Hi Gronbog,
Glad to hear that you are a professional software engineer, could you or is it possible to write a "stand alone" blackjack strategy calculator to generate perfect strategy WITHOUT INTERNET CONNECT base on exact remaining cards ? I am thinking to use it to play online game.
Here are the perfect BJ strategy calculator : https://wizardofodds.com/games/black...nd-calculator/
Unfortunately this calculator required internet connection before it can generate perfect strategy.
James
Yes. Of course, I have to clarify. The gain is in percentage over the base. So, say HiOpt II + ASC wins $50/hr. Tarzan will win $52.50, or a 5% gain. Perfect play will win $62.50, or a 25% gain. The ratio of the two gains is 5/25 or 20/100.
Just my estimates, but they aren't to be trifled with! :-)
Don
Can't you tell from Don's post that he pulled those numbers out of the air. I am sure they get the just of the ratios but I wouldn't expect then to be exact. If Don intended them to be taken as exact he would have made that clear by his wording. Clarification would be to put an error range to the estimate, which only Don could do. We have posts on perfect play and Hiopt2/ASC by Eric. They may not be apples to apples sims but if you assume they are then you could fine tune Don's approximation.
Here is a thread by Eric that might be applicable. Start at his first 2 posts, post #1 and #2:
https://www.blackjacktheforum.com/sh...452#post214452
As you can see Hiopt2/ASC bets near perfectly with optimal betting and optimal play coming in at $34.90/hour while Hiopt/ASC with optimal play at $33.18. But with hiopt2/ASC with its full indices it only makes $27.01. The difference between $27.01 and $33.18 is the increased performance from perfect play, which is aided by a 9.99% increase in the optimal bets. This is for S17, DOA, DAS, SPL1, no surrender, 4.5/6.0 pen, floored indices with 1/2 deck resolution, spread 1-16.
The closest T Count comes to this is where the comparative stats are:
So the strongest versions of T Count has a range increase over Hiopt2/ASC of 0.59% to 0.9067% in the shoe game defined.
So increasing Hiopt2/ASC's $27.01/hr by that amount gets a range of $27.16/hr to $27.25/hr to compare to perfect play of $33.18/hr. Using T Counts strongest version for this shoe game comparison, while assuming the percent increase with 1/2 deck deeper pen, a smaller spread and worse rules would not change much (which you can bet is at least somewhat flawed) making the above projection of T Count into Eric's work reasonable, and using Hiopt2 as 0 and perfect betting and playing as 100:
T Count would have a number of around 4.
A 6 deck shoe was the only info I had to try to make a comparison. I am sure Don was talking about DD, which would have a larger number on the scale of 0 to 100, as defined previously, since strong play is worth so much more in pitch games.
Last edited by Three; 12-11-2017 at 11:18 AM.
First you are skewing the comparison, as you state assume the percent increase with 1/2 deck deeper pen. The SCORE simulation is for 1 deck cut off. Second the SCORE for Hi-OPT II + ASC would be higher with 1/2 deck cut off. Where did you get the number $33.18/hr for perfect play?
You left out Norm's simulation:
HiOpt II + ASC CVData+CVCX 65.96 65.96
I said in my post that Eric did a perfect play and perfect betting sim and gave the link. That is where the base lines for my comparison came from. I also said that I was making an assumption that the percentage increase would be the same for T count over Hiopt2/ASC in Eric's sim as it is in Gronbog's sim that Tarzan posted. And I pointed out this assumption was flawed and it was just a matter to what degree because the penetration was better in Gronbog's sim but the rules were better in Eric's sim and Eric's sim used a larger spread, 1-12 rather than Eric's 1-16. This is probably because the larger spread was needed to make the lower EV approaches profitable enough in Eric's sim. This was the best comparison that could be made using these sims and you have to understand that your assumption is flawed and why. You can call it garbage in and garbage out if you like but I think it would be reasonably close to what sims with the same inputs would show. After all the difference between T Count and Hiopt2/ASC isn't that much in shoe games. If it were the assumptions that we know are flawed would cause more concern.
I purposely dismissed the sims from Norm's software because they were there only to validate the results from the other software. Adding in a sim using a third set of software didn't seem to make anything more certain. There was no perfect play with Norm's software but Eric did both perfect play and Hiopt2/ASC using his software. I felt including Norm's software's results would muddy the waters rather than make them clearer.
All I was trying to do was get some numbers on how far T Count is between Hiopt2/ASC and perfect play that are based on sim results. With Eric's work being the only thing I could find that makes a comparison between perfect play and counting I was forced to use that as a template for the comparison. It only did 6 deck shoe with one penetration level. I stated the assumptions the data was based on and pointed out that they are flawed assumptions. The degree they are flawed is debatable but I think they aren't going to skew things much. I don't see how that is confusing. Until someone does more sim work with Hiopt2/ASC versus perfect play that fits Tarzan's parameters better this is the best we can get.
Don can feel free to comment on the work I did.
My take about the effect of the assumptions. I see the lower pen of Eric's sim would hurt T count because the playing advantage would really start to get separation in that extra half deck. This would tend to make the number 4 on the 0 to 100 scale in Eric's sim as a slightly high estimate. Eric's sim also used a higher spread, 1-16 versus Tarzan's 1-12. Higher spreads can gain from both BC and PE. This may favor T Count if the betting is as strong as Hiopt2/ASC but my guess is the level 1 nature of T Count has it lacking a tad for betting and beating Hiopt2/ASC for playing making the increased spreads effect hard to define which count it favors so I am calling it a wash. Tarzan's sims used better rules which should favor T Counts stronger play. All in all that looks like the effects of the assumptions based on the sim differences would tend to cancel each other out to have little or no effect one way or another. But that is a guesstimate and should be taken with a grain of salt.
Again Tarzan presented a comprehensive comparison of many many combinations of games and ways of attacking them. The limitation caused by the sims I had to use only allowed me to compare one, which was probably one of the ones that T Count performed closest to Hiopt2/ASC.
Bookmarks