Why most trading strategies die

A strategy "working" on a chart you're looking at, and a strategy that will make you money next year, are two completely different things. Here's the gap between them, in five parts.

This is the single biggest killer. A strategy is built and tuned during one kind of market — say a long, smooth bull run — and it looks incredible, because it was effectively shaped around exactly what happened. Then the market changes character: it starts chopping sideways, or crashes, or grinds down. And the strategy quietly stops working, because the conditions it depended on are gone.

Traders call this regime fit. A "regime" is just the current mood of the market — trending, ranging, calm, violent. An edge that only works in one regime isn't really an edge; it's a bet that the weather never changes. The classic version is a system that made a fortune in a bull run and then handed it all back the moment the bull ended, having learned nothing except "the last few years went up."

This is why we hold back the most recent chunk of history and test on data the strategy has never seen. A strategy that only shines on the years it was built around has told you nothing about tomorrow.

Give anyone enough dials — which moving average, which threshold, which time-of-day — and they can make a backtest look magnificent. Not because they found something real, but because they kept adjusting the settings until the numbers fit the past. That's curve-fitting (also called overfitting): fitting your strategy to the random noise of history instead of to a genuine pattern.

The tell is brutal and simple. A curve-fitted strategy looks better than an honest one on the data it was tuned on, and worse on data it wasn't. We built a page that shows exactly this: two versions of the same idea, one honestly built and one tuned over thousands of settings for the perfect fit. In-sample the tuned one looks twice as good. Out-of-sample, the boring honest one wins. See the two curves for yourself →

Some strategies genuinely do have a small edge. Then you subtract the cost of trading — the fee on every entry and exit, plus slippage (the gap between the price you wanted and the price you actually got) — and the edge vanishes. A strategy that makes a tiny amount per trade but trades constantly is especially exposed: the more it trades, the more the fees eat.

We've tested patterns that were real and robust in a raw backtest, and still died the moment realistic costs were applied — the edge was smaller than the toll. That's why every result here is charged 0.06% on every entry and exit, and compared against simply buying and holding. "Made money before costs" is not a strategy. "Beat doing nothing, after costs" is the only bar that matters.

Plenty of famous strategies aren't broken so much as empty. They win on some coins, lose on others, and when you line the results up they're indistinguishable from luck — except you paid fees the whole way for the excitement. This is the most seductive failure of all, because it feels like discipline. You're following a system, taking clean signals, being patient. It just doesn't beat having done nothing.

The Golden Cross is the poster child. Ask a hundred traders for a "safe" signal and most will name it. We backtested it faithfully: it beat buy-and-hold on exactly 5 of 10 coins — a coin flip — while underperforming the "do nothing" benchmark on the biggest of them. Read the autopsy → Across everything we've tested, this "no edge" verdict is the most common result by far.

Even before any of the above, a lot of backtests are quietly cheating in ways that are easy to miss:

Very little — and that's the honest headline of this whole site. Of the strategies we've put through the full test, only a handful cleared the bar of beating "do nothing" across most coins and holding up on data they'd never seen. The survivors aren't money printers either: usually a low win rate and an ugly, patient ride. But they're real, not a bull-market coincidence.