Janhaus: Q regarding indices in BJAttack 3

[Janhaus]

Quick question for Don or anybody else out there that might be able to answer my question:

In BJAttack3, why are the indices between Illustrious18+Fab4 (p.62) & Catch 22 (p.375) different from the simulation indices (p.213)? Take 12v2 for instance, the index is 3 as shown in the Il18+Fab4 & Catch22 (Risk Averse or not), but in the simulation it is +4 for multideck and +5 for single deck. In a situation like this, which index is right?

(P.S. If it helps any, I'm aiming to find the right index for 6deck, das, ls, h17)

[Don Schlesinger]

> Quick question for Don or anybody else out there that
> might be able to answer my question:

> In BJAttack3, why are the indices between
> Illustrious18+Fab4 (p.62) & Catch 22 (p.375)
> different from the simulation indices (p.213)? Take
> 12v2 for instance, the index is 3 as shown in the
> Il18+Fab4 & Catch22 (Risk Averse or not), but in
> the simulation it is +4 for multideck and +5 for
> single deck. In a situation like this, which index is
> right?

They were all "right" in their time. Wong's came first, and he used his software, his assumptions about rules, numbers of decks (4 in this case), etc. Next came Cac's Catch 22, and he used his software and his assumptions. (I believe both Wong and Cac truncated, whereas Norm floored, but with a positive index, it makes no difference, so that's not a factor.) Finally, Norm used CV Data, 6-decks, and flooring. I trust his findings, although I suspect that using +3 instead of +4 won't make the slightest difference for the rest of your life. :-)

> (P.S. If it helps any, I'm aiming to find the right
> index for 6deck, das, ls, h17)

None of the above rules (das, ls, or h17) has any bearing on determining the index for hard 12 v. 2 (presumably, h17 could, but it turns out that the s17 and h17 indices are the same).

Don

[Francis Salmon]

All the reliable sources agree that the index for 12v2 is +3 and this matches my own findings (precise index 3.0).Indices don't change over time.They are either right or wrong.
I fear that Norm's +4 has more to do with oversimplistic assumptions (or unacceptable prescriptions!) about the way we humans calculate TCs wrongly than with any real progress in BJ-science.
I have always stated that sims are not the right way to establish indices.This obvious outlayer here is one more proof to it.

Francis Salmon

> Quick question for Don or anybody else out there that
> might be able to answer my question:

> In BJAttack3, why are the indices between
> Illustrious18+Fab4 (p.62) & Catch 22 (p.375)
> different from the simulation indices (p.213)? Take
> 12v2 for instance, the index is 3 as shown in the
> Il18+Fab4 & Catch22 (Risk Averse or not), but in
> the simulation it is +4 for multideck and +5 for
> single deck. In a situation like this, which index is
> right?

> (P.S. If it helps any, I'm aiming to find the right
> index for 6deck, das, ls, h17)

[Norm Wattenberger]

1. Indexes depend on assumptions about penetration and deck estimation accuracy. There is not one "right" index for all cases.

2. The Index for 12v2 is +3 when estimating deck depth exactly and +4 when estimating to half decks. Half decks is more realistic and is why I choose +4. Almost noone estimates deck depth exactly.

3. CVData removes the dealer up card and player cards when calculating Hit/Stand indexes. Most simulators do not. It also uses full simulation, not the less accurate "sampling by replacement" used by some others, including Stanford's original indexes. Sampling was required in those days because PCs were extremely slow compared to today's.


CVCX Online

[Don Schlesinger]

Thanks for the further clarifications, Norm. You are, of course, correct in every instance.

Janhaus, Francis means well, but in this case, what he wrote above was utter nonsense. The indices on p. 213, and the simulations that follow, were generated according to the principles outlined on pp. 211-13. Too many people jump right to the data without reading the important information that precedes the charts.

As Norm points out, virtually no player reckons true count in shoe games to any precision greater than half-deck. That's what was used to generate the 12 v. 2 +4 index -- and all the other multi-deck ones. (Note: quarter-deck precision was used for single- and double-deck TC generation and indices; see p. 211, number 6.)

Indices must be generated according to the manner in which you are actually going to use them when you play in a casino. Francis deludes himself into thinking that he has more precision than he does, when he plays, but that's his problem, not ours.

We've been down this path before with him, so when he responds -- which he will, as sure as the sun rises in the east! -- it is likely that Norm and I will ignore him. I suggest you do likewise.

Please write back if you have further questions.

