THE BOOK--Playing The Percentages In Baseball

My problem with hockey-fights.com is their determination of wins. 

Perfect example would be Oct 15, 08
Thornton v Laraque
Laraque 38.5%
Thornton 38.3%
Draw 23.2
517 Votes

This is declared a win to Laraque.  Shouldn’t this be a draw?  The winner should need greater than 50% because the results are not binary.

I’d call it a win if the difference between the two is at least 10%.

So, 35 v 25, with 40% draw
or 45 v 35 with 20% draw
or 55 v 45 with 0% draw

Anything where the gap is under 10% is too close to call, and so, a draw.

Unfortunately, all I scraped was the winner, and his percentage.  I’m waiting for the site owner to send me all three percentages, and then I can run ELO using those values instead of binary outcomes.

This is genius!  Question: on the linked page I don’t see the actual Elo Ratings listed anywhere.  Have they been published?

Andy also has a method to handle this (no different than a strength of schedule).  He did it for the head-to-head ethics questions from last year.

GZ - I didn’t think the ratings were all that meaningful, so I left them out.  Steve C points out a problem with my dataset - I’ll post some numerical ratings once I get the right data.