Martingale

Martingale is a betting rule before it's a trading system: after every loss, double the position size, so the next win recovers everything you've lost plus a profit. Win once and the whole losing streak vanishes from the equity curve as if it never happened. That's the pitch, and it's mathematically true for as long as the account can keep doubling.

The problem is that losing streaks aren't rare enough to make that safe. Five losses in a row means the sixth trade risks 32 times the original size. Eight in a row, and it's 256 times. Win rate looks spectacular right up until it doesn't. A martingale sequence can post winner after winner for months and still be one streak away from wiping the account. See win rate for why that number lies on its own.

Few bots call themselves martingale. Most call themselves grid trading or DCA, and the doubling logic sits underneath, undisclosed, doing the actual work. Any system that adds to a losing position to lower its average entry, instead of cutting the loss, is running some version of the same bet. The tell is always identical: a smooth curve, then one candle that erases a year of gains. That candle shows up in drawdown, not in the win-rate column. That's exactly why win rate alone never catches it.

We tested 76 grid, DCA, and martingale systems across futures, FX, and crypto. Every single one failed. A 100% rejection rate, the only category in the whole audit to hit it. Not most. Not the weak ones. All 76.

That's not a settings problem. Tighter grids, smaller multipliers, earlier cutoffs: none of it changes the shape of the risk. A martingale sequence is short volatility with unlimited downside and a capped upside, and the geometric growth of position size means the account only has to meet one bad streak, ever, to be over. You can delay that streak. You cannot design it away.