In the last two weeks since I put K’s account on my “cherry-picked” trading system, she is up 200 pips net so my experiment of putting outcome ahead of process is working.
But even if you are going to focus on results rather than adhere to statistical norms you still need a method for determining which pairs and strategies to select.
To review, last week I argued that traditional statistical methods of system analysis do not work. Classical statistics teach that a system is “valid” only if it works across all pairs and multiple time frames. That may be true for physical phenomena like coin-flipping but is less than useless for psychological ones like trading markets.
All trading systems fail and the best way to approach them is to exploit the recency bias for as long as it works. Thus the title of last week’s column — forget Mr. Right, choose Mr. Right Now.
But even if you are going to abandon statistical orthodoxy you still need a process to make your choices. So here is my method for cherry-picking the trades. First and foremost you still need a decent sample size. So even if I am going to backtest just this year’s data, I still want to see 50 — ideally 100 trades to give me some confidence that this is an exploitable pattern.
Secondly, you may get a lot of marginally positive results but what you really want is some reasonable margin of error. So forget Sharpe ratio, profit factors and the myriad of statistical tools at your disposal. I keep it simple. Expected profit per trade. So if am I trading on an hourly chart I want to see 6 pips or better. Why? Because I assume that at BEST I will be able to achieve only half of that in real life so having a buffer is crucial. For daily charts, my minimum expected target is 10 pips per trade and ideally, I want to see 15 pips. Again, I assume in real life I would be lucky to get half.
Third, get paid even if you break even. I make sure I get a rebate on every trade I make. My strategies generate about 1000 trades per year so even without any leverage at all I can make 2% on the account. At 3:1 lever I can still make 5-6% even if my projected 25% backtested returns end up a Big Bagel.
Fourth and last. Do. Not. Lever. I never trade more than 3:1 lever on my auto strategies. Why? Because in real life your actual leverage is actually much higher. If you are trading automated strategies across several pairs you will often float 5 trades at once. At 3:1 lever your account is effectively levered to 15:1 at that point. Furthermore, that strategy is very likely correlated to a specific market regime so all five positions could flip against you and the equity hit at that actual leverage could wipe out 20% of your account in one fell swoop.
Finally, cull every three months. If the pairs are working, keep riding them. But if the strategy starts to fail you turn it off and start running what’s working in backtests now. Yeah, I know this is anathema to most system traders, but once you cross to the dark side there is no turning back. There is no loyalty to your ideas, only respect for your results.