hi folks, i am having a somewhat perplexing problem: i moved to win7 x64 and AB 64bit 5.30 version simultaneously (from win-xp and AB5.20). Everything went smoothly since i reinstalled AB from scratch, except that sharpe ratios for some systems are off (significantly).
Details: I took some systems and ran them on both old and new system, just for comparison. i noticed that the reports for the most part are the same (expectedly - this is deterministic software after all!). However, i noticed that for a couple of my systems, i get a negative sharpe ratio (even though i know it is not negative ;)) in the new install, (it is nicely positive as I expected in the old.). The calculation is ok for some other systems. I checked the obvious things: I am running on the same basket of stocks, with quotes from the same vendor (PremiumData EOD). I have the same start and end dates. I dont configure any settings from the wizard, but set my own in the AFL itself. The errof is seen with different start dates and end dates. potential sources for error: (a) It is possible that some default settings are different in AB 5.30 as opposed to 5.20, but i did a brief eyeballing and it seems ok. I dont think this is the issue, since a couple of my systems report the same sharpe as before (b) maybe the sharpe calculation changed its assumptions (i.,e start date etc.,) i noticed that while the first trade doesnt happen for a while in my system, the built-in Drawdown plot / graph shows negative, but doesnt affect the max dd results. I was wondering if this was the source for some difference in calculations. I do have a workaround, which is to write my own sharpe calculator and stick it in the CBT custom metrics for my own sharpe reporting, but just thought it would be worthwhile to understand the source of this error (who knows what else it masks..).. An associated question: - are the default calculations for sharpe ratio stored in some AFL / folder ? or are they "built-in" to AB source code. Reason i ask is I could put trace statements in there to debug whats going on. thanks for your time and attention to this long email. Again, likely its my error, not AB's of course, so any tips are most welcome. Any clues before i roll up my sleeves and dig in ? bgpl
