Thanks for the post, though the first section seems to have swapped the definition of a and b (compared to the linked wiki page) -- the formula here requires decreasing the leverage for very large a, which does not make sense if a is profit.
You mention that if we only consider the constant and linear term of the Taylor expansion, we get f = mu/mu_2, which is equivalent to the non-skew non-kurtosis approximation of Kelly (mu / sigma^2). Wouldn't mu / (mu_2 - mu^2) be equivalent to non-skew Kelly? mu_2 is just E[X^2].
Thanks for the post, though the first section seems to have swapped the definition of a and b (compared to the linked wiki page) -- the formula here requires decreasing the leverage for very large a, which does not make sense if a is profit.
Correction applied, thank you!
You mention that if we only consider the constant and linear term of the Taylor expansion, we get f = mu/mu_2, which is equivalent to the non-skew non-kurtosis approximation of Kelly (mu / sigma^2). Wouldn't mu / (mu_2 - mu^2) be equivalent to non-skew Kelly? mu_2 is just E[X^2].