Multiparameter sims in CVData

[Photonman]

I've been trying to do a multiparameter sim in cvdata but can't figure out how. Norm, bless your heart, I know you're swamped so I thought I'd try here.

In the playing page you are offered the option of adding a multiparameter side count. When you click on it all you get is the "sample multi-parameter strategy" as an option. I tried to create one in create new system, and could name the values I wanted to side count, but had no way to assign the point values. Neither the manual or the "Help" feature addresses the side count feature other than the ace side count.

What I'm specifically trying to do is side count 5s vs 6s (5=-1,6=+1) for the 16 v 10 decision which comes up a lot. If the rc sum is <1 you hit ow stand. (See Griffin) I want to combine this with my usual count which is the Uston Advanced Plus Minus (3-7=+1, 10/A=-1) and test the results, both just using the 5/6 count for the 16-10 decision and UAPM for every thing else or perhaps combining them some how. (This came up on Green Chip.)

As an aside I also have tried a side count of A=+1 vs 3+-1 that I add to the running count for playing decisions in 2d. This converts the UAPM to Griffin 1 which is the most powerful level one count from a PE standpoint. I did this by using the UAPM values as a "betting side count" in betting strategies and the Griffin (4-7=+1, 10=-1) values for the primary count in playing strategies. I found that contrary to the manual, CVD doesn't add the two counts ie the primary and "betting side count" but uses the "betting side count" by itself as a separate count for betting purposes, so it's not actually a side count.

I personally have found it easier to keep the A vs 3 using my finger tips than using a level 2 count. The advantage being that unlike ace correcting for each betting decision, you're ace correcting for the occasional playing decision.

So if any of you CVD users have figured out how to use the multiparameter feature please let me know.

[Norm]

CVData allows a balanced side count for betting, but not playing. The multi-parameter side count for playing counts excess cards of your choice to apply against multi-parameter index tables.

[Photonman]

"The multi-parameter side count for playing counts excess cards of your choice to apply against multi-parameter index tables."

So how do you designate the cards and their tag values and determine their "excess"?

Also "CVData allows a balanced side count for betting, but not playing." As I mentioned, from what I've found it seems that the balanced side count stands alone, ie that it is NOT added to the playing count. Is this correct?

As a way around all this I tried to determine by simulation what the contribution of 5 vs 6 counting would be by creating a playing strategy with 5=-1 and 6=1 in CVIndex, (with a new system name) and tried to determine the index. IE the system played BS except for the 16 vs X decision. I got the result "always hit" which is flagrantly wrong, so I must have done something wrong. Per Griffin any excess of 5s over 6s one should hit and vice versa with an efficiency >.6. When I forced a standard sim with the system using the above rule I got worse performance than with BS, which again per Griffin shouldn't happen.

[Norm]

You designate the cards in a multi-parameter strategy. Go to Define/Edit Strategies and create a strategy based on the Sample MP strategy in the Multi-Parameter Strategies folder. There are no tag values. I'm not sure what you are thinking about is an MP strategy. See http://www.qfit.com/blackjack-side-counts.htm. MP side counts are described near the end.

[Photonman]

Upon my death send my corpse to the ammo factory.

Forget about the multiparameter issue for now.

I did some controlled tests in CVD and determined that indeed the betting side count is added to the main count and not used instead of it. I don't know what I did to get the weird results I got in the first place but all it takes is checking the wrong box etc and everything's affected.

Getting back to the issue of testing 5 v 6 for the 16 v T H/S decision, I still get nonsense.

The effects of removal for 5 and 6 for 16 v T are large and of opposite sign. Per Griffin any excess of 5's over 6's is 60% efficient in the 16 v T decision. Accordingly I generated a new CVIndex config, a new strategy based on composite and even complete BS, tagged 6=-1 and 5=+1, and ran the index generator. The results were "Always Stand," which isn't even basic strategy, whereas per Griffin a running count of <=+1 one should hit, and where I would expect the index to be.

What have I done wrong? I know it probably seems trivial but a conflicting result such as this impeaches my confidence in other results.

Rest massless,

Photonman

[Norm]

Sorry, I don't understand what you're trying to do. A strategy where six is -1 and five is +1 will generate gibberish.

EDIT: OK, I guess I see what you're doing; although I've never heard of anyone using CVIndex for this purpose. If you want to test for an excess of fives, five must be -1, not +1. You have five set to +1.

[Photonman]

Being composed of U238 I actually tried it both ways. Gibberish both ways. I don't see why you'd tag 5 as -1, it acts as a "small card" for 16 vT whereas 6 acts as a "big card." By convention we tend to hit with low/negative counts, ie when the remaining pack is rich in small cards/deficient in big cards an excess of 5's will be revealed by a negative count, ie when more 6's have been played.

In principle I don't see why CVD shouldn't be able to consider any single parameter counting system even if its utility is strictly for one decision. We aren't using it for betting.

[Norm]

Being composed of U238 I actually tried it both ways. Gibberish both ways. I don't see why you'd tag 5 as -1, it acts as a "small card" for 16 vT whereas 6 acts as a "big card." By convention we tend to hit with low/negative counts, ie when the remaining pack is rich in small cards/deficient in big cards an excess of 5's will be revealed by a negative count, ie when more 6's have been played.

In principle I don't see why CVD shouldn't be able to consider any single parameter counting system even if its utility is strictly for one decision. We aren't using it for betting.

It has nothing to do with small vs. big. You are looking for an excess of fives. Therefore, you must count it as a negative. In this situation five is a good card. Normally, ten is a good card. You can't good cards as negatives.

In any case, CVIndex was not designed for this purpose as no one has ever requested this in the past.

[bejammin075]

There's a short list of what casinos generally look for in card counting, and varying your play on 16 vs 10 is on that short list.

[Photonman]

Now wait a goldarn minute. If I take a single deck, deal 2 fives off the top and count that as -2 I would by convention be inclined to hit a 16 v T, no? However there is a deficiency of fives remaining. Ergo one should count the fives as positive. A ten is a bad card for this decision, it has a positive effect of removal as does the 6, so for this decision at least their signs should be the same. Certainly you wouldn't advocate hitting 16 v T more often in a ten rich deck, ie where few of them have been played. A ten is a good card because of its betting correlation and a bad card from a playing effect, that's why we bet more in positive counts and hit more in negative counts. In this case we're not talking about betting, only playing, and the 6 is the bad card.

I realize this is of no direct practical value, it started from a question in Green Chip. I was trying to simulate it. I don't understand why CVData shouldn't be able to handle it, unless there was some line of code that prohibited 5 and 6 from having opposite signs, which I doubt. All tag values for a single parameter count are a compromise in the betting efficiency vs playing efficiencies of all the various departures and some systems perform better on some decisions than others, and CVData seems to handle all those other EOR conflicts OK.

Thanks for you patience.

[Norm]

Now wait a goldarn minute. If I take a single deck, deal 2 fives off the top and count that as -2 I would by convention be inclined to hit a 16 v T, no?

There IS no convention with such a count. It's an unconventional count and would act differently depending on the hand. In any case, CVIndex was not designed to handle such a count.