JCapper Message Board

JCapper 101
-- Analyzing shippers by track

Home	Register Log In

Analyzing shippers by track

tanix
11/20/2011
8:13:20 PM

I'm trying to figure out the best way to utilize the data window to analyze the performance of shippers at a particular track broken down by the tracks that they shipped in from. Let's say that I wanted to look at Gulfstream and determine how shippers from various tracks perform there in their first start.

I know there has got to be a relatively easy way to accomplish this, but right now it is eluding me. Any ideas?

jeff
11/20/2011
9:59:25 PM

Before diving in, I should probably mention the SQL Cheat Sheet.

I recently added a new section towards the bottom titled SQL EXRESSIONS (A LITTLE MORE ADVANCED).

As it so happens, that section deals with how to identify shippers, specify the shipped from track, and how to apply date ranges to make things meet specific.

Here's a link:
http://www.jcapper.com/HelpDocs/HelpFileSQL.txt

The following sql expression flags starters racing at GPX shipping in from another track:

SELECT * FROM STARTERHISTORY WHERE TRACK = 'GPX'
AND TRACKLAST <> 'GPX'
AND STARTSLIFETIME > 0

Note: A min number has been specified for lifetime starts because the value in the tracklast field for first time starters is a zero length string or "" (which gets flagged by TRACKLAST <> 'GPX')

Hint: You can make it easier to see what is going on if you display individual plays in the Data Window while adding an ORDER BY clause to the sql expression so that individual records returned by the query are sorted by the shipped from track.

Example:

SELECT * FROM STARTERHISTORY WHERE TRACK = 'GPX'
AND TRACKLAST <> 'GPX'
AND STARTSLIFETIME > 0
ORDER BY TRACKLAST

Hope I 'splained most of that in a way that makes sense.

-jp

.

~Edited by: jeff on: 11/20/2011 at: 9:59:25 PM~

ryesteve
11/21/2011
8:24:40 AM

Jeff, have you found this data to be useful? My feeling has always been that shipper success if a function of the trainer, which means you have to deal with the huge noise of small samples. But have you found ways to use it in aggregate? EG, "toss out UPR1 if the horse is shipping in from tracks a,b,c or d"?

tanix
11/21/2011
8:25:08 AM

Still in playlist mode... should have mentioned that :-)

jeff
11/21/2011
2:41:45 PM

Steve, As you know I generally avoid using micro data in my models (unless I can attribute results observed to a physical cause.)

Two areas where I've spent time working with micro data and enjoyed success are: trip notes and track bias. However, I'll re-emphasize understanding why the data behaves the way that it does helps greatly.

Not saying shipped from micro data isn't valid - there may in fact be some nuggets there - especially for those who can glean physical cause and effect from observations in the data.

In my own case I haven't gone there yet because there are only so many hrs in a day.

-jp

.

jeff
11/21/2011
2:50:45 PM

It's a little tougher in playlist file mode (but doable.)

Create a UDM:

1. Track Code is: GPX

2. Track Last Out (Avoid) is: GPX

3. Track Last Out (Require) is: a track code or list of track codes other than GPX

4. Starts (lifetime) min val is: 1

Run that through the Data Window. In playlist file mode you can't sort the output like you can in sql mode. So you'll have to experiment by analyzing individual track codes in the Track Last Out (Require) field.

Hint: Query speed can be improved upon by using steps 1,2, and 4 above to create a UDM which in turn is used to drive creation of a quick index file that is used for queries involving individual track codes in the Track Last Out (Require) field.

-jp

.

JCapper Message Board

JCapper 101 -- Analyzing shippers by track

JCapper 101
-- Analyzing shippers by track