Most of the algorithms that I have developed so far (none of them has been any good, btw, that has to be said) have depended in one way or another on estimates of expected returns. Expected returns are, if I understand it correctly, of course, required for portfolio optimization, CAPM related approaches and even momentum ranking. I have found it very difficult to obtain their estimated values that would have been statistically significant, though. My question to the community is therefore the following:
1) Do you yourself use such estimates in your algorithms?
2) If so, have you been able to estimate them reliably?
Sincerest thanks for any comments that you are willing to provide.