BJ Majish: CVCX and Benchmark SCORE

[BJ Majish]

Hi, everyone.

Having just upgraded to BJA3 and purchased a copy of CVCX, I'm excited to start experimenting. Before starting to plug in my own variations, I decided to see if I could replicate some of the numbers in BJA3 Chapter 10. Unfortunately, I'm having no luck at all. The numbers aren't even close. For example, for 1-Deck, S17, 31/52 pen, and 1-2 spread, using the canned CVCX sim with settings to match the book's (I think), I get 34.99 for the c-SCORE. The benchmark SCORE as given in the book is 41.41.

Reading through the CVCX Help, I found the following:

"Bankroll Growth - This is the default method of bet optimization. Bets are calculated to maximize bankroll growth. This is the same as calculating bets for the highest desirability index with a risk of ruin at about 13.5%. Instead of using the method in Blackjack Attack Chapter 10, a newer technique used in BJA Chapter 11 is used based on a paper written by Brett Harris in October 1997. This technique provides superior optimal bets."

It's possible that this is the reason for the difference, but if so, I'm surprised at the magnitude of the difference.

I also found the following in Help:

"CVCX does not use the same method as BJA Chapter 10. You can use the More button to change the calculation methodology to match BJA. But, the default method is the newer and more accurate method used to create BJA Chapter 11."

I can't seem to find this option. Can someone tell me where the More button is?

Any other ideas for what might explain the discrepancies?

Thanks!

[BJ Majish]

In investigating further, it appears that these Help sections might be outdated. The current Chapter 11 discusses team play, and there don't appear to be any SCORE calculations. I'm guessing that the reference is to the current Chapter 9.

On page 152 of Chapter 9, however, Don says this:

"Indeed, there was a better way. That new and improved methodology was used not only to create the studies in this chapter but also was adapted for the following chapter, as well."

Thus, it looks as though I'm back to square one. Why do my numbers not match those in the book? :-(

[Don Schlesinger]

The bet sizes in BJA3 seem to be about 11% higher than those in CVCX, but this would explain only about half of this rather large discrepancy.

Perhaps Norm has some further explanation.

Don

[Norm Wattenberger]

Ahh. I need to update the documentation. Blackjack Attack 3rd edition now uses the same techniques as CVCX since it was generated with CVCX. The documentation refers to the earlier version of the Chapter X tables which did not use optimal betting.

But the canned sims will still not match Blackjack Attack 3rd edition because they don't use the same indexes or True Count calculation assumptions. The canned sims use the indexes and methods in the respective books as the attempt is to compare how people generally play that have learned from the books. BJA3 uses a more modern set of indexes generated by CVData. Yes, in Single Deck games such differences can make a large difference.

Fortunately, CVCX allows you to specify exactly the TC calculation method and indexes that you personally use such that you can generate your own stats.

[BJ Majish]

Aha! Thanks for explaining this, Norm. It makes sense. I had just assumed that since CVCX was used to create the tables in the book, that the canned sims would match.

This does raise another question I wanted to ask you. Is there any way to see what indexes and strategies were actually used to create the canned sims? The file I've been working with is "Hi-Lo, II18 & Fab4." So it uses the same indexes as the book (although the book also adds 8 vs. 5 and 6 for single deck, and I don't know if you did this). So it appears that the main difference must be in the actual index values, correct? Any way to find out what they were?

Another example: In a 1999 Blackjack Forum article, Arnold Synder said:

"in all of these sims?WGBJS and risk-adjusted?the advanced version of the Red Seven system cannot be simulated as published in the 1998 edition of Blackbelt in Blackjack [editor?s note: the Advanced Red Seven can now be simulated using CVDATA, which was not available at the time this article was written]. In shoe games, for example, I provide index numbers for the Advanced Red Seven that are to be used only in the second half of the shoe. The player is advised to use only the six ?simple Red 7? indices in the first half of the shoe, then switch to the advanced indices for the second half."

So CVDATA will handle this; I'm curious to know whether that was done for the "Advanced Red7, Shoes" canned sim, or which Advanced Red 7 strategy was used.

Thanks to both of you for the reply.

[Norm Wattenberger]

All CVCX canned sims (except specifically noted) use exactly the indexes in the respective books - except that the Ill18 & Fab4 sims remove any indexes other than those 22. That's why there are sometimes multiple sims for the same strategy. Because there were multiple editions of the books.

