Systems Archive 4

[ Read Responses | Return to the Index ]
[ Previous | Next ]

How come no one seems to trust the

Posted by Clarke Cant on 27 March 2002, at 6:35 p.m.

Corelaton Coeficient in general? The CCC on yahoo has been spamed a bit over ridiculous claims about which cards are worth more. Even Kim Lee has gotten into the act by posting on that a half the tens ace value is 75% as effective as side tracking the ace. It floats between the boards and never seems to end.

The measure of the effectiveness of a count, for betting purposes is essentialy its corelation coefficient to precise effects of removal for a given game. It is never the effectiveness of just the ace or just the ten anytime. It is the overall corelation coefficient. A good estimate of the playing efficiency can be found from the corelations between a reasonable count and a count at level 180, as decribed in the Blackjack Formula and the update in Blackjack Therapy (the free version online here).

Why don't people trust it? This isolation of card values is absurd, unless you want to all go back to ten counts.

None of the people involved could ever explain, for example, how come the Victor APC, (5s +3, the other low cards +2, 8s 0, 9s -1, Xs -3) why this count has better betting than Hi Opt I. And forget trying to explain its high playing efficiency. This count is almost as good at PE as the Uston APC, and is a bit better at BE. Yet it is just a bentup version of the Expert Count or Hi Opt I, if you look at it the way these people in these "which card?" threads have been examining things. It has higher corelation coefficients and that is that, and those accurately predict its performance in sims. Period. comeon people PERIOD PERIOD PERIOD.....

What is it that makes people not believe such correlation coefficient measurements?

Even Peter Griffin got into the act a bit in this, in edtions 4 onwards of The Theory of Blackjack, where he uses the average effect of removal for a card, to describe algebraic multiparameter indexes. That average effect of removal method was abandoned, for his final method, based on the correlation coefficient, by Arnold Snyder, when he was "scolded by Wong," who was the first person to figure out that his original algebraic method was actually a combination of how the correlation coefficient and the RMS ratio of the point count and a set of EORs, would transform the full deck favorability, from the EOR tables, into a useful point count index. Wong pointed out then that the corelation coefficient covered all such situations and not just for single level counts. Yet once Griffin slipped back into, for multiparameter systems, what is the same as the alternative method Snyder thought he needed for multi-level counts, before Wong corrected him, everybody seems to have backtracked. The Moss method for algebraic indexes is just Griffin's revival of Snyder's original error.

These are not ploppies doing this, but a wide range of math folks, who suposedly include the sharpest math minds to ever examine the game.

So what the hell is it about the CC that brings about such lapses?



The Systems Archive 4 is maintained with WebBBS 2.24.060398.