Quote from zentrader:
The bottom line is your edge, ie:
win % * average winner - loss % * average loser.
You can trade like acrary with small losses and the occassional home run, or you can go for a high winning percentage with a more even distribution in the magnitude of winning and losing trades. I don't think it really matters how you get there, as long as the number is positive.
Great post. Obviously the optimal style (in theory) is to play trending/breakout/pyramiding methods in markets displaying those tendencies, and play the fading approach in the more frequent choppy/rangebound markets.
Acrary's position-sizing algorithm (used by many good traders) is basically a simple method of ensuring that this is indeed the case. It uses the initial positions as "tests" of the character of the market. If these positions are profitable, this raises the probability that the market is in trending mode. If further adds are profitable, this raises the probability even more. When the probability is higher, obviously you want a large position on to bet on the continuation.
In a similar fashion, good fade traders use increasingly large losses to tell them that the market has changed from a fadeable ranging character, into a trending character. All fade traders know when they have to eventually get out on of a nasty loser, on the rare occasions that a true trend continuation is occuring. Once again, they are using the P&L from their positions as a simple method of detecting the character of the market.
They key is to be able to distinguish between profits or losses caused by "noise", and profits or losses caused by a change in market character. If you exit due to noise then you lose. If you stay in whilst mistaking changing character for noise, you also lose.
For this reason, I think it is better to observe the character of the market directly, rather than via the P&L of "test" positions. When you have positions on, you get psychological distortions, and are taking risk. If you can observe the character of the market equally well or better without these tests, then you avoid the risk and distortion problems, and are more likely to correctly identify noise versus signal.
As for biases, personally when I find I have a bias, I prefer to eliminate the bias rather than stick to markets which suit the bias. This is not always easy but is always more profitable. If I think the bias is unsurmountable, then I will follow Zentrader's method of simply avoiding markets where it will impair profitability.
