zacorbul: statistical arbitrage does not seek a market direction. in the options world, selling premium is also a good example. luckily, volatility is certainly mean reverting (statistically and should make sense logically). good current example is the amount of premium priced into spy despite the beta on qqq.
captain jack. i hate the term black swan. it is meaningless. if you have a distribution with excess kurtosis as we have observed, the notion of a 6+ standard deviation event is not impossible to see in our lifetimes (and we have). as for seeing patterns on multiple time frames. much analysis has been done to fit pattern occurences on randomly generated stochastic processes. to your point they occur often on different scales. unfortunately this indicates that there is little predictive power. one point that you stated regarding a sort of building process in which a small events propagates into a larger one - take a look at autoregressive processes. what you describe is well studied and quantified in processes with unit roots. so called "shocks" propagate for infinitely long timeframes. another similar phenomenon is the presence of strange attractors in dynamical systems. but overall i think the key here as you said is a lagging signal. pm me if you want to discuss a model i use - you seem legit.
captain jack. i hate the term black swan. it is meaningless. if you have a distribution with excess kurtosis as we have observed, the notion of a 6+ standard deviation event is not impossible to see in our lifetimes (and we have). as for seeing patterns on multiple time frames. much analysis has been done to fit pattern occurences on randomly generated stochastic processes. to your point they occur often on different scales. unfortunately this indicates that there is little predictive power. one point that you stated regarding a sort of building process in which a small events propagates into a larger one - take a look at autoregressive processes. what you describe is well studied and quantified in processes with unit roots. so called "shocks" propagate for infinitely long timeframes. another similar phenomenon is the presence of strange attractors in dynamical systems. but overall i think the key here as you said is a lagging signal. pm me if you want to discuss a model i use - you seem legit.