Peter Muller, a 43-year-old trader at Morgan Stanley, is used to markets behaving more or less as he expects. But in late July, some unusual patterns perplexed him. Certain investing strategies that historically had posted steady gains started faltering for no evident reason. ...Though few on Wall Street know about it, his group at Morgan Stanley has been among the investment bank's most profitable operations in recent years. Known as PDT, for Process Driven Trading, it produced profits of roughly $3.5 billion in the 10 years through 2006, people familiar with it say. They add that PDT, which now contains about $6 billion of Morgan Stanley's money, accounted for 7.2% of the bank's net income last year by producing $540 million in profits. But between the last week of July and Aug. 9, PDT lost approximately $500 million, according to traders. Neither Morgan Stanley nor Mr. Muller would comment on the losses or on PDT's trading strategy. Morgan Stanley said it is fully committed to the quantitative trading business... http://online.wsj.com/article/SB118912592144720147.html?mod=googlenews_wsj Come on Morgan Stanley - at least I hoped you guys drove past the QUANT mess - sh.t, I forgot, they are also part of PPT, or ?
Models that anticipate every contingency may have very low draw downs, but they will never have out sized gains. Further, the more contingencies that you hard wire into a model, the more bloated and curve fit the model is. The better way is to have circuit breakers by human beings that understand model statistics to take if off line, or to filter out trades. More likely than not, the relationships will snap back the way a rubber band that is too stretched does. The worry isn't that the rubber band is stretched by more than you have ever seen, the real worry is that it breaks. nitro
I found this to be a far more interesting article. http://www.nytimes.com/2007/08/26/m...d=1&ei=5070&en=b26548f2bbf6aea2&ex=1189483200 AFAIK, it is interesting that Taleb doesn't pick up on nature tail events and the way markets (mis)price them. nitro
I think that many quantâs models did not factored the power of forced redemption. IOW , many of the âVâ shape positions were liquidated at the lowest point of PnL curve.