qboy: Q for Norm et al. - calculated vs. published inds

qboy · 06-25-2003, 02:18 PM

I've noticed when I calculate Hi-opt II indices using simulation software, I get different values than those published elsewhere. My calculated values tend to be smaller in absolute value...I don't think this is a divisor issue (I am using per deck values?).

For example, CV indicates Hi-OptII indices for 12vX as (5 2 0 -2 -5) where I get (4 3 1 0 -2). The pattern is similar throughout all tables. Am I doing something wrong? Are the published indices adjusted for risk aversion?

Norm Wattenberger · 06-25-2003, 06:27 PM

Several possible answers. First, what software are you using and what are the settings? As for the books, the indexes in various books were generated in various manners. Some include a bit of risk aversion (AO-II.) Some are compromise indexes for simplicity (KO, Hi-Lo Lite.) Some are simply not precise (ProBJ Hi-Lo.) And the assumptions on deck resolution and penetration, both of which can affect indexes, are unknown.

Serious Blackjack Software

Don Schlesinger · 06-26-2003, 09:59 AM

. . . truncating, rounding, flooring issues.

Don

Norm Wattenberger · 06-26-2003, 10:31 AM

> . . . truncating, rounding, flooring issues.

Yeah that one confused a lot of people when the 1994 edition of ProBJ didn't use the same technique as the 1981 edition of the same book.

BJStats

qboy · 06-26-2003, 01:28 PM

Software used is SBJ. Indices in question are hi-optII from CVDATA.

I thought of the round/truncate/floor issue, but that doesn't explain the pattern I'm seeing.

I'm trying to see how penetration has an impact on index calculation...could you give me an example of that?

> Several possible answers. First, what
> software are you using and what are the
> settings? As for the books, the indexes in
> various books were generated in various
> manners. Some include a bit of risk aversion
> (AO-II.) Some are compromise indexes for
> simplicity (KO, Hi-Lo Lite.) Some are simply
> not precise (ProBJ Hi-Lo.) And the
> assumptions on deck resolution and
> penetration, both of which can affect
> indexes, are unknown.

Norm Wattenberger · 06-26-2003, 06:07 PM

I assume that you mean SBA. SBA and CVData will generate indexes that are nearly always identical. There are some slight differences due to the fact that CVData takes into account resplits, possible differences in handling burn cards, the CVData option of varying deck estimation resolution by shoe area, and occasional differences in extremely close indexes due to methodology and standard error. But these differences don?t matter. The calculation of absolutely exact indexes is not possible with current technology. But both packages provide indexes that are accurate enough and more accurate than any other available commercial packages. (If I had to bet, I?d say Caccarulo?s private software is the most accurate.)

The indexes supplied in all CV products for Hi-Opt II were not generated by CVData. They are the indexes supplied by the original author works. My point in including the original author?s indexes is that?s what most people use. You can generate your own indexes if you wish.

CVData supplied Hi-Opt II indexes were generated by Julian Braun in 1976 (supplied with permission from Dr. Humble.) Braun?s work was simply amazing considering the time and the hardware used (IBM 709.) I am very familiar with the IBM 709 as back in the 60s I had considerable experience with 7040s and 7044s which had the same instruction set. These were million dollar machines filling large rooms. Yet, they had less power than chips designed for use in today?s cell phones. The indexes he created are perfectly useable today.

As for penetration; accurate indexes vary by area of the shoe. This is obvious with unbalanced strategies. But, it occurs to a lesser extent with true counting strategies. The effect is small in TC strategies, but could change an extremely close index. There are many such close calls. In fact, it?s possible that a back-counter should adjust his indexes since he doesn?t play the first part of a shoe. But, for true counted shoe strategies ? the effect is so minor as to be of only academic interest.

Serious Blackjack Software

AsZehn · 06-26-2003, 07:05 PM

Which indices are correct for Hi Lo ( 6dk & 8dk assuming floor); PBJ '81, PBJ '94, BJA Chapter 10, BJA Chapter 12 or none of the above? TIA

AZ

Don Schlesinger · 06-26-2003, 07:14 PM

> Which indices are correct for Hi Lo ( 6dk
> & 8dk assuming floor); PBJ '81, PBJ '94,
> BJA Chapter 10, BJA Chapter 12 or none of
> the above? TIA

The BJA Chapter 12 indices on p. 317 are the ones generated by the most recent version of SBA and are those incorporsted into BJRM's Systems 101 (except for the 2/-2 typo for 12 v. 5). These are floored, and I believe that they should be the most accurate.

Don

Norm Wattenberger · 06-27-2003, 05:29 AM

> Which indices are correct for Hi Lo ( 6dk
> & 8dk assuming floor); PBJ '81, PBJ '94,
> BJA Chapter 10, BJA Chapter 12 or none of
> the above? TIA

As Don says, BJA Chaper 12. I believe the PBJ indexes were created using sampling with replacement. This is a very fast method of index generation. But, less accurate than true simulation.