Best reference for EV-maximizing indices

[Gronbog]

I'm not seeing your rules and conditions anywhere in this tread and the image of your results is too fuzzy for me to read.

[Chuckles!]

Here are the rules and conditions for my game, hope this helps! (if I left any out, feel free to ask):

3:2 BJ payout
6D
52 card pen
H17
DAS
No RSA
LS
SP4
No ENHC
Simulation played heads-up with only 1 hand the entire time.
Always play until end of shoe (will not wong or end simulation mid-shoe)
Burns 1 card at top of each shoe (not added to count)

I'll try putting the results in a public google sheet, that should have better resolution. Let me know if there are problems accessing the link: https://docs.google.com/spreadsheets...it?usp=sharing

Note that my index range is -9 to +10, so indexes with those values are likely rounded and greyed out.

Format for index chart:

H [1+:S]3.17

H = Basic strategy play (S = stand, H = hit, D = double, DS = double or stand, P = split, F = surrender)
[1+:S] = index. (stand at tc1 or above, 0/neg indices are rounded downwards)
3.17 = rough accuracy. Calculated by taking the EV difference of top 2 actions in the closest decision, and dividing by the sum of each action's std err. It's not the best method, and I plan to change it. But if I'm understanding correctly, then it should give a lower value than Dr. Ethier's formula.

Values are compared against Table 1.1 in the Hi-Lo book. Only 10v10 has been corrected.

[Chuckles!]

In the book, the estimate of remaining decks is rounded to the nearest half deck. That is, the RC can be divided by 6, 5.5, 5, 4.5, 4, 3.5, 3, 2.5, 2, 1.5, 1.
Then the division is rounded down (floored).

Also, no deck alterations have been used, and estimation/tc conversion follows Cacarulo's advice

[G Man]

What is the correct reading of "No ENHC"

[Freightman]

What is the correct reading of "No ENHC"

Believe your referring to European No Hole Card

[Freightman]

What is the correct reading of "No ENHC"

Believe your referring to No European No Hole Card

[DSchles]

Believe YOU'RE referring ...

Don

[G Man]

Believe your referring to No European No Hole Card

LOL, I know perfectly well what ENHC means, it's the term "No ENHC" that is funny, something like "No S17" instead of "H17"

[Gronbog]

Agreed, I do plan to run it for 200 billion at some point. But since I'm still seeing repeated patterns of errors (13/14/15/s16/s17/s18/s19 vs 2/3/4 is consistently high by 1)

The reason I asked for your rule set is that I wanted to check these plays against the raw data from the sims I did for the HLCCS book.

• We did not publish an index for A5 vs 2 because the strike point occurs too rarely to be of any consequence. This would make it difficult to converge.
• Of the other plays above, several were close calls (13 vs 3, 14 vs 3, 15 vs 2, A5 vs 3,4, A6 vs 2,3, A7 vs 2,3,4) and may not converge after only 20 billion rounds.
• Of these only A6 vs 3 failed to converge after 200 billion rounds.

I suggest simulating with the full 200 billion rounds to see if any of these converge before looking blindly through your code for errors.

[Chuckles!]

Ok, sounds good. I should have results for 200B by next Wednesday.

[Chuckles!]

Looks like my results are in earlier than I thought. I've added some sheets here (200 billion round index chart + full EV data across all upcard/hand/tc combinations): https://docs.google.com/spreadsheets...gid=1585507930

Reminder that the accuracy score (number after each index) is only a rough approximation and is most likely less than the standard errors produced by Dr. Ethier's calculation. StdErr listed for each individual EV on the upcard charts is calculated using 16 samples of 12.5 billion. Next simulation, I plan to have more reliable numbers to estimate accuracy.

Aside from 3 indices that are off by more than 1, all of the errors have a rough accuracy of <4. I'm sure some are the result of lack of convergence, but it seems to me there's still a clear pattern of slight overestimation for the hard and soft hands, and some very strange things going on in the splits.

[Chuckles!]

Dave, do you happen to have the house edge calculated for the simulations you ran? (using BS or indices) I'm trying to find if my error is in the game simulation, or how I'm handling the results. If the BS HE matches, I believe that would indicate the later.

The values I've calculated for my sim are:
-.564% BS HE
-.403% HE w/indices

[Chuckles!]

Ok, good news! I tried simulating both sets of indices in CVCX to get the house edge from an independent source. (10 billion rounds each, which should give a standard error of 0.00115% if I'm not wrong)

Basic Strategy: -0.556%
HLCCS Indices: -0.405%
My Indices: -0.401%

This is NOT to say that my indices are more accurate. I think what happened is that there's some microscopic configuration difference and the CVCX sim I set up happens to match my config better. But it does seem to indicate that there aren't any significant issues with my code which is mainly what I was trying to verify. Please do let me know if I made a mistake somewhere, and thanks again to everyone who helped me out!