[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Ratio Money Management


  • To: RealTraders Discussion Group <realtraders@xxxxxxxxxxxxxx>
  • Subject: Ratio Money Management
  • From: John Sweeney S&C <JSweeney@xxxxxxxxxxx>
  • Date: Fri, 19 Sep 1997 14:30:30 -0700 (PDT)
  • In-reply-to: <2.2.32.19970919161343.0095a2a8@xxxxxxxx>

PureBytes Links

Trading Reference Links

Extending our comments on the fraction of trading capital to be used in any given trade (See previous notes from Bob Pisani and Louis Kates [snipped below]), can we all agree that it's unlikely that, on any given trade, we face the same probability distribution we faced on any other trade?
John
Bob Pisani: >It comes out of a very simple model in gambling theory, whose simplicity >makes it only remotely applicable to real life investing and trading. As >you may know, the result goes back to a paper by Kelly on information theory >and to a couple of theorems by Breiman showing that the strategy is >"asymptotically optimal". The Kelly criterion was publicized in the early >70's by Ed Thorp and evidently some people just took the idea and ran with >it. But it is in fact optimal only in a very special situation, namely when >the distribution of your profit is the same on every bet -- ie when you are >making precisely the same bet over and over, and only that bet. In the real >world, you will have different situations with different possible outcomes >and differing probabilities associated with those outcomes. If the >distribution changes from bet to bet, the assumptions behind these results >will not pertain, and the fraction of your capital you should bet in order >to achieve maximal capital growth will depend upon the particular >distribution of profits that characterizes your current bet, and this >optimal fraction will typically not be fixed at all but will vary from bet >to bet.
Louis Kates: Just to add specificity, fixed ratio money management in the case of independent bets in which you double your bet with probability p or lose it with probability q=1-p is a system where you bet a fraction f=p-q on each bet. Of course, if p differs from one bet to another then the corresponding f clearly differs as well. There is no implication that there is one f that works for all p.
John Sweeney, Tech. Editor Technical Analysis of Stocks & Commodities Technical Analysis, Inc. The Traders' Magazine 4757 California Ave. S.W. Phone: 206 938-0570 Fax: 206 938-1307 Seattle, WA 98116-4499 USA Web: http://www.traders.com/ ____________________________________________________________________ Contents may not reflect official opinion of Technical Analysis, Inc.