Database

[shoelessjoe]

Hi All,A buddy of mine is putting together a database at Penn National.He is seperating the horses by class and trying to get an average time of what they run.I told him that it might be inaccurate because he is not taking into consideration the track variant.I started thinking maybe I gave him some bad advice.I would appreciate any help that you may have to offer.Shoeless

[cj]

There are many different ways to do it, but I tend to agree, without using some type of variant, you will get very skewed results. This is especially true at a track like Penn that has very extreme weather situations.

[headhawg]

shoeless,

Is he trying to make par times? I'm no statistics wizard but I would think that if the sample is large enough you would get a stable average time for each class. Variants should only matter if you're making speed/pace figs.

If he is trying to make pars, Dave Schwartz has recently posted a link to pages referring to par times creation. Do a search and it should pop up.

HH

[michiken]

Has anyone ever made par times and variants based on the season the of the year?

For example, during the winter when the track is frozen, snowy you could create pars for say November - Feb?

[Tom]

Here's wha tI used to do, back in the day before commercial par times, Beyer numbes, internet sitses.....on my trusty old Tandy 1000 ( and the Radio Shack Color Computer before that!)


1. Calculate the average time for each distance, raw times.
Throw out the fastest and slowest 25%of the data, so you only use the middle half.

2. Convert each race time to a speed figure bases on this data. I used 80 as my base average for each distance

3. Adjust each race using the template from Quirrin's book - ie, +2/5 for fillu and mare races at a sprint, +3/5 at the route. add +5 for a maiden claiming race at a sprint, +7 for a route.......

4. Go through and calculate a DRF type variant for each day and use it to modify the times. I would throw out any race variant that was large enough or small enough to change the daily variant by more than 2 points.

5. Now sort by class and use these modified times to calculate class pars.

With Excel, it is a pretty easy task

[midnight]

One idea would be to use the median figure, not the mean. That would help negate the effect of the skewed times. In other words, use the figure that's in the center of the data sample (the median), not the average (the mean) of the sample.

[Tom]

Good point, midnight. I use Excel to calculate both - if they are equal, I have some confidence, but if they are off by much, I look in that direction for reasons.

[shoelessjoe]

Actually I dont really know if he is making par times or not.It just seemed to me when that tracks starts to freeze as somebody has already pointed out that the variant would come into play.Shoeless