mean reversion
Buy the oversold, sell the overbought, expect price to come back to its average. Wins are frequent and small; the losses come from the day it does not come back. The backtest has to fill the stop on the real 1-minute path, not at the candle close.
Official Binance broker · Non-custodial · Live only when you confirm · Updated
RSI below 30 and a close back above it, a close below the lower Bollinger band, a z-score of price to its 20-bar mean beyond −2. Exit at the mean or after N bars. Works best in ranges and on pairs or baskets where relative value is more stable than price.
Stop fills. A mean-reversion stop sits where volatility is already high, so the fill is often far worse than the stop price. OptAlgo fills stops on 1-minute data with the exchange's rules, which is why many mean-reversion backtests grade A and still show a drawdown their author did not expect.
Say the extreme, the exit, the stop and the hold limit.
“Mean-revert ETHUSDT on 1h: buy when the 20-bar z-score is below −2 and RSI(14) is below 25, exit at z-score 0 or after 24 bars, 3% stop, taker fees, 2022 to now.”Single-coin mean reversion rarely survived our walk-forwards; the 2022 trend down turned every fade into a loss. Pairs and baskets with a volatility filter did better. Treat a great in-sample fade as suspect until the holdout agrees.
faq
The grade says the simulation was honest. It lost because the stop fills on the real path were worse than assumed, or because the regime trended. Both are what the grade is for.
Yes; the backtest charges funding on the hold and shows it as a cost line, which matters when you fade a crowded move.
No. Describe the rules to Claude or ChatGPT through the OptAlgo connector and the assistant writes the strategy against the OptAlgo SDK, validates it for free and backtests it. You can read and download the code.
Every run returns an overfit verdict based on how many variants you tried. Then ask for a walk-forward or a holdout: the strategy must keep its edge on data it never saw before it is worth paper trading.
50 backtest units a month on the Free plan.