Ann: November Trading Update, page-66

Not sure where that school of thought comes from but if that’s the case it is frankly stupid.

The Kelly calculations are a good theoretical framework upon which to make bet optimisation decisions, however, unlike fixed systems where probabilities are exact, betting in equity markets is inexact. I’ve found the best way is to start with a guess of what the aggregate distribution is then sense check with my own distribution (which gets more accurate as more time passes and there are new data points). The aggregated results (need to know % gain when correct, % loss when incorrect and % calls correct), can be used to work out theoretical bet limits. Even assuming the distribution is 50-50 of calls correct, the probability of 7 incorrect calls in a row is <1%, depending on your loss taking personality or rules, you can then work out the maximum drawdown you can tolerate at a portfolio level and then work out a maximum size.