The NEW Factor Ranking method

[Ted Craven]

This is just a summary of the new, changed factor ranking method to be introduced in RDSS with this Leverage/Pace module under construction. This factor ranking method applies to ALL existing and new factors which have a rank.

The old/existing method dealt with tied ranks as follows: say you have 2 ranks tied for 2nd. The top 5 ranks will look like: 1-2-2-3-4. The new method will skip a rank after the tied ranks: 1-2-2-4-5. Here, the tied ranks 2s occupy the 2nd and 3rd positions in the rank set, though they both are of equal value.

For model and study tracking purposes, now when you see such and such a factor has a hit rate in say, the Top 3, you know that there are only 2 horses better than it in the race it came from (e.g. the ranked 1 or 2 or the 2 tied rank 1s). Under the old method, there could be 3 or more horses better than that rank 3 and thus 4 or more horses included in those Top 3 stats.

This would lead to a certain 'inflation' of the factor hit rates. Now, although, per the above example, there could still be 3 horses included in the Top 2 (or 2 horses in the Top 1 with for example ranks 1-1-3-4-5), at least you know the number of horses better than a given rank. IMO, when faced with a given race and a collection of ranked factors, when referring to your factor models, it has the overall effect of narrowing down the number of horses under consideration in the race - and making the models more effective.

Note: the old method will still be available to those who wish to retain it. When this new release of RDSS goes public, along with the explanation for this factor ranking change - folks can choose to revert to the previous method if they choose (and understand yet disagree with the reasoning). Right now in this testing software - the ranking method is FORCED to the new one: you can't choose to revert it back to the old one.

Note: this difference in ranking methods will have the effect that existing factors which are made up of summing or weighting ranks of other factors (Primary Line Score, VDC, Rx) will not compare identically with output under earlier versions if you are comparing readouts with earlier versions having identical Config settings, please be aware of this difference. For example, your BLBL rank order, or Rx ranks may differ somewhat. Other main factors not based on ranks (Velocities, energy, TPR, Segments, etc) will remain unaffected. The proposition is that it will be better - although as usual, any reasoned or statistically justified opinions to the contrary are always welcomed.

Ted