Can ET veterans please explain why we use average instead of median win/loss ratio when calculating Kelly? The distribution of returns for both losses and wins are often skewed and using simple average doesn't seem to reflect this. Yana
Because the Kelly assumes perfect information with regards to gains and losses. In that case it is necessary to use the average otherwise you could end up betting too little or too much in case there is a significant divergence to the median. The average will be affected by the outliers more than the median which provides the formula additional information about the situation. This is rarely the case in trading but it is the holy grail in casino games
The question you should really be asking yourself is why you're bothering with the win/loss ratio at all when calculating Kelly.
You're probably using the wrong form of the Kelly Criteria if you're using the win loss ratio, you need to look at the continuous version. As for the use of Gaussian based statistics, well nothing is perfect in trading and using non-parametric statistics isn't going to be the difference between winning and losing, we just want good approximations.
If the PnL is not normally distributed, but you trade many many times, then over time the formula for optimal kelly rule is reduced to a fractional kelly rule due to law of large number. njrookie