you say, you don’t need an advanced degree, but surely it gives you an edge in statistical arbitrage trading? Not necessarily. I received a PhD from one of the top physics departments of the world (Cornell’s). I worked as a successful researcher in one of the top computer science research groups in the world (at that temple of high-techdom: IBM’s T. J. Watson Research Center). Then I worked in a string of top investment banks and hedge funds as a researcher and finally trader, including Morgan Stanley, Credit Suisse, and soon. As a researcher and trader in these august institutions, I had always strived to use some of the advanced mathematical techniques and training that I possessed and applied them to statistical arbitrage trading. Hundreds of millions of dollars of trades later, what wa... (查看原文)
By its very nature, quantitative trading is a highly automated business. Sometimes, the more you manually interfere with the system and override its decision, the worse it will perform. (查看原文)
A historical database of stock prices that does not include stocks that have disappeared due to bankruptcies, delistings, mergers, or acquisitions suffer from the so-called survivorship bias, because only “survivors” of those often unpleasant events remain in the database. (The same term can be applied to mutual fund or hedge fund databases that do not include funds that went out of business.) Backtesting a strategy using data with survivorship bias can be dangerous because it may inflate the historical performance of the strategy. This is especially true if the strategy has a “value” bent; that is,it tends to buy stocks that are cheap. Some stocks were cheap because the companies were going bankrupt shortly. (查看原文)
in general, the more rules the strategy
has, and the more parameters the model has, the more likely it is go-
ing to suffer data-snooping bias. Simple models are often the ones
that will stand the test of time.
This is not to say that no methods based on AI will work in prediction. The
ones that work for me are usually characterized by these properties:
1:
They are based on a sound econometric or rational basis, and not on ran-
dom discovery of patterns.
2:
They have few parameters that need to be fitted to past data.
3:
They involve linear regression only, and not fitting to some esoteric non-
linear functions.
4:
They are conceptually simple.
5:
All optimizations must occur in a lookback moving window, involving no fu-
ture unseen data. And the effect of this optimization must be co... (查看原文)
