Martin from Bulgaria emailed this week asking for my opinion on a strategy. He ran a マルチンゲール EA through the backtester. 驚くことはないです。, it flopped. A pattern in the winning trades, しかし, caught his attention.
Martin wondered, “if I only took trades after 7 losing signals, would I expect my trade to win?” His idea is to not accept all signals, but to only place real trades when the signals lose too often.
The data showed 5 うち 593 total trades experienced 7 consecutive losers before winning. The EA never experienced 8 consecutive losers. Is this a good time to trade?
The best way to answer the question is through a counter point. If the backtested outcomes were random, then the signals are bad. If the outcomes are not random, then the signals are useful.
Is this random?
Martin provided a key bit of information. The stop loss was 50 ピップ離れて. The take profit was also 50 ピップ離れて. A random outcome would make us profitable half of the time (0.5 または 50%).
The formula for predicting the percentage of 7 consecutive losers among all trades is losing percentage^number of consecutive losers. この場合, it’s 0.5^7. Roughly 0.78% of all winning trades should experience 7 consecutive losses before winning.
The stats that Martin provided were 593 winning trades. Of those, 5 experienced 7 consecutive losing trades before finally winning. 5/593 = 0.80%. That’s almost exactly what a random outcome should produce.
Randomness implies independence between trials. If you flip a coin 10 times and it comes out tails every time, the chance of coming up heads on the next flip does not change. The odds of winning stay at 50% regardless of what happened in the past.
Martin’s idea was to use the information in the signals to place trades. Because the signals show random results and because random results means each signal is independent, the idea is not viable.