My systems just run on their own without any intervention on my side. This is a prerequisity for any system I develop : it has to run on its own as I don't want to sit behind my computer daylong. I use a portfolio of different strategies on different time scales, in order to be able to identify quickly if one is not working anymore and hence to do a rollover of strategies in my portfolio. Everything is automatised. Thanks linux for stability... and perl for being, let me say, fault tolerant...
How can you say that? high frequency systems are much less vulnerable to black swans than long term systems. Yes, tighter stops.. but if you are hedged appropriately your loss is going to be a few percent max.. And then the system should be smart enough to realize the markets have went into 'uncharted territory' and cease trading until some human helps it along. An ultra long term system is highly vulnerable to black swans because structural breaks in the system are more likely to occur and when you realize it'll be too late.
The amount of cycles a high frequency system makes from profit to drawdown takes years for long term systems to realize. A high frequency system is able to shift its outlook much quicker than a long term system and thus the chances of getting hit by a black swan is less. It can be argued that it will actually profit from it as its outlook would shift 180 degrees.
I'm interested as to how this could happen.. I'm always nearly completely market neutral so even if the exchange catches fire I should be pretty safe.
I am interested to know, as well how they blew themselves up. Naturally all systems should be stress tested and I typically will manually intervene during high stress events in terms of the net risk exposure taken. I actually have had the opposite results than you. I have found that extremely long term (weekly/monthly) models tend to be very volatile in their results due to the longer cycles between profitability and drawdowns.
My faith in diversification is also reduced due to the recent increase in correlation of markets during high stress events. Unrelated instruments tend to display a high correlation during high stress events than traditional theory would otherwise suggest. Currently I am doing some research on the implications of short term prices in markets during high stress news events and this phenomenon is displayed based upon preliminary data.
Fascinating, how are you performing your analysis? I'm also studying related ieas using cointegration, causality, automatically generated directed graphs, etc.
I am using simple correlation to determine which can be considered to be statistically unrelated in price action. After that its a matter of identifying high stress news events and using different volatility measurements to look for the affects. The next step will be to determine the impact of these events in various time lengths which could show possible opportunities.
Not sure about this technical speak.. I don't usually look at charts because i think they can be deceptive.. But..I don't hedge with options, I find highly cointegrated ETFs, indices, etc. short one, long the other, or use baskets of etfs, etc. It is not a perfect hedge but it has performed well during all sorts of major events.