Don

[Janhaus]

Thanks a lot for all of the replies, and I now have a clearer understanding of how everything works. On a related thought, DS, how come there isn't a (complete) table of indices in your book for a few commonly seen decks? I have read the chapter on how two dozen indices (Illustrious 18, etc.) capture most of the advantages of counting, but what about for those of us who wouldn't mind remembering a few more numbers?

> Thanks for the further clarifications, Norm. You are,
> of course, correct in every instance.

> Janhaus, Francis means well, but in this case, what he
> wrote above was utter nonsense. The indices on p. 213,
> and the simulations that follow, were generated
> according to the principles outlined on pp. 211-13.
> Too many people jump right to the data without reading
> the important information that precedes the charts.

> As Norm points out, virtually no player reckons true
> count in shoe games to any precision greater than
> half-deck. That's what was used to generate the 12 v.
> 2 +4 index -- and all the other multi-deck ones.
> (Note: quarter-deck precision was used for single- and
> double-deck TC generation and indices; see p. 211,
> number 6.)

> Indices must be generated according to the manner in
> which you are actually going to use them when you play
> in a casino. Francis deludes himself into thinking
> that he has more precision than he does, when he
> plays, but that's his problem, not ours.

> We've been down this path before with him, so when he
> responds -- which he will, as sure as the sun rises in
> the east! -- it is likely that Norm and I will ignore
> him. I suggest you do likewise.

> Please write back if you have further questions.

> Don

[Don Schlesinger]

> Thanks a lot for all of the replies, and I now have a
> clearer understanding of how everything works.

Glad to hear that.

> On a
> related thought, DS, how come there isn't a (complete)
> table of indices in your book for a few commonly seen
> decks? I have read the chapter on how two dozen
> indices (Illustrious 18, etc.) capture most of the
> advantages of counting, but what about for those of us
> who wouldn't mind remembering a few more numbers?

Since Hi-Lo was the count of choice for most of the studies in BJA3, I found it superfluous to include any more indices, when Wong's "Pro BJ" has them all for Hi-Lo, in his Appendexes.

I don't have more because it became clear to me that, past the I18, especially for shoe games, the rest aren't worth very much. Nonetheless, you have the next four important ones, which were added by Cacarulo to form the "Catch 22," namely doubling 8 and A,8 v. 5 and 6.

Don

[Francis Salmon]

Janhaus, Don has told you not to listen to me but I think you're perfectly capable of making your own judgement based on reason and logics.
I have no reason to believe that the Don's Ill18 are not the most important ones but it's not a big deal to memorize a few more. Actually I myself use a lot more.
Quite an important index which is not included by Don is the one for spitting 9/9 vs. 7.Under DAS it's +3 (+3.1 exactly).
This is just one example but I can give you more on request or answer specific questions about indices.

Now a general remark. You might have noticed that the promotors of this site have quite a dogmatic view about index generation and index use.As a professional player,I have a different approach:
TC is defined as the RC divided by the number of decks remaining. The result is a number which reflects deck distribution regardless of the number of decks used or the level of penetration (that's why we didide by the number of decks).For the index generation, the initial hand causes a slight distortion especially for one deck but for all shoe games this is negligible. This means that we can safely use one single index for any shoe games with a given set of rules.There is no reason to skip decimals, not in the index itself and not in the TC calculation.The deeper you are in the shoe the more important the deck estimation but even if you are able only of half deck accuracy this will do.There are some short cuts for calculation that you need to know.For instance if there are 2.5 decks left, you shouldn't divide your RC by 2.5 but simply multiply by 4 and put the decimal point. So if the RC is +8, your TC will be +3.2.Since this number is higher than the index for 9/9 v. 7 you know immediately you should split.
Wasn't too difficult,was-it?
The basic idea is simple. The more precision the better!
I look forward to Don's bashing!

Francis Salmon

[MGP]

> Quite an important index which is not included by Don
> is the one for spitting 9/9 vs. 7.Under DAS it's +3
> (+3.1 exactly).

Here you seem to be advocating using an exact index, and then you do the same again here:

> So if the RC
> is +8, your TC will be +3.2.Since this number is
> higher than the index for 9/9 v. 7 you know
> immediately you should split.
> Wasn't too difficult,was-it?
> The basic idea is simple. The more precision the
> better!

And yet, then you go and say:

> For the index generation, the initial
> hand causes a slight distortion especially for one
> deck but for all shoe games this is negligible.

Hmmm. See below.

> This
> means that we can safely use one single index for any
> shoe games with a given set of rules.There is no
> reason to skip decimals, not in the index itself and
> not in the TC calculation.

