Originally Posted by
podonne
I'm not sure I follow all the math in these two posts, but I have to agree with wantitall4moi that Kelly is not appropriate for sports betting.
Kelly requires you to know your exact edge in any given betting situation, not an estimated edge based on your past results with this type of situation, which is all you can ever get with sports betting. Kelly works wonderfully only in mathmatically-defined games like a card counter in blackjack.
Actually, I'd argue the two are almost exactly the same. Unless you're using computer software when playing BJ or have access to Rain Man, you're most likely using one of several card counting systems. These systems only provide
estimates of the expected true deck probabilities, which in certain cases may be off substantially.
The same dynamic exists with sports betting. We don't know the exact probabilities of a given outcome occurring, but nevertheless, through a combination of quantitative and qualitative analysis we can still determine (hopefully) unbiased
estimates of event win/push probabilities.
Compare the probabilities implied by whatever counting system you use with the true prior probabilities over a few hundred million deck runs, and what you'll probably see are that your system's predictive figures possess a similar degree of accuracy as that which might be expected of an advantage sports bettor.