As many know, for the past few years I have conducting research on the ranking of elite level Judo. Specifically, I have been running and experimental ranking system for Olympic qualification using the ELO ranking system and all the IJF ranking events.
To date (not including the 2011 Tokyo Grand Slam which is happening whilst I write this), there have been 17,542 fights in the qualification period. The qualification period started in the 2010 Tunis Grand Prix and there have been 64 events so far with approximately 150 nations and approximately 3,781 athletes all in the mix.
You’ll notice I said “approximately” there. That was intentional. Since the qualification started I have identified over 300 typographical mistakes in the names of athletes reported by the tournaments. Thats approximately 11% of error. Now, my data is sourced from Ippon.org and that technically is not the definitive record. However, it is generated directly from the tournament software used in the stadium, so I can imagine that the errors carry over to the official written records.
The IJF’s own ranking system is I think accurate (at least as far as I can see) this is mainly due to the diligence of the hard working lady who looks after the list. The IJF ranking list is maintained manually and it’s hard to fathom quite how much work goes into maintaining it! I know from my experiences running my experiment ( on http://www.rwjl.net by the way), that maintaining accurate information is very hard. I am lucky to have had lots of help from people, not least of all from http://www.judoinside.com
In my experimental ranking system for Judo, accuracy is vital. My system awards or deducts ranking points from athletes after every fight. Unlike the IJF ranking system, my system looks at every fight; the IJF look only at the final positions. So the IJF system requires less accuracy arguably, as you need only make sure the final results are right. After all, 50% of the names are gone after the first round at most events so need not be entered into the IJF system. On my system however, those names need to be right.
The reason they need to be rights is that points are awarded based on the predicted chance of player a beating player b in that specific match. As I calculate that chance for each fight, it matters as the ranking points are awarded after each match. Lets say you have a “Joe Bloggs” who wins points in competition A, getting a ranking of perhaps 1580. If they are mistakenly called “Joseph Bloggs” in competition B, then my system thinks they are a new person and there 1580 points are not applied and a new athlete enters the system.
This is bad as I would earn more points at competition B for beating a player with an existing rank of 1580 (Joe Bloggs) than a new athlete (Joseph Bloggs) who would start with 1500 points. In the IJF system, if Ilias Illiadis beat Joe Bloggs in round one it matters not at all if his name is wrong; because who you beat does not matter, only the position you get to. If Joe Bloggs makes it far enough to earn points, then there name being wrong would matter. However, as the IJF system is manual it’s perhaps going to get picked up at that point.
It is thanks to “encouragement” from others that I have put more and more effort into the accuracy of the names. I have about 330 corrections and as it’s automated so it should stay right. Of course, there is another problem which is one I have not yet attacked. What to do about athletes who have the same exact name! It is difficult to find workable solutions to that problem. How do you decide if two names that are identical are not the same person? How do you decide two identical names are not to different people?
As the software I have written matures I hope to find good solutions these sorts of problems. For now I have 330 specific corrections;I have found by manually going through the data and fixing mistakes people point out (please do check http://rwjl.net and see if you find any mistakes).
My future plans are to publish a final report on the experiment after the London2012 Olympic games, this will go along with the poster presentations I have given at the European and World Judo Championships. I also plan to open source the software and data, so that others can improve on what I have done and take the research further or in different directions. I have already created a data API, so that people that want it can easily obtain machine readable access to my record of the 17,542 fights that make up the qualification race for London2012. If you would like access please let me know.
Regards,
Lance Wicks.
