ikkyu2 (99.52)

Accuracy banking

August 05, 2010 – Comments (5)

I quit paying daily, close attention to CAPS in Jan of 2008, shifting from a sort of day-trading to a longer-term strategy, checking in every couple of weeks.

Somewhere in there, the way accuracy is calculated changed - and TMF was a little vague about the new formula.  I'm going to give a brief summary of what I derived from reading the TMF description, and then my gentle readers can chime in at the comments to tell me what I got wrong.

1)  All players with more than 100 picks still compete on a level playing field with each other, percentile ranked by their accuracy score.

2)  The score cutoffs for ended picks are the same: <0, inaccurate; 0-4.99, not counted; 5.0+, accurate.

3)  The score cutoffs for active picks have changed:  <0, inaccurate; 0, not counted; >0, accurate.

4)  Players with fewer than 100 picks are treated differently from players with 100 or more picks.  TMF is vague on exactly how, but they do mention that the calculation is processor-intensive.  I assume that they are handicapped by some permutation of the binomial theorem - maybe their accuracy rank is multiplied by (1-p), p being the probability (one-tailed) that their level of accuracy would have been produced by a coin flip.  Or maybe they are subject to some kind of Bonferroni correction - heh.  I doubt it.

Anyway, in order to get any kind of serious score points and be in any kind of contention for a top Fool hat, one probably has made more than 100 picks - and I have - so for my purposes, I can disregard #4 entirely.

Also, my current accuracy of 77% puts me in the 100th percentile as far as accuracy - and that means that I no longer need to accuracy bank, in fact, I shouldn't try.  I should focus on getting more points if I want to up my CAPS ranking.

What that means to me is a strategy shift.  Previously, it was always end accurate underperform picks and re-up them (covered in a prior blog entry, contains assumptions about the long-term performance of S+P, yadda yadda).  That way, I could 'bank' accuracy points and gain the maximum amount of score.  Outperform picks were to be let run forever.  

That is no longer my optimal strategy, however.  Consider my RIG outperform pick, made at the top of the market at a price of $142.

RIG's current price is 57.  Because of S+P movements I'm down 43 points on the pick.  Let's assume for the moment that the S+P won't move in the future; and think just about the performance of RIG.  Previously, I did not want to lose an accuracy point, so I would let that pick run.  If RIG ran back up to where I picked it, I gain about 43 score points and my negative accuracy point goes away and turns to positive, net gain accuracy +1.

However, I'm already at 100%ile accuracy.  I don't need more accuracy points. I need score.  I need to end that pick - leaving my accuracy score unchanged.  If I re-up and RIG goes to 142, I get more than 100 score points off the same run, and I get an accuracy point - net gain accuracy 0.  (Of course, if I re-up and RIG keeps falling, I get -2 accuracy points.)

I guess my conclusion is that optimal strategy hasn't changed from the CAPS changes in terms of calculating accuracy; but the fact that I am topped out in accuracy actually has changed my own personal strategy: if I want to get a better overall CAPS ranking, I need to quit fooling with accuracy banking and go for total score.

Comments welcome.