Page 1 of 1
Kelly criterion and investing
Posted: Sun Mar 24, 2013 10:14 am
by BP
Anyone have any experience with the use of the Kelly criterion with investing?
https://en.wikipedia.org/wiki/John_Larry_Kelly,_Jr
The article references Jim Simons:
http://www.insidermonkey.com/hedge-fund ... ologies/5/
Re: Kelly criterion and investing
Posted: Sun Mar 24, 2013 5:15 pm
by MachineGhost
Sure, I've played around with it. It's best viewed as the the upper maximum to ever bet because if the future distribution of returns shifts to the left, it will become non-optimal and over-risky. If you want to learn more about it, look for Ed Thorp's book who used it extensively to make millions with his hedge fund.
Edit: This is the book:
http://www.amazon.com/Fortunes-Formula- ... 809045990/
Re: Kelly criterion and investing
Posted: Mon Mar 25, 2013 4:37 pm
by BP
Thanks MG. Based on your response, I assume that your results did not lead you to believe that you had found the "magic" money making machine like RenTech?
Re: Kelly criterion and investing
Posted: Tue Mar 26, 2013 12:36 pm
by rocketdog
Search YouTube for "The Trillion Dollar Bet", a 48-minute BBC documentary that aired on Nova. It's a sobering account of what can happen if you rely too heavily on "proven" mathematical formulas when investing. Whenever you encounter an investment "system" that's designed to beat the market, just remember what happened to these guys.
Re: Kelly criterion and investing
Posted: Tue Mar 26, 2013 3:29 pm
by MachineGhost
BP wrote:
Thanks MG. Based on your response, I assume that your results did not lead you to believe that you had found the "magic" money making machine like RenTech?
Kelly doesn't lend itself very well to a portfolio, only one at a time trades. RenTech makes money because they use quantitative analysis, not because of Kelly. No form of money management will make a mountain out of a molehill.
That being said, Kelly is on my to do list to see if it can work at the portfolio level. The problem at the portfolio level is diversification is better done by equal-weighting or similarly to achieve more simultaneous trades. One at a time trades usually suffer from limited statistical significance, even though they can span a long period of time. Kelly assumes the probability of winning will not change which is not true in reality, trades will swing above and below the win percentage mean. It's far better to focus on risk by "risk targeting" and let the wins take care of themselves.