Disliked{quote}1.2. You can use something like a variance calculated Sharpe ratio and only use systems above a certain point. Reducing variance after the backtestings ends is the hard thing because you enter the realm of the unknown.Ignored
QuoteDisliked2.1. To be sincere I have never seen if the systems that fall into a negatively biased random walk have a positive autocorrelation (do you mean a positive autocorrelation of the return series?).
Yes, and it is just a guess. If it really is a random walk then of course the returns are random and not correlated. But if there is a trend (negative bias) in the random walk then maybe this would show up on a statistical level (acf) on either the return or trade series. Maybe I need to do some simulations in R...
QuoteDislikedI have some strict criteria for discarding systems and once they are discarded I no longer monitor them. I generally discard systems at a 99.5% confidence for the tests I use for this purpose.
This implies that there is a higher probability after the statistical criteria to discard is met, that performance will continue to deteriorate or at least not be positive. This seems like an hypothesis that could be tested. Or maybe it is just wiser to reduce the unknowns by discarding a system whose performance doesn't match its historical performance distribution? Could a second hypothesis test be used to turn a system back on after it has been deactivated? Would this off/on switch allow for lower thresholds than 99.5% confidence and thus quicker exits from downward equity curves and quicker re-entries to upward sloping equity curves?
I have some systems that need to be re-optimized about every 6 months or performance deteriorates. It is also true that all systems have periods where they perform well and poorly. If performance degrades below a certain threshold such that the distributions are really different then it makes sense to discard the system. But from a practical standpoint, when you are at that point you have typically already endured a fairly sizable drawdown. But I'm not sure if I have any insight as to how to overcome this problem. It's sort of like taking a stop loss. Your future loss is limited but you've locked in the current loss at the worst point up to the current time.