CVDATA: Trouble reconciling IBA calculations

[RustyNCrusty]

I’m not able to correlate IBAs from CVDATA sims to published IBAs. The IBAs correlate closely for dealer up cards 2-9 but then differ widely with dealer up cards of A and T. Here’s a brief example using the hand A,2 to avoid TD vs CD issues. Sim run for 6D S17 DOA DAS RSP4 nRSA nLS Flat Bet + Complete Basic Strategy using 2e9 rounds.

BJA3 Tables A50, A58, A59 for 6D S17, EVs from table are for Basic Strategy

A,2 v A -0.059450
A,2 v 10 -0.102342
A,2 v 9 -0.034089

Now, from the CVDATA sim:
A,2 v A -0.35043
A,2 v 10 -0.1686
A,2 v 9 -0.03365

All IBAs for 2 -9 show less than a 2% difference with the BJA3 tables but, as can be seen, are not close at A and T. I attempted to work through the probabilities associated with a possible dealer BJ but was unable to reconcile the differences. I’m stumped.

[Norm]

The purpose of the BJA tables is determination of basic strategy. I would imagine dealer BJs were ignored as they wouldn't affect results without ENHC. This would also explain why the Cumm. Freq. columns in BJA have lower numbers with dealer ace-up hands. The CVData tables are general purpose and include dealer BJs.

Of course, the numbers will not match within std. err. due to the cut card effect.

[RustyNCrusty]

Good point on ENHC. I was able to find the error in my calculations...brain fade on my part. As you point out the published EVs ignore dealer BJs, but if adjusted for dealer BJ the EVs line up with CVDATA results. Adjusting the published EVs to include the probability of dealer blackjack:
Evadj=EV x (1-Pbj) -1xPbj
For the rules mentioned above, and for A,2 vs A, the probability of 10 under is: Pbj=6x16/(312-3)
Evadj = -0.05945 x (1-96/309)-96/309 = -.3517 , vs. CVDATA = -0.35044
Similarly for A,2 vs T
Evadj = -0.102342 x (1 –(24-1)/309) – (24-1)/309 = -0.169, vs. CVDATA = -0.1696
Note, the EVs in BJA3, Wizard of Odds, and Cacarulo are the same, the frequencies are not. Cacarulos frequencies include dealer blackjacks but BJA3 and Wizard of Odds don’t…I was a bit confused by this at first.
So, why am I mucking around with Basic Strategy comparisons? I’m working on ranking the index plays for various game conditions similar to Schlesinger’s I18. Basic startegy EVs are used to compare return gains for each hand – absent bet ramps. The basic strategy comparisons are also used to check and verify my calculation methods. Earlier I had started down the path of attempting to use the method outlined by Schlesinger ( per Griffin) however, I was not able to rationalize the weighting method used on p59. I’ve opted to use a more straight forward sim approach with CVDATA even if slightly more time consuming.
Again, thank you for your response.