Sorry for the delay. Internet was down.

Yes, that's what I'm doing.

Maybe I'm misreading something. So let me do a step by step rewrite to see if I'm understanding correctly.

'p' = probability, which is the Jones Probability that we already know I'm doing right.

A hit is scored as -LOG(p), and a miss as LOG(1-p). LOG being the natural base e logarithm. Is that negative sign on the hit formula correct?

The expected value for each prediction is: (spacing added for clarity)
ev = -p * LOG(p) + (1-p) * LOG(1-p)

And the variance for each prediction is: (brackets and braces for nesting clarity)
var = p * (1-p) * {LOG[(p)*(1-p)]}^2

Since both the above formula only depend on probability 'p', then in the case of the 3 day windows, where the probability is always the same for all predictions, then the expected value and variance for all predictions is also the same, right?

Now, for a set of predictions, the score, expected value and variances are computed for each prediction. The set of score, ev, and var for all predictions are summed, such that we get sumscore, sumev, and sumvar.

Then, the standard deviation is sqrt(sumvar).

The final score is (sumscore-sumexpected)/standard deviation.

How'd I do?

Brian