Hmmm again... In the same sentence you say use one general index for ANY number of decks with ANY rules in a shoe game but, "There is no reason to skip decimals." I'm sorry, I don't get it - you can't have it both ways - accuracy and generalization don't mix.

Let's go back to that other statement:

> For the index generation, the initial
> hand causes a slight distortion especially for one
> deck but for all shoe games this is negligible.

Now let's look at the very simple case of insurance. If you use a general, e.g. floored index, this is true. For shoes with 4-8 decks, the main exception is AA which has an insurance index of 2 for 4,6 decks but 3 for 8 decks. All other hands have a floored index of 3.

But if you look at the exact indices to just 1 decimal point, then we have a different story. In 4 decks the exact indices range from 1.9 to 3.3. In 6D and 8D the range is from 2.4 to 3.3 and 8 Deck it's 2.6 to 3.3.

If it's sooo important to calculate your TC to one decimal place, why is it sooo unimportant to use that information? I'm sorry but I just don't get why you're bothering with the TC estimation to 1 decimal point. What exactly are you gaining, besides the ego boost and showoff factor just because you can estimate it with that accuracy?

Not to mention that this is just for insurance. How in the world can you honestly expect the rules and number of decks to have a negligible effect on specific hand indices to 1 decimal point?

The data for the insurance calcs is below. My numbers are for Hi-Lo, CA generated and are as close to exact as you'll ever find:

 

1 Deck 

Hard Hands	General		Exact  

2, 3		2		1.925926  

2, 4		2		1.925926  

2, 5		2		1.925926  

3, 4		2		1.925926  

2, 6		2		1.925926  

3, 5		2		1.925926  

2, 7		1		1.106383  

3, 6		2		1.925926  

4, 5		2		1.925926  

2, 8		1		1.106383  

3, 7		1		1.106383  

4, 6		2		1.925926  

2, 9		1		1.106383  

3, 8		1		1.106383  

4, 7		1		1.106383  

5, 6		2		1.925926  

2, 10		2		2.418605  

3, 9		1		1.106383  

4, 8		1		1.106383  

5, 7		1		1.106383  

3, 10		2		2.418605  

4, 9		1		1.106383  

5, 8		1		1.106383  

6, 7		1		1.106383  

4, 10		2		2.418605  

5, 9		1		1.106383  

6, 8		1		1.106383  

5, 10		2		2.418605  

6, 9		1		1.106383  

7, 8		1		1.106383  

6, 10		2		2.418605  

7, 9		1		1.106383  

7, 10		1		1.333333  

8, 9		1		1.106383  

8, 10		1		1.333333  

9, 10		1		1.333333  

		  

Soft Hands	General		Exact  

A, 2		1		1.106383  

A, 3		1		1.106383  

A, 4		1		1.106383  

A, 5		1		1.106383  

A, 6		1		1.106383  

A, 7		1		1.130435  

A, 8		1		1.130435  

A, 9		1		1.130435  

A, 10		1		1.181818  

		  

Pairs		General		Exact  

A, A		-2		-2.418605  

2, 2		2		1.925926  

3, 3		2		1.925926  

4, 4		2		1.925926  

5, 5		2		1.925926  

6, 6		2		1.925926  

7, 7		1		1.106383  

8, 8		1		1.106383  

9, 9		1		1.106383  

10, 10		3		3.058824  

 

4 Deck 

Hard Hands	General	Exact 

2, 3		3		2.983606557 

2, 4		3		2.983606557 

2, 5		3		2.983606557 

3, 4		3		2.983606557 

2, 6		3		2.983606557 

3, 5		3		2.983606557 

2, 7		3		2.754966887 

3, 6		3		2.983606557 

4, 5		3		2.983606557 

2, 8		3		2.754966887 

3, 7		3		2.754966887 

4, 6		3		2.983606557 

2, 9		3		2.754966887 

3, 8		3		2.754966887 

4, 7		3		2.754966887 

5, 6		3		2.983606557 

2, 10		3		3.127819549 

3, 9		3		2.754966887 

4, 8		3		2.754966887 

5, 7		3		2.754966887 

3, 10		3		3.127819549 

4, 9		3		2.754966887 

5, 8		3		2.754966887 

6, 7		3		2.754966887 

4, 10		3		3.127819549 

5, 9		3		2.754966887 

6, 8		3		2.754966887 

5, 10		3		3.127819549 

6, 9		3		2.754966887 

7, 8		3		2.493150685 

6, 10		3		3.127819549 

7, 9		3		2.493150685 

7, 10		3		2.909090909 

8, 9		3		2.493150685 

8, 10		3		2.909090909 

9, 10		3		2.909090909 

		 

