Javascript simulation with graphs
Hello; I have some information for the general community, and some questions for Don Schlesinger, if I may. I had wanted to see if I could calculate basic strategy so that I could actually see what was happening under the hood. (I do not know how software like CV does it.) I used javascript because 1) I already know how, and 2) it is very well suited for graphic output via html. The main problem is that it is relatively slow. I calculated win rates for each hand using each applicable action (hit, stand, double, split, surrender), which takes the better part of an hour for 1,000,000 sims of each hand/upcard/action combination (5,000,000 would be better). I did this for a range of counts, then created graphs for each hand that show the win rates vs. count for each action. You can see the relative return for each action, and more importantly, where the win rates cross over indicating the index for a departure from basic strategy.
Then I ran sims to determine the count frequency for a given system and number of decks, and the hand frequency for each count using the strategy determined by the win rates for that count. (Everything is specific to the count, and rules etc.) Combine this with the bet ramp and you have win rate * hand frequency * count frequency * units bet; sum this for each count and you get the actual advantage obtained from each hand. This information can be used to calculate “illustrious” indices. The same thing can be done for insurance, getting the odds of the dealer drawing a 10 on an ace for each count.
Yet another sim calculates the real-life win rate for each hand, including player and dealer blackjacks, that are not used when doing strategy win rates. Adding all these * hand frequencies gives the overall player advantage for the game.
I ran a complete sim using the parameters provided on page 57 of Blackjack Attack. I have attached some screenshots of the results. The colors for the graphs — green: hit; red: stand; blue: double; gold: split; black: surrender (-0.5); gray: baseline reference (0); cyan: hand frequency. There is a vertical line for the starting count, which in these examples was zero.
My top 18 indices are similar, but not identical, to the illustrious 18 in the book. Some of my questions involve the following:
I ran all the sims with a bailout count of zero, meaning that no data were obtained for counts that went negative. The bet ramp on page 4 implies that, but page 57 does not come out and say it. Also, I used counts up to 10 (0, 2, 4, 6, 10); I don’t know if that was the right way to do it either. A full set of data would include every possible count, but high counts take forever as decks with each count have to be found at random (I reused them a number of times, shuffling the cards still in the shoe by several means, or it could take 10x as long or more for high counts). I do not know if I could exactly duplicate the illustrious 18 if I used the right parameters. Also, I used the baseline rules (S17, no DAS, no HSA, no RSA, no surrender). [edited]
Just wondering if there is any interest. What I have is a collection of html documents whose output needs to be combined and run together for the final result. And there is some hand-coding involved, and judgment about how many threads to run. (In other words, not ready for prime time any time soon).
-BC
graphs.jpg
indices.jpg
insurance.jpg
count.png
bs.jpg
Reply With Quote
Originally Posted by BackCounter
Bookmarks