Today I will continue developing the point count system I began last time. I will continue the process until I develop a complete system. I should inform you, my readers, that this is a work in process. Meaning that I am developing this and testing my concepts and theories as I write these articles. That being said there may be gaps along the way as I may run into dead ends, inconsistent results and assumptions that don't pan out. When I am finished, I do believe we will have a true and tested system ready to deploy at the tables.
Point count systems are nothing new. Mike Cappelletti included a very rudimentary system in his 2003 book "How to Win at Omaha High-Low Poker" for 2-card high hands only.
Wilson in their software package includes a number of profiles and a number of point count systems. The beauty of Wilson's software is that you can copy any of their systems and modify the point count values. You can then assign them to profiles and test the system. That is exactly what I have done. I used a profile named C.
Chan who typically plays point count system or style (K). I modified style (I) by replacing its point values for pairs with my own. I then ran 500,000 simulations and recorded the results. Before I display those results again let's take a look at the point counts assigned under each system.
The chart above shows the style (K) as packaged with the C. Chan profile. The standard style (I) and the modified style (I), into my own style (S), are also displayed. The major difference between these 3 styles is the middle and lower value pairs. I assigned a higher value to the to the low value pairs whereas Wilson assigns a constant or diminishing value as in some of the other styles.
My studies indicate that the low pairs, especially the deuces play better with two random cards then do the pair of kings. This may be accounted for by Wilson in the values he assigned to low hands. Indeed I may have to adjust them when we take a look at that aspect of the count system. The results from the substitution of my values indicate an overall improvement as indicated in the chart below.
The gain in net win is a modest 3 cents per average. That 3 cent gain combined with an 8% win rate yields a $19,636 gain in profit over the 500,000 hands dealt. Small differences do add up in the long run. There are several high card categories in addition to the low category. Small gains are meaningful.
The chart above shows how well the point count coordinates to the net win rate of the pairs. The match is not perfect but this is normal whenever dealing with an uncertain situation and correlations. There are other factors, too numerous to get into but some deal with rounding in the ranks for instance. How do you play a J and a half? Does one consider it a Jack or a Queen?
The proof is to actually play the system over a sufficient number of hands and then compare the results. This is exactly the procedure that I used to work out the values assigned to high card holdings.
High card holdings are cards that do not aide in the formation of a low. They are cards of rank 9 or higher. As mentioned earlier I based my model on Wilson's basic (I) style. Listed below is Wilson's (I) style point values and my own (S) style values. The differences are minor in terms of point count values. I dropped the value of the Ace from 7 to 6 and increased the value of the Queen from 1 to 2. How did I determine these values you ask? I used systematic trial and error and the running of simulations. So let me explain.
I first changed the value for the Ace by one increasing it to 8 from 7. I ran 500,000 simulations using that value and compared the result.
If the average net win was greater I would then increase the value and run the simulation over again until the average net win showed no increase or a decrease. If the results of the first simulation above produced a decrease I lowered the point count by one and ran the process again. By utilizing this approach I was able to develop the best value for the Ace that maximized the net win, given the prior changes to the pair values.
With the new or revised values, six in this case for the Ace, I proceeded to the King and utilized the same logic. I then worked my way down through the list, rank by rank and recorded the changes.
This is a time consuming method but I believe it will produce the best results. My starting numbers will either be Wilson's "I" style values or numbers based on my database values. In either case I will continue to run simulations to isolate the best values in conjunction with the other changes. I will continue to report the results back to you through these articles.
The ultimate goal is to produce the most effective, yet simple point count method. Next time I will answer some readers question with respect to issues raised from statements made by other authors in their book.
By the time you read this article I should have available my first book which consolidates my articles over the first 3 years of my writing. The book will retail for $29.95 plus S&H of $5.95 for priority mail.
