Kelly - my decision on a wagering scheme

I thought I had completed my Kelly analysis and noted it was time for me to determine my own personal strategy on a method of sizing bets. I still have about half a baseball season to go, and football is on the horizon. I had done my analysis by using multiple simulations of seasons with 100 bets per season. That’s a reasonable number for the way I bet. I had determined that I agree that using the Kelly Criterion is probably the best approach to growing your capital over the long term and a great number of bets. But the variance it introduces coud be problematic in the short term. I’m 73 years old, I don’t have a long term. So, what to do now?

I went back to my simulator and added a couple of wager methods to it. I wanted to compare 4 methods of determining wager size. They were Kelly, half Kelly, a star system, and flat wagers. The star system assigned a star for each percentage point edge that I had. They ranged from 1 star to 5 stars. I had a base star wager size and multiplied it by the number of stars. This is almost identical to Kelly, except it is constant throughout the season and is independent of what occurred earlier in the season. I adjusted the star and flat bet sizes to be the same as the full Kelly on the average. Following is a table of the results.

Description	Kelly	Half Kelly	Star	Flat
Number of profitable seasons (based on net profit)	619	678	728	705
Average Season Ret/$	$1.036	$1.051	$1.064	$1.050
Best when measuring season's net profit (loss)	193	169	330	308
Best when counting number of winning seasons	0	0	885	148
Best when computing Ret/$	38	68	497	397
Average Bet	$65.69	$30.06	$65.69	$65.69
Average Net	$4.27	$1.98	$4.22	$3.31
Highest Single Season Profit	$5,610.41	$1,680.21	$2,376.73	$1,958.88
Biggest Single Season Loss	-$827.79	-$564.75	-$1,803.47	-$1,427.19
Total Ret/$	$1.065	$1.066	$1.064	$1.050

A famous gambler (I believe it was Nick the Greek) reportedly said “gambling isn’t about money, that’s just how you keep score”. So, if you wanted to go for the biggest gain in a season, you would choose full Kelly. If you wanted a minimum loss, you would choose half Kelly. In my case, I’m just interested in finishing the season with some positive results. From the chart it’s obvious that I should choose a star system.

That leaves another quandary, how to assign stars. In the simulations, I assigned a star for each 1 percent improvement in my edge. But that’s because the simulation had 100% correct edge. In real life, my computed edge almost certainly isn’t 100% accurate. So, it’s off into another study. If the results are interesting, I’ll post them. Follow @ole44bill on Twitter to know if and when I post this.