Scaling Ratings

[mwilding1981]

Do you think it is better to have all your ratings put on the same scale or not? I currently don't but it would seem to me to make a lot of things easier if I did. What is everyone's opinion on this?

[Sinner369]

Perhaps, you should explain what "scaling ratings" means? Same weighting on variables (factors)?

[GameTheory]

I think he means that if you have a rating for class and a rating for speed and a rating for jockey ability, etc, then should you put them all on the same scale, e.g. 0 - 100 so that 50 at one rating "means" the same thing (in terms of that rating) as 50 at another.

Well, that depends what you are going to do with them. The problem is that each rating has a different inherent value in terms of predicting races. It is not like you can look at a horse with speed = 25 but jockey ability = 75 (avg 50) and determine he's equal to a horse with speed = 75 and jockey ability = 25 (also avg 50). Let's fact it, if the bad jockey can just hold on, the faster horse is going to win most of the time. Now if you could build into your ratings a weight adjustment for the inherent importance of the factor so that 50 of any factor actually equaled the same thing in terms of winning chances, and you really could just take the average for a super power rating, you'd really have something. But doing that would mean NOT having them on the same scale -- maybe you'd have speed on a 0-200 scale and jockeys on 0-50 or some such thing. (Don't mean to pick on jockeys here -- plug in any example you like.)

Of course, if you're comparing similar ratings -- early speed vs late speed -- well, your numbers probably are already on the same scale. But even that is tricky because they can be on the same scale, but with different meanings depending on the scale. For instance, early speeds are faster than late speeds in general, right? I mean coming from the same horse and in terms of raw times, that is. Well, time is a standard scale -- seconds, tenths of seconds, etc. But if you put early & late on THAT same scale and convert to a rating (just a different way of looking at the raw time) so that faster = higher, well now your early ratings are going to tend to be much higher than your late ratings. Jim Cramer makes his pace figures like that -- he extrapolates the velocity as if the horse was going to run the whole race at that speed and gives it a rating on the same scale as his other figures. ( See this: http://www.rspos.com/pace.html ) This makes it easy to see energy patterns of a horse. But maybe you don't want that -- maybe you want 50 on a scale of 0 - 100 to mean "median for the rating" -- now the rating is measuring a level of performance rather than a velocity. But either way, they are "on the same scale" in a sense, and yet are completely different. So...it's a quagmire, like most things in handicapping.

But if you do want a "quick and easy" way to compare different numbers from different scales, that's what z-scores are for:

http://en.wikipedia.org/wiki/Standard_score

[ddog]

...and since an avg jockey(how do we know avg?) on a fast horse is one thing, but lots of bad jocks NEVER get a fast horse, so a jockey rating is ............

[mwilding1981]

That is exactly what I meant GT. Once again thank you for the excellent replies and links. For some reason I had completely forgotten about z-scores although I used to use them all the time.