Quote:
Originally Posted by bobphilo
Just realized that the primary problem with this study is that it is going about things all wrong.
When planning a study, first one decides what effects a variable has on a given population. Then compare what are the effects of different values of this variable.
To give a simple example: To determine the effect of being on the early lead in the population of 6F races, you compare the Win%, Impact Value, ROI , etc of horses who lead early in that population with these values of those that don't. That's how Brisnet does their Bias numbers in their Race Summaries
Therefore if you want to know the effect of breakthroughs on the starter population in general, you compute the Win%, IV, ROI, etc of breakthroughs with the general population.
If you want to know the effect of breakthroughs on favorites in order to expose vulnerable favorites, you take the population of favorites and compare the Win%, I.V., R.O.I. of those who breakthrough with those that don't.
The problem with this study is that it does neither. All it does is try to tell you out of the population of horses that breakthrough whether they do better as favorites or not. It only helps if it had already been determined that breakthroughs are a disadvantage in a previous proper study, whether or the favorite or other is the better bet. Not nearly as useful. Even at that this study seems to have problems with data correctness.
|
The performance of all favorites in general is very easy to find. It will clearly show that those that don't break through the gate (if noted in the chart) do MUCH better than those that do. Most people know how favorites do these days from a Win%, IV, and ROI perspective. I would say though it wasn't presented by o_crunk that is pretty much because it is a known quantity. Even those that don't really know can look at the break through horses and know they are much worse.
I also happen think you guys are greatly overstating the data problem here. It isn't perfect, but it isn't near as bad as you guys are pretending and certainly not bad enough to skew the statistics posted from being quite negative.