Quote:
Originally Posted by jeebus1083
I know that many of the systems/analysts in HSH are based on composite (?) factors and weighted ranks. The original NewPace employee concepts such as population samples, weighted/composite numbers. Dave fine tuned the approach over time, but the point is that an analyst does not use a single paceline to represent “the” horse. It attempts to take into account the horse’s history (or comparable parts of it) and weighs it against the other runners and expected profit/loss ($net).
|
That is exactly right. Thanks, Jeebus.
The idea is to emulate paceline selection without actually selecting pacelines. It works amazingly well.
Consider the old Sartin concept of EP-SP-W being the only factors.
You wind up with 10 pacelines, each with 3 possibilities=30 choices. Then you interpret the model and ultimately use one column. (or maybe more)
Thus, if you chose the 2nd paceline back and decided that EP was the column to use...
All we've done is built a modeling process that begins with the 3 columns, based upon things like Last Race, Best of Last 2 Races, 2nd Best ever, etc. In this example, there are actually 39 possible choices.
Thus, if our model said that the most important factor was EP, Best-2-of-Last-3, we'd select this:
We actually model 8 factors x 13 measurements, so there are 104 possible choices.
All of this, without selecting an actual paceline.
BTW, in the new software I am building which I hope will be in beta before the Derby, there are around 10,000 combinations.
Dave