Bert Nommel: Some Interesting Sims

[Bert Nommel]

I just bought the BJRM. Something funny I just noticed. For the RPC that comes with BJRM(6D, S17, 5 of 6 dealt) the frequency for RPC at TC 0 is 44.28%. When I simmed this with SBA I got frequency at TC 0 of 14.61% for RPC. Also the RPC advantage at TC +1 was 0.487% in the included data, but my sim showed TC +1 advantage of only 0.025%. Now, either my SBA is malfunctioning of the data in the BJRM is wrong or I'm missing something fundamental in my understanding of the BJRM. I floored my TC, but that should not make much difference. Sim it yourself and see.

Sincerely,

Bert

[Parker]

> Now, what is true-counted TKO?

Redundant?

The devil made me do it! :-)

[Zenfighter]

> I just bought the BJRM. Something funny I
> just noticed. For the RPC that comes with
> BJRM(6D, S17, 5 of 6 dealt) the frequency
> for RPC at TC 0 is 44.28%. When I simmed
> this with SBA I got frequency at TC 0 of
> 14.61% for RPC. Also the RPC advantage at TC
> +1 was 0.487% in the included data, but my
> sim showed TC +1 advantage of only 0.025%.
> Now, either my SBA is malfunctioning of the
> data in the BJRM is wrong or I'm missing
> something fundamental in my understanding of
> the BJRM. I floored my TC, but that should
> not make much difference. Sim it yourself
> and see.

> Sincerely,

> Bert

".....my sim showed TC +1 advantage of only 0.025%"

Maybe here lies the problem, a month ago I've got
for RPC and 5000 millions rounds this:

TC = 1 ev = 0.354% SE = 0.0047%

Lets take 3 sigma to be 99.7% sure

0.354% +/- 0.0141% = [ 0.213% and 0.368%]

Since these TCs are also floored down, your figure
looks not real, at least for me, obviously.

Sincerely
Z

[Don Schlesinger]

> I just bought the BJRM. Something funny I
> just noticed. For the RPC that comes with
> BJRM(6D, S17, 5 of 6 dealt) the frequency
> for RPC at TC 0 is 44.28%. When I simmed
> this with SBA I got frequency at TC 0 of
> 14.61% for RPC. Also the RPC advantage at TC
> +1 was 0.487% in the included data, but my
> sim showed TC +1 advantage of only 0.025%.
> Now, either my SBA is malfunctioning of the
> data in the BJRM is wrong or I'm missing
> something fundamental in my understanding of
> the BJRM. I floored my TC, but that should
> not make much difference. Sim it yourself
> and see.

I've tried to impress upon you and anyone reading how extremely difficult it is to perform a truly pure apples-to-apples comparison, expecially when using more than one piece of software. Inadvertent errors creep in.

The true counts in BJRM are truncated. A count of zero includes all true counts greater than -1 to less than +1. That's where the 44% comes from. If SBA says that a TC of zero occurs only 14.61% of the time, that sounds like the frequency of a true o****of exactly zero, with another 15% or so on either side until we reach -1 or +1.

When you floor the TC, you eliminate the entire left side of the TC = 0 symmetry from truncating. But, if you also don't count the TC of exactly zero, you might get the 14% you allude to. if you include it, you'll have to get more than 14%.

For the edge, using indices, a TC of +1 means the edge in the interval between +1 and +2. It is certainly not just 0.025%.

Perhaps if Karel or John reads this either of them can help you further.

Don

[Bert Nommel]

If you divide by 1/2 decks instead of whole decks, and if you truncate, the frequencies and advantages almost match BJRM.

Appreciate your help.

Bert

[Cacarulo]

> Here are the results.
> 6D,S17,DOA,DAS,SPA1,SPL3,NS,5/6 (260 cards
> dealt),Catch22 indices,5000 million rounds
> (1000 million rounds is not enough) and
> heads up. True Count estimated to nearest
> 1/2 deck.

> UBZ(true counted): SCORE = $38.44/100 hands
> TKO: SCORE = $36.48/100 hands
> RPC: SCORE = $35.4/100 hands

> Cacarulo, would you perform this sim too and
> import to BJRM to validate my results? I
> don't think others will believe it.

Bert, I have run only the TKO sims. The only difference is that I estimated the TC exactly and used my own simulator. Apart from this I have also run Cac/7 and Hi-Lo since there were some doubts about Cac/7 outperforming TKO.

Here are the results (1-12 spread):

Cac/7 (using Hi-Lo indices except for INS) = $37.13 (in fact this could be a little more)
Cac/7 (using exclusively Hi-Lo indices) = $36.86
TKO = $36.38
Hi-Lo = $34.84

Notice that Cac/7 using exclusively Hi-Lo indices has a PE = 0.509 and IC = 0.789 whereas TKO has a PE = 0.522 and IC = 0.813. Where does the difference come from then? The answer is in that "tiny" difference in BC. Cac/7's BC = 0.984 versus TKO's BC = 0.977.

In terms of % Cac/7 is 2.06% better in SCORE than TKO and 6.57% better than Hi-Lo.

I will post the complete analysis on DD in which other spreads were analyzed.

Sincerely,
Cacarulo

[just a counter]

Using BJRM I noticed when comparing KO against the RPC and Halves (the exact same rules),KO out-performs. Is this actually TKO, or some mistake?

Since BJRM trunicates from +1-1.99 TC for betting purposes and shows an average ev of .37, what is the actual ev at an exact TC of +1?

I'm now a little apprehensive about placing the optimal bet BJRM calls for at a +1 TC(s17). I'm seriously thinking of waiting untill the TC reaches 1.5 before making the optimal bet,what's your opinion?

JAC

[just a counter]

I believe I read a previous post of yours saying you use the 7 side count for only playing purposes. Since a depleted 7 is benficial to BC,why not also use it for betting purposes?

JAC

[Don Schlesinger]

> I will post the complete analysis on DD in
> which other spreads were analyzed.

Please do the RPC, so that I can see how it compares to TKO, which, I am presuming, it will out-SCORE.

Don

[Cacarulo]

> I believe I read a previous post of yours
> saying you use the 7 side count for only
> playing purposes. Since a depleted 7 is
> benficial to BC,why not also use it for
> betting purposes?

No, you have it backwards. I use the seven for betting purposes. Note that I'm using Hi-Lo for playing where the seven is not counted.

Sincerely,
Cacarulo

[Cacarulo]

> Please do the RPC, so that I can see how it
> compares to TKO, which, I am presuming, it
> will out-SCORE.

I agree with you. RPC should be better and I think I know where the problem is. I think Bert is using the count "-2 1 2 2 2 2 1 0 0 -2" but with the indices of "-1 0.5 1 1 1 1 0.5 0 0 -1".
He needs to double those indices before running the sim. Bert: Could this be the problem?

Sincerely,
Cacarulo

[Cacarulo]

BJRM2000 using the I18 gives $37.33 for RPC and $37.21 for UBZ2 (in RC mode). With Catch22 should be a little bit more.

Sincerely,
Cacarulo

[Cacarulo]

> I agree with you. RPC should be better and I
> think I know where the problem is. I think
> Bert is using the count "-2 1 2 2 2 2 1
> 0 0 -2" but with the indices of
> "-1 0.5 1 1 1 1 0.5 0 0 -1".
> He needs to double those indices before
> running the sim. Bert: Could this be the
> problem?

Another alternative would be to use "-2 1 2 2 2 2 1 0 0 -2" with the indices of "-1 0.5 1 1 1 1 0.5 0 0 -1" but with the divisor (SBA feature) set to 2 (not 1).

Sincerely,
Cacarulo