My use of "identical" (with a number of different meanings) was misleading. Sorry. I blame my blunder on the after effects of insomnia from this debate.

Let's try again.

My data for 1200m races at Kembla has 29 days with both Maiden and Open handicaps. I used only those 29 comparisons to establish a relationship between the two classes.

Others would have averaged all the 43 Opens and 167 Maidens present.

But the more I think about this, the more comfortable I am that my method is more precise.

For this exercise the 138 unmatched Maidens are rubbish. Why allow them to contaminate the calculations?

Now even though either methods seems to work fine for Kembla, there are other situations where my method excels:

[*]Flemington 900m races (restricted to 2YOs) only occur in very early October. Comparing them with other times of the year is asking for trouble.


[*]75% of 2YO 1400m races occur in often rainy May, June, July.


[*]Very little data is available for new (extra) tracks like Randwick and Sandown.[/list]



These examples suggest that like for like comparisons (like mine) are far safer.


_________________
jfc
~

[ This Message was edited by: jfc on 2003-06-19 08:37 ]