Findings · 2026-11-18
The "86% win rate" MACD strategy wins 41% of the time on crypto
A MACD and 200 EMA strategy from a YouTube video with over 3.5 million views promises to win 86% of trades. We ran it, and two other famous "high win rate" setups, on ten large coins for six years. All three lost money after fees.
The claim. “This MACD strategy wins 86% of trades.”
What we found. It won 41% of trades on the years we had not looked at, and lost money after fees. So did the other two.
Follow a signal channel? We score every call it made, after fees. About a minute.
The three strategies
The MACD strategy: only buy when the price is above its 200-period average, and buy when the MACD line crosses up through its signal line below zero. The stop goes under the recent low and the target is one and a half times the risk. Selling is the mirror image. We ran it on 1-hour bars.
The Bollinger and RSI bounce: buy when the price closes below the lower Bollinger Band with RSI under 30, sell when it gets back to the middle band. The triple EMA scalper: trade in the direction of three stacked moving averages when the Stochastic RSI crosses back out of its extreme, with a target of two times the recent average range and a stop of three. Both are shared everywhere as "high win rate" setups.
What happened
The MACD strategy won 41% of its trades from September 2023 to October 2026, before fees, not 86%. With a target only one and a half times its stop, that is not enough, and it lost 0.11% a trade after fees. Re-tests by other people on gold and on the S&P 500 found the same thing: about one trade in three.
The Bollinger and RSI bounce won 48% of the time and lost 0.20% a trade. The scalper got closest to its promise, winning 61% of its trades, but its stop is wider than its target, so each loss wipes out more than one win. It lost 0.07% a trade over more than 15,000 trades.
Why win rate is the wrong number
Any strategy can show a high win rate if it takes profits small and lets losses run. What pays is the average trade after costs: how often you win times how much, minus how often you lose times how much, minus fees. None of these three is positive on that count.
And the prop-firm challenge
We put all three through prop-firm challenge rules, starting a new challenge every week for three years. They passed between none and one in ten, and collected almost nothing afterwards. You can read the same test on eight TradingView strategies in our Findings.
How it was tested
- Binance USDT-M perpetual futures, ten large coins, 1-hour candles (15-minute for the scalper) from data.binance.vision, January 2020 to October 2026.
- Signals on closed bars, filled at the next bar's open; stops and targets checked against each bar's high and low, and when both were touched in one bar we counted the stop. Costs: 0.107% taker fees per round trip plus funding.
- Rules locked on 17 November 2026 before the results were run. Figures quoted are from trades after 20 September 2023. Uncertainty measured across weeks, not trades.
Research on past data, not financial advice. Every test we run, failures included, is on the Proving Ground.