My count sys uses 2-6 as plus and 9,10 as minus; with Ace side count. I've run the numbers with SBA and my indice for 2,2v 8 is plus 4 and 3,3 v 8 is plus 10. Reason 2,2 works and 3,3 works in my count is on the splits you can draw 2,9,10 or 2,9,9 or 3,9,9; or other combos because heavy proportion of 9's are in the mix.
> why?
No idea. I've never heard of the program.
An index of +10 is way off for 3,3v8 for Revere +/-. But, it?s not surprising that an index generator would pick it. First it must be understood that indexes are not linear. Nothing in BJ is linear except Insurance. But indexes, and in particular defensive indexes, are not only not linear; they can bounce back and forth between decisions as the count increases. This can confuse an index generator. 3,3v8 is an interesting example. Below is the advantage for Split vs. No Split for 3,3v8 by TC.
At low TCs, you should not Split. When the count gets high enough Split. But, at a very high count, don?t Split. This is not surprising when you think about defensive splits like this. Do you really want two hands of three at a very high count?
OTOH, when the lines are this close, it doesn?t matter what you do. Might make a good cover play to always Split.
> I had briefly consider running sims to see the
> difference in the indexes. But it would take tens of
> billions of hands, the difference is so tiny.
I wrote to Wong, and his answer is, ostensibly, confirms exactly what I wrote:
"For doubling allowed after splitting, 3-3 vs 8 is basically a tossup, and it's not sensitive to the count. There is almost nothing to be gained by deviating from basic strategy. So any index numbers are basically worthless. I suppose I could have suppressed them for one counting system or faked some in for the other to make them look consistent, but instead I chose to go with the results generated by the software."
Don
I ran ten billion rounds each for 33v8: Always Split, Never Split and an index of +4. Following are some conclusions:
[*]Use of the index had a gain in SCORE of 0.01. That is, one penny per hour with a $10,000 bankroll at full Kelly.[*]ONE card difference in penetration has 40 times the effect on SCORE as using an index.[*]Always splitting had a loss of 7 cents per hour as opposed to Basic Strategy. I believe this makes it a candidate for a cover play since the play looks stupid, is an obvious violation of BS and costs a tiny amount.[*]The difference between indexes of +4 and +5 is immeasurably small.[*]Using an index accurate to one-tenth?s is smaller than immeasurably small - assuming such an index even exists.[/list]
I agree with Wong on all three conclusions: that the hand is insensitive to the count (as indicated by my chart in the ?way off? post,) that the index is useless and that if you really wanted to use the index, it is +4.
Disclaimer: The play is so close, 30,000,000,000 hands is not enough for absolute accuracy.
CVCX Online
Who will challenge these data? Simulation has borrowed 5 seconds,Russian program "Korovin" -
Rules 6 decks,ENHC,3 split,DAS
2,2 vs 8 3,3 vs 8
TC Hit Split Hit Split
0 -0.1587 -0.1737 -0.2166 -0.2294
1 -0.1661 -0.1770 -0.2256 -0.2358
2 -0.1744 -0.1806 -0.2356 -0.2430
3 -0.1828 -0.1845 -0.2457 -0.2504
4 -0.1915 -0.1882 -0.2560 -0.2582
5 -0.2004 -0.1924 -0.2666 -0.2665
6 -0.2098 -0.1969 -0.2774 -0.2754
I'm sure Korovin's program is not a simulation but a calculation,so called combinatorial analysis or shortly CA. It is based on the most likely deck distribution for every TC.
There is total agreement between Korovin's data and mine up the third position and the slight differences from the forth position onwards are easily explained by the fact that my program assumes an infinite deck.
Francis Salmon
> Who will challenge these data? Simulation has borrowed
> 5 seconds,Russian program "Korovin" -
Sorry, I don't understand what you are trying to say. I've never heard of this program; but you cannot come up with accurate numbers in five seconds.
> Sorry, I don't understand what you are trying to say.
> I've never heard of this program; but you cannot come
> up with accurate numbers in five seconds.
I wish to tell, that Wong's data not correct in occasion of these Split.
But why do you insist that this program is correct? Representative subsets are not an accurate method of index generation. The assumption of infinite decks also makes for inaccurate index generation. What this program does is make a calculation based on inaccurate assumptions.
> I wish to tell, that Wong's data not correct in
> occasion of these Split.
Precision: 6 sigma
Maximum number of pairs of hands at index: 1000 million
SPLITTING (DAS) 2,2 versus 8 up
SPLITTING NO SPLITTING
5: Adv.:-19.569, SEr 0.045 Adv.:-19.753, SEr 0.025 Dif.: 0.184, SEr 0.047
Determined index(es): 5
Wong is right here, as we see.
Precision: 6 sigma
Maximum number of pairs of hands at index: 1000 million
SPLITTING (DAS) 3,3 versus 8 up
SPLITTING NO SPLITTING
4: Adv.:-25.792, SEr 0.031 Adv.:-25.753, SEr 0.016 Dif.:-0.039, SEr 0.034
5: Adv.:-26.777, SEr 0.045 Adv.:-26.793, SEr 0.024 Dif.: 0.016, SEr 0.050
6: Adv.:-27.565, SEr 0.052 Adv.:-27.778, SEr 0.028 Dif.: 0.213, SEr 0.058
7: Adv.:-28.465, SEr 0.080 Adv.:-28.840, SEr 0.043 Dif.: 0.374, SEr 0.089
8: Adv.:-29.362, SEr 0.091 Adv.:-29.860, SEr 0.049 Dif.: 0.498, SEr 0.101
9: Adv.:-30.620, SEr 0.146 Adv.:-30.937, SEr 0.079 Dif.: 0.317, SEr 0.163
10: Adv.:-31.416, SEr 0.162 Adv.:-31.945, SEr 0.087 Dif.: 0.529, SEr 0.181
11: Adv.:-32.560, SEr 0.279 Adv.:-33.362, SEr 0.150 Dif.: 0.803, SEr 0.312
12: Adv.:-33.939, SEr 0.303 Adv.:-34.201, SEr 0.162 Dif.: 0.262, SEr 0.339
Determined index(es): 5
No significance results. In other words here the ?differences? between Tc 4 and Tc 5 are not trustable
due to the standard errors of the sample even after a run of 1000 million of pairs of hands and a protective 6 sigma = 6 standard deviations of precision. Thus Wong is ?right? again. In other words feel free to use 4 or 5. You won?t gain the price of a hot dog at the end of a year for being precise here.
Sincerely,
Zf
> Thus Wong is ?right? again.
My selective memory tells me Wong is seldom found to be wrong even unto areas as insignifigant as this.
Is my memory correct and if so, did he have at his disposal the computing power available these days?
Thanks.
Posting Permissions
-*-
Bookmarks