Anyone who's still doing that needs to fire their quant/portfolio construction team. This has been known problem for decades. VaR, in generally, isn't used much in the front office. It's usually reserved for regulatory reporting. If you are using VaR is actual portfolio management, you suck. (btw, LTM is an example of bad front office risk management leading to failure; the CDO/Subprime blow up is a lot more subtle than that. As someone who was in that space when it happened and before it happened, I'd argue it was a huge regulatory/ratings arbitrage that eventually played out in the way these things always do - tears). Anyway, "money management" (or, as we call it, trade sizing, liquidity management, risk management, and portfolio construction) is a relatively "new" area still. I've seen more and more tools being developed lately that are more concerned with how bad can things get vs how-much can we leverage based on blah blah blah (which is essentially what VaR answers).
Hi dtrader98, Just for the record, I have read Pounderstone's Fortune's Formula and just re-checked again. In that book, he never made the claim that Citadel, Medallion, and D.E. Shaw are using or even aware of Kelly. The only place those firms are mentioned is around page 319 where he said those firms are the competitors of Thorp's stat arb fund. I believe the author of that report, James Case, made a FALSE statement regarding something Pounderstone never said. It's a great book by the way.
Good catch, as I haven't looked at it in awhile it is possible that maybe he obtained this information from Poundstone outside of the book or made the inference implicitely. There is certainly a well known relationship between medallion and its origins going back to axcom and Berlekamp (who had C. Shannon as an advisor!). I would suggest that they are quite aware of kelly, and would consider it reasonable to assume that Berlekamp used an information theoretic approach likely incorporating kelly, although not necessarily divulging to the public his methods. Berlekamp himself reviewed fortune's formula (American Scientist), and had this to say, "No one who has made a legitimate fortune in the markets believes the efficient-market hypothesis. And conversely, no one who believes the efficient-market hypothesis has ever made a large fortune investing in the financial markets, unless she began with a moderately large fortune. Of the stories presented in Fortune's Formula, the case of Ed Thorp presents the greatest challenge to the efficient-market hypothesis. Poundstone devotes only a single paragraph to the even stronger cases of Ken Griffin, D. E. Shaw and Jim Simons, presumably because financial wizards as successful as these have always been unwilling to discuss their formulas in public." http://www.americanscientist.org/bookshelf/pub/bettor-math
In fact, the fallout between Simons and Ax was argumentatively due to drawdowns that Simons was uncomfortable with. Their parting of ways led Simons to Berlekamp, who's methodology likely addressed these concerns and performed quite well. I have no doubt those methodologies incorporated kelly type money management considering the background of players and events leading up to that point.
... There's no reason for Citadel or DE Shaw or any of these shops to be "aware" of kelly's. Kelly's ratio is basically a particular case of a general class of problems called optimal stochastic control. It's in fact a solution only when you throw in tons of assumptions. Some of these, like independence of out comes, stationarity of distribution, etc, - if written elsewhere most people here would cry murder. Yet, it's embedded in Kelly's. Kelly's criterion is far more simplistic and "flawed"[*] than VaR or any of the other measures that people like to bash. [*] it in itself is not flawed. it's just not meant to be used for trade position sizing. The usage is flawed.
Thanks for the book review dtrader98. Btw where did you learn about the historical background of Simons? He is so secretive that I have only seen one Bloomberg report about him. That was an interesting digression, but I am still interested in how portfolio allocation is done in the professional money management industry. As an outsider, the only method I have heard of is the mean-variance optimization approach. From my amateur point of view, I would argue that any method based "strictly" on historical variance is risky because everything tends to become highly correlated during crisis. Eric Rosenfeld highlighted this point when he gave a talk at MIT: http://www.marketfolly.com/2009/05/long-term-capital-managements-collapse.html Another topic I am unsure about is how to apply Kelly Criterion to portfolio allocation. The original Kelly Criterion seems to be applicable to a single stream of events. Can Kelly Criterion be applied to optimal portfolio allocation at all? If so, I would be interested to know which direction I should take to learn more about it.
I totally agree. I would even go as far as to say that all the financial models are never wrong. It is the people who mis-use them who are wrong. They applied the methods without considering whether the underlying assumptions behind the models are realistic or practical to the real world.
Plenty of information out there. Dig with diligence. yes, I mentioned the correlation aspect a few threads back. No easy textbook answer to give, but it is something that certainly requires thought. There are other ways to spread risk outside of position sizing. Plenty of ways to think about this. I thought Mr. Vince was here to describe one. I'll let him take the stage. ---------------------------- As pointed out by others, the common interpretation of kelly towards markets has pitfalls. The original idea gives a nice fixed answer with a nice fixed distribution, based on nice binary outcome assumptions. We all know reality is not that way. That does not preclude the concept as being a useless one if understood properly and applied in a useful manner. I had hoped the anecdotal back-story on Berlekamp and medallion might give some testimony to that, but feel free to investigate and draw your own conclusions; just don't expect a typical textbook approach to be the only one.