Yes, I remember that G.Soros mentioned LTCM in his last book. I do not think that their bankruptcy discredits the statistical approach completely. The usage of 1:30 leverage was not a very wise idea from the money management point of view. Also I think that one should not use the statistical methods during the «metastable» periods, at which the rules of the game are changed and the process becomes nonstationary. That nonstationarity (turbulence, phase transition) explains «10 sigma events», which never happens in stationary processes. The problem is how to discover the beginnig of metastable
period, but that is also a statistical problem of «critical phenomena» approach.
One should not ignore the rare event risk. Maybe the problem is in not being too greedy J