Soft Hands	General	Exact 

A, 2		3		2.491017964 

A, 3		3		2.491017964 

A, 4		3		2.491017964 

A, 5		3		2.491017964 

A, 6		3		2.491017964 

A, 7		2		2.197183099 

A, 8		2		2.197183099 

A, 9		2		2.197183099 

A, 10		3		2.618705036 

		 

Pairs	General	Exact 

A, A		2		1.937888199 

2, 2		3		2.983606557 

3, 3		3		2.983606557 

4, 4		3		2.983606557 

5, 5		3		2.983606557 

6, 6		3		2.983606557 

7, 7		3		2.493150685 

8, 8		3		2.493150685 

9, 9		3		2.493150685 

10, 10		3		3.275590551 

 

 

6 Deck 

Hard Hands	General		Exact  

2, 3		3		3.108696  

2, 4		3		3.108696  

2, 5		3		3.108696  

3, 4		3		3.108696  

2, 6		3		3.108696  

3, 5		3		3.108696  

2, 7		3		2.948454  

3, 6		3		3.108696  

4, 5		3		3.108696  

2, 8		3		2.948454  

3, 7		3		2.948454  

4, 6		3		3.108696  

2, 9		3		2.948454  

3, 8		3		2.948454  

4, 7		3		2.948454  

5, 6		3		3.108696  

2, 10		3		3.263598  

3, 9		3		2.948454  

4, 8		3		2.948454  

5, 7		3		2.948454  

3, 10		3		3.263598  

4, 9		3		2.948454  

5, 8		3		2.948454  

6, 7		3		2.948454  

4, 10		3		3.263598  

5, 9		3		2.948454  

6, 8		3		2.948454  

5, 10		3		3.263598  

6, 9		3		2.948454  

7, 8		3		2.769231  

6, 10		3		3.263598  

7, 9		3		2.769231  

7, 10		3		3.072727  

8, 9		3		2.769231  

8, 10		3		3.072727  

9, 10		3		3.072727  

		  

Soft Hands	General		Exact  

A, 2		3		2.75  

A, 3		3		2.75  

A, 4		3		2.75  

A, 5		3		2.75  

A, 6		3		2.75  

A, 7		3		2.611872  

A, 8		3		2.611872  

A, 9		3		2.611872  

A, 10		3		2.84153  

		  

Pairs		General		Exact  

A, A		2		2.394737  

2, 2		3		3.108696  

3, 3		3		3.108696  

4, 4		3		3.108696  

5, 5		3		3.108696  

6, 6		3		3.108696  

7, 7		3		2.769231  

8, 8		3		2.769231  

9, 9		3		2.769231  

10, 10		3		3.284211  

 

 

 

8 Deck 

Hand Indices		 

Hard Hands	General		Exact 

2, 3		3		3.157894737 

2, 4		3		3.157894737 

2, 5		3		3.157894737 

3, 4		3		3.157894737 

2, 6		3		3.157894737 

3, 5		3		3.157894737 

2, 7		3		3.068852459 

3, 6		3		3.157894737 

4, 5		3		3.157894737 

2, 8		3		3.068852459 

3, 7		3		3.068852459 

4, 6		3		3.157894737 

2, 9		3		3.068852459 

3, 8		3		3.068852459 

4, 7		3		3.068852459 

5, 6		3		3.157894737 

2, 10		3		3.261324042 

3, 9		3		3.068852459 

4, 8		3		3.068852459 

5, 7		3		3.068852459 

3, 10		3		3.261324042 

4, 9		3		3.068852459 

5, 8		3		3.068852459 

6, 7		3		3.068852459 

4, 10		3		3.261324042 

5, 9		3		3.068852459 

6, 8		3		3.068852459 

5, 10		3		3.261324042 

6, 9		3		3.068852459 

7, 8		3		2.902325581 

6, 10		3		3.261324042 

7, 9		3		2.902325581 

7, 10		3		3.107569721 

8, 9		3		2.902325581 

8, 10		3		3.107569721 

9, 10		3		3.107569721 

		 

Soft Hands	General	Exact 

A, 2		3		2.899628253 

A, 3		3		2.899628253 

A, 4		3		2.899628253 

A, 5		3		2.899628253 

A, 6		3		2.899628253 

A, 7		3		2.769230769 

A, 8		3		2.769230769 

A, 9		3		2.769230769 

A, 10		3		2.977099237 

		 

