Glitch in CV Index Generator?

Below is a link to excellent HiLo indices posted on another message board using the CV Index Generator (6D, .75 Pen, SP to 4, No RSPA, 1 card to split aces, LS, DOA2, DAS, S17). Thanks to Southpaw for generating these indices and making them available to us.
While the vast majority of the indices make sense, there are a few cases where the recommended action appears to differ from the indices.
For example, in the Surrender Table, 8,8 vs X, the recommended play is Surrender >=1. However, the indices seem to indicate that it is better to surrender even at a TC of 0 (Surrender -0.50, Play -0.536691).
Other examples are Double 9 vs 3 and Double 8 vs 6.
Can anyone explain how the recommended play was determined based on the indices generated? Is there a glitch in the program or am I just misinterpreting something?

http://www.blackjackinfo.com/bb/showthread.php?t=20292

Are you sure you asked for an e.v.-maximizing index and not a risk-averse one?

Don

I understand from Southpaw that he asked for e.v. maximizing indices.

The issue can be illustrated by the following 2 examples below based on the output from the CV Index Generator.
In the case of 12 vs 3, the recommended index properly falls between the negative Diff (-0.010569 for +1) & the positive Diff (0.003427 for +2), generating a recommended index of +2.
Most of the tables in the CV Index Generator follow this pattern.

However, there are a few tables where this is not the case.
For Surrender 8,8 vs X, the e.v. for Play is worse at both 0 and +1 yet the recommended index is +1. This does not seem to make sense. It appears from the info below that there should be a negative Index # where the e.v. for Play is greater than -0.500000(e.g. -0.49xxxx) resulting in a negative Diff, and that the recommended index should be between this negative Diff & the next higher Index which has a positive Diff.

Surrender Table-8,8 vs. X: Surrender>=1
Index Surrender--Play----Diff
0 -0.500000 -0.536691 0.036691
1 -0.500000 -0.540229 0.040229

Hard Hit/Stand Table-12 vs. 3: Hit < 2
Index Stand-----Hit -----Diff
1 -0.243992 -0.233423 -0.010569
2 -0.231635 -0.235062 0.003427

I would say that the correct 8,8 v. 10 surrender index for Hi-Lo with six decks and DAS is +2. That's what I've always known. Meanwhile, the e.v.s that you show look very much like generic 16 e.v.s and not pair-of-eight 16s. Something is wrong.

The e.v. for playing 8,8 v. 10 at zero is surely not -0.5367. You're not using ENHC are you??

Don

FYI, following is Norm's response posted on message board at blackjackinfo:

Ahh, two things:

On extremely close calls, like 9v3, CVData will select the less risky decision, even if the Risk Averse indexes option is turned off. This is not the same as making RA calculations -- it's just a common sense decision.

On the 88 Split, this is a bug. The index is correct, but the numbers are displayed incorrectly. The porblem only affected 88 surrender, and oddly is part of the hole-carding code. I put up a fix.

MikeyD,

I believe the 88 vs 10 error is the mixing of numbers for the EV of the hand before the dealer has checked for blackjack and the EV of the hand after the dealer determines that he does not have an ace in the hole. At a neutral count, before the dealer checks, the EV of 88 vs 10 for surrender is about -0.5385 (lose � bet - 12 out of 13 times and 1 bet - 1 out of 13 times). The -0.50 would apply for early surrender.

As you can see, -0.5385 is between the values you listed (-0.53669 and -0.540229). Thus one should surrender at a count of 1 but should split at a count of zero.

Dr 21

Thanks Dr 21. Apparently that is what Norm was referring to when he said this bug is part of the hole-carding code.

I understand that the HiLo RA index for Surrender 8,8 vs 10 is -3.
I wonder if it makes sense to use this RA index or to stick with the EV index of +1.
Or do RA indices only make sense at a high count?
Note that in the table on page 375 of BJA3, the RA indices that differ from the EV indices are at TCs of +3 to +7.
Does it matter than the 8,8 vs 10 situation comes up when you are most likely betting only 1 unit, whereas the shaded RA indices on page 375 come up when you are probably betting 8 to 12(or more) units?

Mikey D,

IMO there is no reason to use RA indices for a 1 unit bet. For a zero or negative true count, I generally split my eights vs a dealer ten. A RA index might make sense for a very large bet (high count before the deal especially in a single or double deck game) with numerous tens dealt such the count was neutral by the time you need to act on your pair of eights. Then, if the bet was a large percentage of my bankroll, I would probably surrender instead of split based on RA considerations.

Dr 21