[MrX]

Originally posted on 12/08/2009:

If quantification is the issue here, how are you sure that you have an edge or that you aren't overbetting your edge when you are flat betting?

I agree with this 100%. Inability to quantify your edge is overwhelmingly the number 1 argument given against Kelly, and I think it should be way down the list, if it should even be there at all. If you're sure enough that you have an edge to be betting, you should have a good enough estimate of that edge to utilize Kelly.

What doesn't get mentioned enough, in my opinion, is that almost no one has the utility function that Kelly assumes. It sounds like a no-brainer that maximizing long-term bankroll growth is what we want, but there's a lot more that goes into quality of life than long-term bankroll growth. There is a lot of short-term stuff that is extremely important to a person's happiness.

Most people experience diminishing returns with larger and larger wins. There's a much bigger psychological difference between losing $500 vs winning $500 than there is between winning $3000 and winning $4000. I think that a big part of the reason that Chase systems are as popular as they are is because their resulting high frequency of small wins and small frequency of large losses results in much greater utility for the average bettor than other types of money management.

Take two recreational gamblers who are going to lose $5000 over the course of the year. One is flat betting and experiences a fairly steady decline of his bankroll over the course of the year. He wins about half of his days. The other chases losses and wins 19 out of 20 days. His occasional bad day wipes out all of his wins and then some. Most people will have a lot more fun being the second gambler, especially if they're making minor changes (filters!) after each loss so they can continually think they've got it right. I think that for most losing recreational bettors, a chase-type money management system is far superior to a flat-betting system.

Professionals and other gamblers serious about long term profits have a different utility function, of course, but I think most make the mistake of forgetting that the goal is to maximize quality of life, not just bankroll. Most underestimate how much the short-term affects their quality of life.

When I started out, I used full-Kelly for my first season. My model over-performed and I sextupled my bankroll, so I can't say anything bad about my experience with full-Kelly. After that season, though, I looked a little deeper into what full-Kelly would entail going forward, and I never went back to it.

I think anyone with a quantitative model and the ability to project seasons would be well served to experiment with the results of different staking strategies and make an honest assessment of how the results would have affected your life throughout the entire period. Don't focus only on the final bankroll. Take a good look at the entire time-line and how it would have affected you. How many swings would have you miserable and doubting yourself and your model? How many times would you wonder if you pissed off God or something?

Those times are not fun, and lessening their frequency and brutality is often worth sacrificing some long-term growth. Most with the ability to develop a profitable model will get rich and will run into difficulty getting down as much money as they should long before retirement age anyway, so why not make the ride more enjoyable?