Pairs	General	Exact 

A, A		3		2.623853211 

2, 2		3		3.157894737 

3, 3		3		3.157894737 

4, 4		3		3.157894737 

5, 5		3		3.157894737 

6, 6		3		3.157894737 

7, 7		3		2.902325581 

8, 8		3		2.902325581 

9, 9		3		2.902325581 

10, 10		3		3.305084746 

[Francis Salmon]

First, I have to clarify something. I meant to say 'For a given rule set we can safely use the same indices for any shoe game (4 or more decks)'.Of course, we have to use different indices for different rules.For instance splitting indices for DAS are totally different from the ones for NDAS.
Your composition dependent indices are very interesting but they just prove my point.
I recognized distortions in single deck, but they are minor for multideck.Of course you took the case with the clearly most important impact of composition in hilo:insurance.
But you do not seriously expect us to use a different index for each initial hand.So we have to use a single index for all compositions which should be the weighed average of these.I bet you that the result of this would be almost identical to the one that didn't take composition into acount at all (within decimal range for most cases).And this is the index I use (nothing to with egoboost or the like).Even if there might be some situations where the general infinite deck index given to the decimal point differs from the exact index for the specific situation this error will always be less than the one caused by rounding to whole numbers which is an inherent potential error of 0.5 TC.
As to calculation,I don't look for decimals. They are just there. Or what is 7 divided by 2? It's neither 3 nor 4,it's 3.5.
Of course we need to make a compromise between precision and practicability while we are playing, but nobody can make me believe that deliberate imprecision on all counts will produce a good endresult.
You didn't analyse the case I presented. So I assume that your index for 9/9 vs 7 matches mine.

Francis Salmon

[MGP]

> First, I have to clarify something. I meant to say
> 'For a given rule set we can safely use the same
> indices for any shoe game (4 or more decks)'.Of
> course, we have to use different indices for different
> rules.For instance splitting indices for DAS are
> totally different from the ones for NDAS.

Ok, then that makes more sense.

> But you do not seriously expect us to use a different
> index for each initial hand.So we have to use a single
> index for all compositions which should be the weighed
> average of these.I bet you that the result of this
> would be almost identical to the one that didn't take
> composition into acount at all (within decimal range
> for most cases).

The floored TC index is 3 for all cases but varies form 2.9 for 4Deck to 3.1 for 8D with variations depending on the penetration. Cacarulo made a nice table on BJMath:

http://www.bjmath.com/bin-cgi/bjmath...ames;read=5688

So if you are saying you should use decimal places, then they are not the same.

> And this is the index I use (nothing
> to with egoboost or the like).Even if there might be
> some situations where the general infinite deck index
> given to the decimal point differs from the exact
> index for the specific situation this error will
> always be less than the one caused by rounding to
> whole numbers which is an inherent potential error of
> 0.5 TC.

These indices are not infinite deck calcs. They use all subsets of the relevant counts for finite decks. But here you are making my point. The inherent mistakes in TC estimation and practical usage makes the idea of using indices with decimal points unecessary.

> As to calculation,I don't look for decimals. They are
> just there. Or what is 7 divided by 2? It's neither 3
> nor 4,it's 3.5.
> Of course we need to make a compromise between
> precision and practicability while we are playing, but
> nobody can make me believe that deliberate imprecision
> on all counts will produce a good endresult.

But my point is that deliberate Precision on some counts doesn't really make a difference - so why argue for it when it can make life more confusing and difficult, especially for non-professional counters?

> You didn't analyse the case I presented. So I assume
> that your index for 9/9 vs 7 matches mine.

Actually my CA only does insurance calcs at this time so I don't have one, but if Cacarulo ever published one I'm sure it's the most accurate so you might want to see if he did.

[Francis Salmon]

Insurance is about the only case in which I don't use the infinite deck index (3.3).It would be too far from the truth because of this bloody ace that has been counted the wrong way. I use 3.0 which is the exact index for 6-decks and happens to be a good average for all multideckgames.
Thanks for the link to Cacarulo's table.I think highly of him.
I still don't see why errors in TC-calculation should make decimal indices unnecessary.I just rediscovered an interesting article from Richard Reid about Index generation and usage on bjmath.com which atually confirms the validity of my approach.
This might be a subjective view but I find this whole business with flooring,truncating,rounding,rounding up,rounding down etc. totally confusing.For me it's easier to remember clearcut decimal indices and just keep the decimal from the TC-calculation. On top of it, it's more precise.

Francis Salmon