BJ Majish: Question About Index Generation

[BJ Majish]

Hi, all.

I've been arguing with myself over which of the three DD forums would be most appropriate for this post--I think an argument could be made for all three--and finally decided on the most general forum. I assume that, if necessary, it can be moved.

I'm trying to generate the missing K-O Catch 22 indexes, using the CVIndex component of CVData, and several issues and questions have arisen. I'm sure I'll be able to get some good help here. (I'm using the IRCs as given in the book.)

1. As helpful as the group indexes are in the new Qfit programs, it appears that CVIndex does not understand them. As a result, it seems that I will have to create separate strategies for 1, 2, 4, 6, and 8 decks and run a separate simulation for each. Correct?

2. What philosophy is used when creating a generic strategy for all decks and conditions, as with K-O (other than dropping some of the indices for multiple decks, of course)? Is it best to use the 4-deck or 2-deck indices as most representation of all deck variations? Or should the highest indices be used to avoid expensive errors? What about differences between S17 and H17, or DAS?

3. What's the best way to apply the "reduced and rounded" philosophy to the resulting indices? Is it best to just round them up to the next highest group index?

4. In the case of index numbers such as 6 and 7 (which I got in a preliminary sim), I can see only two ways to maintain the reduced/rounded groups. The first would be to "round" these numbers down to the pivot point of four; the second would be to create a new, fourth group (probably with an index of 6 or 8) to contain the new indices. Is this a valid assumption?

Thanks in advance for any help you can give me in understanding this process!

[Norm Wattenberger]

> 1. As helpful as the group indexes are in
> the new Qfit programs, it appears that
> CVIndex does not understand them. As a
> result, it seems that I will have to create
> separate strategies for 1, 2, 4, 6, and 8
> decks and run a separate simulation for
> each. Correct?

Group Indexes are used for compromise indexes. Compromise indexes really cannot be created without human intervention as they are art as much as science. So, yes you need to run separate sims and then create your own Group Indexes if you wish to create new compromise indexes.

> 2. What philosophy is used when creating a
> generic strategy for all decks and
> conditions, as with K-O (other than dropping
> some of the indices for multiple decks, of
> course)? Is it best to use the 4-deck or
> 2-deck indices as most representation of all
> deck variations? Or should the highest
> indices be used to avoid expensive errors?
> What about differences between S17 and H17,
> or DAS?

As for decks, you need to decide this yourself depending on the time you expect to spend at each game. If you see a large difference in an index, look at the index report to see if the EV difference is large or small between the two indexes. The odds are that in a situation where the index is substantially different - the index difference in EV is small anyhow. As for rules, H17/S17 affects plays when the dealer has an Ace or Six; DAS affects Split indexes except for Tens and Aces. But you probably will have none that are affected anyhow.

> 3. What's the best way to apply the
> "reduced and rounded" philosophy
> to the resulting indices? Is it best to just
> round them up to the next highest group
> index?

The best way is trial and error. Try different values and run sims. You can start by using the KO categories. And you can use the index reports to get an idea of the EV lost by compromising a specific index.

> 4. In the case of index numbers such as 6
> and 7 (which I got in a preliminary sim), I
> can see only two ways to maintain the
> reduced/rounded groups. The first would be
> to "round" these numbers down to
> the pivot point of four; the second would be
> to create a new, fourth group (probably with
> an index of 6 or 8) to contain the new
> indices. Is this a valid assumption?

Depends on the balance of effort/accuracy desired. Again, trial and error is the only way to know for certain the effect on overall EV.

You might also want to read:

How to Create a Card Counting Strategy

[BJ Majish]

Thanks for the quick reply, Norm. Very helpful, as always.

I had aready read your Web page on the subject, but will read it again.

I am aware of the effects of H17 and DAS. My question had to do with the choice of indices that cover both S17 and H17, and nDAS and DAS for a generic strategy such as K-O. I guess the answer again is to run simulations for both cases and compare EVs.

Thanks again.