I'd forgotten about Arnold's '99 BJF comment. (Although I probably was the source as I called him just after BiB2 came out to discuss a few points.) Yes CVData can specify different indexes for different parts of the shoe. But, this was not done in the canned sims. The difference is tiny in this particular case - which is why Arnold says not to bother with all the indexes early in the shoe. Although frankly I think it's just as easy to use the same indexes throughout the shoe.

Incidentally, in unbalanced strategies; it is quite likely that adjusting the running count halfway through the show will slightly improve results. But I've never done a study along these lines. John Auston may have looked at this.

[BJ Majish]

> All CVCX canned sims (except specifically
> noted) use exactly the indexes in the
> respective books - except that the Ill18
> & Fab4 sims remove any indexes other
> than those 22. That's why there are
> sometimes multiple sims for the same
> strategy. Because there were multiple
> editions of the books.

Not trying to be a smart aleck, but how do I know which "respective book"? As you point out, there are two editions of "Professional Blackjack" and three of "Blackbelt in Blackjack," and I think the high-low has been described in books by several authors. The download page says nothing about which book was used for each sim. (Maybe that information is on the CD; is it?)

> Yes CVData can
> specify different indexes for different
> parts of the shoe. But, this was not done in
> the canned sims. The difference is tiny in
> this particular case - which is why Arnold
> says not to bother with all the indexes
> early in the shoe.

Good point. The fact that the canned sim used all the indexes throughout probably didn't significantly improve the EV or the C-SCORE.

> Although frankly I think
> it's just as easy to use the same indexes
> throughout the shoe.

Yes. If you already have them learned, anyway....

> Incidentally, in unbalanced strategies; it
> is quite likely that adjusting the running
> count halfway through the show will slightly
> improve results. But I've never done a study
> along these lines. John Auston may have
> looked at this.

Yes, that's an interesting idea. I'd be very interested to know if anyone has done so.

Thanks again for your help.

[Norm Wattenberger]

> Not trying to be a smart aleck, but how do I
> know which "respective book"? As
> you point out, there are two editions of
> "Professional Blackjack" and three
> of "Blackbelt in Blackjack," and I
> think the high-low has been described in
> books by several authors.

Tables from both editions of PBJ are included and labeled as such. BiB1 and BiB2 Zen tables are included. For canned sims 1980 and 1998 Zen sims are included and labeled. HiLo pre- and post-1994 would show almost no difference. HiLo means Wong as that's what people use.

But when you get down to it - you need to run your own sims. Then you know you are simming exactly how you are playing. Frankly the books are nearly all fuzzy on exactly how TC is calculated in balanced counts. Renzy's book is an exception as he uses tables. I think Blackwood's new book also uses tables.

[BJ Majish]

> Tables from both editions of PBJ are
> included and labeled as such.

Where? At http://www.qfit.com/cvcxArchivesV3.htm there are 14 Hi-Lo files, but they appear to be differentiated by rules only, with no mention of edition or publication date.

> BiB1 and BiB2
> Zen tables are included. For canned sims
> 1980 and 1998 Zen sims are included and
> labeled.

Yes, that is the one system (and the only one) for which I see dates. On the other hand, there are three Red 7 tables, but no indication of whether they come from 1st, 2nd, or 3rd edition.

> HiLo pre- and post-1994 would show
> almost no difference. HiLo means Wong as
> that's what people use.

Okay.

> But when you get down to it - you need to
> run your own sims. Then you know you are
> simming exactly how you are playing.

I'm sure you're right. What I would like to be able to do, though, is to compare a few of my first efforts to the results of simulations that I know are accurate, just to be sure that I'm doing things right.

> Frankly
> the books are nearly all fuzzy on exactly
> how TC is calculated in balanced counts.
> Renzy's book is an exception as he uses
> tables. I think Blackwood's new book also
> uses tables.

I haven't seen either of their books. I'll have to look into them.

I appreciate the time you're taking to "unconfuse" me.

[Norm Wattenberger]

> Where? At
> http://www.qfit.com/cvcxArchivesV3.htm there
> are 14 Hi-Lo files, but they appear to be
> differentiated by rules only, with no
> mention of edition or publication date.

The strategy tables are included and labeled for both. All the sims are for the latest version of PBJ. I should have made that more clear.

> Yes, that is the one system (and the only
> one) for which I see dates. On the other
> hand, there are three Red 7 tables, but no
> indication of whether they come from 1st,
> 2nd, or 3rd edition.

Red7 sims are labeled 'Advanced Red7' indicating BiB2. (Actually Advanced Red7 was introduced in BJF between BiB1 and BiB2.)