3 out of 3 members found this post helpful. Did you find this post helpful? Yes | No

## CVCX Update – Readapting Kelly

A free update to CVCX V6 and CVData V6 is now available with an overhauled optimal betting calculation engine. First, a few words on Kelly criterion:

First described in 1956, Kelly criterion is designed to optimize bets for maximizing the logarithmic growth rate of a bankroll. The methodology has reportedly been used by successful investors like Warren Buffett. I believe Thorp was the first to apply it to blackjack. When applied to blackjack, some difficulties arise. The major problem, as Don has recently pointed out here, is the need to keep the bets in a range of a minimum bet and that amount times the desired spread. There is an iterative algorithm, long used by CVCX, to determine the optimal minimum bet. But it is not perfect. For example, some bets may be made with negative EV, you cannot bet pennies, and simply rounding bets to the closet integer bet is less than optimal. CVCX has always included code to deal with this. Any constraint on Kelly reduces the optimization.

Other constraints exist, some of which are becoming more bothersome to the calculations. For example, table minimums are growing, which may not allow the optimal minimum to be used. You may not wish to use dollar chips or \$5 chips. For example you may want all bets to use green and higher chips. Or you may want bets to be multiples of the minimum. You may desire a small spread for cover purposes. You may want a very low risk of ruin. You may have a limited bankroll. Each of these can substantially affect the calculations and more so in combination. Given the increasing constraints, I decided to overhaul the code a few months ago. Kelly criterion is still used at the heart of the new process. But there are now several additional levels of polishing to find a more optimal solution taking into account all of the user’s inputs. Of course you can still tweak the results to find a betting ramp that you find more comfortable, while seeing the immediate impacts on risk and win rate.

The point of this exercise is to squeeze more dollars out of the casinos while improving the ability to fit your desired risk. If your situation is simple—a large bankroll compared to your risk of ruin, a decent game, reasonably large spread, no large constraint on chip sizes—you may often see little or no improvement. The more constraints, the more likely you will see a substantial improvement in win rate and/or a result that is much closer to your desired risk.

There is also a new Table Minimum setting. This is different from the Min. Chip Size setting. So if it is a black table, set it to \$100 and all calculated bets will be at least that amount.

Note: the time required for a solution will increase in many cases. Most will still require a fraction of a second, but more stringent constraints can require substantially more time as some will require a million instructions. As you click through the levels of desired risk, spread, or penetration, you will note that some solutions are instantaneous while others may pause for a few seconds. Fortunately, PCs have been rapidly growing in speed. If you wish to make many setting changes, there is a new field named Suspend Calculations that will stop the constant recalculation while you are changing the settings. When all settings are as you wish, turn it off to get a new calculation.

If you have CVCX V6 or CVData V6, when you start the programs, the bottom left of the main screen will indicate that a new release is available. Click the install button and both programs will be upgraded in a few seconds. I will still be fine tuning for a while. So if you see a situation where you think there is a better solution, please email me.