CVdata win %

[RevJordan]

Just started using CVdata and I'm loving it.

Just noticed something trivial but got me curious.

Simulation 1000 million hands, 6 decks and looking at percentage of win/lose/tie for many different strategies.

Percentage won is slightly less than percentage lost at all true counts up to TC+20, except at TC+18.

What's so special about TC+18 that you win more hands than you lose?

[bigplayer]

Sample size gets very small at extreme True Counts thus you can occasionally get Win/Loss figures within a count bucket that don't follow the usual pattern. (i.e., as sample size for each count gets smaller your standard error within that count bucket gets big). That's why we run simulations with very large numbers of rounds so that even the extreme count buckets get sufficient samples to minimize standard error.

[RevJordan]

Makes sense. Just happened to always be TC+18 that showed the difference in w/l percents. Thanks!

[Norm]

Also, some extreme counts occur with specific unusual subsets of cards. You may see this effect with single-deck, particularly with deep penetration.