BGonline.org Forums

Isight insight: 1 pip = 2% (according to model).

Posted By: Taper_Mike
Date: Thursday, 18 December 2014, at 11:43 p.m.

In Response To: Isight insight: 1 pip = 2% (according to model). (Jeremy Bagai)

The deal-breaker for me on Isight's CPW formula is its large margin of error. On average, the error is slightly more than 4%. (I think that means plus or minus 4%!) In many cases, that is larger than doubling window itself!

I use the Isight CPW formulation, nevertheless, because I find it just as easy to calculate as the alternative Isight formula. I treat it as a doubling metric for money games, not as the CPW. Used in this way, it gives the exact same result as Isight's non-CPW formula.

Igor Erovenko explained to me that no linear approximation can ever be good enough to estimate CPW. Danny Kleinman's formula is a quadratric. That explains why it works so well. Both of the Isight formulas are linear.

When CPW at a given pip count (for the leader) is graphed as a function of race lead (in pips), the result is not a straight line. If it were, the relationship would be linear. Instead, the graph is curved. At any point, the slope of this graph is the value of a pip, i.e., %win per pip = delta(CPW) / delta(race lead).

The value of a pip, therefore, is the derivative (in calculus) of the CPW function.

Mike

Post Response

Subject:
Message: