recently. The strategy didn't fail; the relationship between trader and strategy failed.
The common thread is that none of these killers show up in a naive backtest summary. They show up when you study drawdown depth and duration, stress-test sequencing with Monte Carlo methods, model costs honestly, and size positions for survival first. That is the work. It is less exciting than hunting the perfect entry — and far more valuable. As always: this is education, not advice, and no amount of testing guarantees future results.