i think that the three figures maximum draw down, best month and worst month necessarily increase over time. if they did not than your return distribution becomes steeper and why should it? so in order to remain at the same relative shape of the return distribution all three figures should increase over time. opinions?
if you are assuming a totally static strategy, then the longer you go (the more observations you have), the more/farther outliers you will have eventually/inevitably, hence your statement is correct - in practice however, although my direct knowledge is limited, people get a shock once they experience the worst month/drawdown (if it's a big one) and then tend to (somewhat) adjust their strategy (as simple as cutting leverage) seeking to lower max. drawdowns in the future.
we have several strategies in place at any one point of time and we are constantly adding new ones. my point referred to exactly what you observed: panic like behavior while everything is in shape. ah, and yes, i referred to static for single strat. ...
On an arithmetic basis, they should increase. Otherwise, you're undertrading your system and leaving money on the table. On a percentage basis, you don't want the worst month & drawdown to increase greatly. Otherwise you're likely to re-optimize the system at the worst possible moment before it "returns" to it's normal performance.
well, that is my point. the PERCENTAGE outliers MUST increase if the system remains the same. my point is that if you already traded a system over a decade and you see that it is taking out the best month, the worst month and even the max dd, then this is possibly quite normal. of course there is a too much here as well which should make the bell ring. i think it is important to study this kind of behavior, even in backtests. it makes you more conscious of true processes and enables you act upon fact. so when you trade for longer periods of time taking out the dd is a very normal thing. and by itself maybe just a sign of everything is normal and no reason to pull the trigger. the reason to study this is that you do not want to pull to early, but not too late either, so you have to define where things start to get off line ... and my point is that historical max dd is the wrong figure.