Martingale: why it wins often and when it hurts

The martingale is the oldest idea in trading: if the price falls, buy more, so that your average cost falls with it and a smaller rebound is enough to come out ahead. A9's martingale bot does exactly that, on spot, long only, in cycles.

How a cycle runs

1. The bot buys a first order, an amount you choose.
2. Each time the price falls a set step below the last buy, it buys again, and each new order is larger than the one before (1.6 times by default).
3. When the price rises a set take-profit above the average cost of everything bought, the cycle is closed and a new one starts.

With the defaults, the first step is 7%, and each step after that is 1.5 times wider: 7%, then 10.5%, 15.75%, 23.6% and 35.4%. Five extra buys after the first, so the last one sits about 65% below the first buy. Take-profit is 3% above the average.

Why the win rate is high

Because the average falls with each buy, the bot does not need the price to return to where it started. After two or three rungs, a modest bounce is enough to close in profit. In a market that dips and recovers, most cycles end that way, often within a day or two. A backtest of a martingale typically shows most of its exits closing above cost.

That number is real, and it is also the trap. A high win rate tells you how often the bot wins, not how much it can lose when it does not.

The stuck cycle

The failure is a trend that keeps going. Each rung is bigger than the last, so by the bottom of the ladder most of the cycle's money was spent near the lowest prices. With the defaults, a fully filled ladder has spent about 26 times the first order. If the price has fallen to the last rung, the average cost is still about 40% above it, and the price has to climb back roughly 45% from there before the cycle can close.

Until it does, the bot holds a large position at a loss, and has no more rungs to buy with. Nothing is lost on paper rules alone; the loss is real if the price never returns, or if you stop the bot and sell.

Setting the limits

Three settings decide how deep the bot can go and how much it can spend.

• First order. The base everything scales from. A9 checks the full ladder, not just the first order, against your plan's live limit, so a 10 USDT first order is checked as roughly 260 USDT.
• Number of extra buys and the multiplier. More rungs and a bigger multiplier make cycles close faster after a dip, and make the full ladder much more expensive.
• Step and step widening. Wider steps reach further down with the same rungs. Widening was measured to cut the worst drawdown sharply, which is why it is on by default.

The Fear & Greed gate, when on, stops new cycles from starting while the market is in extreme fear. It never blocks the extra buys of a cycle already running, because refusing the buy that lowers the average would make a bad cycle worse.

Why drawdown is the number to watch

For a martingale, read the maximum drawdown before the return or the win rate. Drawdown is the deepest fall in the account's value from a peak during the test, and for this strategy it is where the stuck cycle shows. Ask whether you could sit through that number in real money, for weeks, without stopping the bot at the bottom. If not, use a smaller first order, fewer rungs, or a different strategy.