Using Elo to pick winners

Facebook turned up something interesting this morning, via the guys at the fantastic Swordsport Productions page:

msi elo

It comes from the guys over at The Fencing Coach. They’ve used the Elo rating system, which was developed for chess but is now used across a range of sports to track player performance and predict match results. It’ll be fascinating to see how it stands up to the experimental test tomorrow in Kazan. It pretty much exactly mirrors my predictions last week, which have been based largely on anecdotal observation of the 2013/2014 A-grade season. Here’s to a Kim v Reshetnikov final!

Those of you who’ve trained at Sydney Sabre may already be familiar with Elo: it’s the system we use for our internal ranking scoreboard. The data we collect is not as useful for predicting the results of a standard competition, as we include the results of matches fenced with handicaps, but it is  an extremely powerful predictor of match outcomes under our standard club training conditions, and functions very well in its primary role as a matchmaking system.

It’s worth noting that this isn’t based on the win/loss results over the season, as a true Elo model should be, only on the FIE ranking at the end of the season. The aim is apparently to do a proper Elo model over 2014/2015, and it will be very interesting to see how that stacks up against the more traditional FIE points-based rankings.

How about tomorrow’s comp? Here’s the big table:
World Champs table

As far as Friday goes, my personal suspicion is that an on-form Szilagyi has the edge on Reshetnikov if the two should meet in the semifinal. Szilagyi, however, is likely to face a couple of difficult early matches against the kind of awkward fencers he traditionally has trouble with.  Given the table above, my pick is for a Kim/Szilagyi final, in which Szilagyi will probably have the upper hand given Kim’s recent spate of injuries.

That said, margins in sabre are so small that attempts at casual punditry are unlikely to end well. This is why a true Elo model for 2015 would be so tremendously exciting.

Image: Wikimedia Commons
Image: Wikimedia Commons