For the unfamiliar, modern sport fencing (sword fighting) is divided into the practice of three different weapons: foil, épée, and saber. The weapons have different designs and bouts using each weapon follow a different set of rules. These differences change how the bout is fought.
As an épée fencer, I often hear that epee is less predictable than its cousins foil and saber. The recent October 2012 NAC in St. Louis provides an anecdote to support this – the #1 seed in Div 1 Men’s Epee, Andras Peterdi, was eliminated by the #192 seed, Anton Piskovatskov, a massive upset.
But is épée really less predictable than foil or saber? This is a statistical question that we can answer with data. So, I have just completed a new paper to answer the age old question - Is Épée Silly?
I encourage you to download the paper (it is only one page) – but the spoiler is that saber is actually much less predictable than foil or épée.
Some quick notes: In discussions with saber fencers, I have been told that the spread of ratings between sabreurs is not nearly as even as it is in épée. I have been told that while there are a couple hundred A-rated épéeists, there are far fewer saberists, and that the same holds for Bs, Cs, etc.
It is true that a lack of spread among sabreur ratings would reduce the correlation between the ratings and final results. This explanation fits; there were far fewer large saber events in the sample*, and ad-hoc comparisons of early (before 2011) saber events and more recent (since 2011) events has shown a large increase in the predictability of saber, as one would expect if there has been an increase in the number of higher-rated saber fencers.
Although interesting, this paper is just the beginning. I would like to analyze the data in other dimensions, and perhaps look at the predictability of DE results based on pool results.
* Although there were fewer saber events, there were still over 4000 individual saber results, so the results are not due to a small sample size. The issue is not statistical, but rather that the actual spread of saber ratings is uneven.
Paper: “Is Épée Silly? The use of USFA Ratings to Predict Results”, James Gibson, 11/14/12. PDF, 469KB.