Ask Andy Beyer

[steveb]

Originally Posted by cj

I thought it was pretty well know that Beyer uses projections now, as do most people that make figures. Pars are so tough to do these days, they are actually nothing but projections themselves in my opinion.

you project horses, i project races(using standards).


if a horse has run an 80 in a race where winner generally runs a 70, then next time it's in that 70 race, imo all it needs to do is run that 70.

i would not project it as an 80

[cj]

Originally Posted by steveb

you project horses, i project races(using standards).


if a horse has run an 80 in a race where winner generally runs a 70, then next time it's in that 70 race, imo all it needs to do is run that 70.

i would not project it as an 80

For me at least, projecting is a lot more complicated than that.

The standard part is tough here, particularly when using time. There simply aren't enough races to come up with a reliable standard with all the various conditions, distances, surfaces, etc.

[steveb]

Originally Posted by cj

For me at least, projecting is a lot more complicated than that.

The standard part is tough here, particularly when using time. There simply aren't enough races to come up with a reliable standard with all the various conditions, distances, surfaces, etc.

it is for me too, but that is the gist of it.



i don't know how they class horses in your country, but most places give them official ratings where 1 point = 1lb or .5kg.
the ratings themselves are iffy, but they allow you to figure how fast these types will run generally(solved by regression .99 correlation times run to official ratings using every horse in race rating).

after any meeting the std dev between expected and actual for any meeting rarely exceeds 5.... 5 would mean 5 metres per 1000 difference actual to expectation which is bugger all(.005%) in the scheme of things.

[cj]

Originally Posted by steveb

it is for me too, but that is the gist of it.



i don't know how they class horses in your country, but most places give them official ratings where 1 point = 1lb or .5kg.
the ratings themselves are iffy, but they allow you to figure how fast these types will run generally(solved by regression .99 correlation times run to official ratings using every horse in race rating).

after any meeting the std dev between expected and actual for any meeting rarely exceeds 5.... 5 would mean 5 metres per 1000 difference actual to expectation which is bugger all(.005%) in the scheme of things.

That is the thing, we don't have a rating system like that here.

[steveb]

Originally Posted by cj

That is the thing, we don't have a rating system like that here.

fair enough.
you could probably do those same regressions using your own timeform figures.



would probably be more accurate too than using official ratings.


my own opinion(and confirmed by checks and balances i do), is that results are far more predictable using a race expectation, rather than horse expectation.


of course it all comes down to your own methodology i guess.

[classhandicapper]

I'm sure the goal is to generally keep dirt tracks consistent. However, one of the complications of doing figures in the US is that dirt tracks tend to be maintained between races. The maintenance crew will add water, allow moisture to dry from sun/wind, float or harrow the track, etc... between races.

So if the speed of a track can vary from race to race as it does (over and above the impact of wind, humidity and temperature on the horses) you sort of have to analyze each race independently and within the context of the day all at once to try to understand what happened.

Also, as CJ was saying, we have an almost endless supply of class designations at each track that don't translate well from track to track even when they are called the same thing. Some classes have a huge standard deviation even at the same track. So just using a PAR/Standard leads to problems in the US. That's not to say they can't help with the analysis at times.