For those who still don’t know, an innovative and forward-looking company, to give judois a pioneer in match analysis in sports.
While there was once an international collaboration to develop this incredible data set, it all originated with former athlete and current director of performance analysis for the Italian Judo Federation, Emidio CENTRACCHIO. Emidio is the eldest of the team Centracchio, Tokyo Olympic bronze medalist, Maria and current European junior silver medalist, Luigi are also judo family members.
It all started when I was an athlete and met some Eastern European coaches in Poland, namely Janusz Pawlowski, Marek Adam from Gdansk University of Sport and the beloved Ryszard Zeniawa who succeeded. When I was there, I was 18 and they gave me a reductive system and with that system they collected all the key technology and data for a contest.
At that moment I had a dream, I asked him, why don’t I do it with a computer and why don’t I do it automatically?
We’re talking more than 20 years ago, at that time the technology wasn’t very advanced and so I had to stop, and obviously I was an athlete so my goal was different. Then when I finished racing in 2012, that dream was still there and I started looking for some technologies that could help me get the job done. I am a judoka player but also the son of a math teacher and I come from scientific studies so it is easy for me to deal with numbers and with this in mind I started creating this manual system.
When we were collaborating internationally, I worked with British judokas and Nigel Donohue, who was actually a big promoter of the project, so some of my business comes from Nigel. It’s very easy for me to talk and interact with them and I have a good partnership with them to develop the tools that I give to everyone.
Today at the Olympic Centre, where the senior national team meets before leaving for Nymburk OTC, we also have an Emidio. Although he works closely with coaches and athletes, his presence at the European Rome Open has another purpose. In Mat 2, the central tatami, four barely perceptible fixed cameras collect footage and data that must be processed by a coding analysis system and the collected data will be used to “train” the artificial intelligence (AI) system.
I started thinking about an AI system because at present, I manually can’t detect all the competitors, because the time needed is large, so before the pandemic situation, I started with an Italian company to study whether it could be done and it was with good accuracy. I have achieved 70% accuracy in recognizing koshi waza, ashi waza, te waza and ne waza with just four cameras in the clubhouse so with a robust system in a real situation I am confident it will work it will definitely be ready to record the Olympics and after the Olympics I’m sure I can provide an up-to-date system that recognizes technologies.
There are many sports that we can say are easier to analyze because the actions are more fragmentary and the points are easy to analyze, as are the technical actions while judo is very complex and very complex. So I said how can I measure what is hard to measure? I started thinking about a system that could log everything that happens, and then translate those records into numbers to provide statistics that could show KPIs [KPIs] of the athletes.
The artificial intelligence system that Emidio intends to redevelop will use the data collected this weekend in Rome to contribute to a more efficient system, detecting not only the techniques, but also the information that we currently lack such as for example the “pointless” attacks , which accounts for 91% of movements as effectiveness currently averages only 9%. For many coaches, athletes, referees and the media, statistics allow us to share information in a universal language. When asked how he feels about people’s views that statistics remove the “feel” element from sports, Emidio makes an obvious point.
The most important thing in judo is judo and how athletes feel about judo and how coaches can train and make athletes understand judo and what they need to do. Statistics and match analysis simply are, and I’m not saying this to devalue what I do, but because it’s just a tool that helps the work of athletes and coaches.
Obviously it’s a really new tool and it’s hard to change what wasn’t the case in the past. One thing should be underlined, because video analysis has always been done, when I was a young athlete there was only the videotape of the Olympics and we watched and rewatched these same races to see what the champions were doing. Then times and technology changed, as an athlete I had a portable video camera so we started recording it ourselves and today we have the power of the internet and the great effort of EJU and IJF to make people see everything.
The goal of building an AI system is to unlock the analysis for each level. We must remember that match analysis was first recorded in the 5th century BC in The Art of War by Sun Tzu and we must always remember to know yourself, know your enemy and the outcome of 100 battles will never be in doubt. So you can study your opponent but you have to know yourself which is why the system will help younger athletes to develop.
This project started before the pandemic but now that it’s back in action, are there any technological hurdles to overcome?
I think now in terms of technology it is powerful enough to not give us any problems in terms of data collection. The challenge is to first create a system that logs everything, in two phases. One is annotation parsing, so annotating what is happening and another part is detailing the logs. This is a second phase because first we need to record as many events as possible, detail as many as possible, and then retrain the system, but from my study before the pandemic, I don’t think I will have any problems.
Not only will it help those using the data, but Emidio himself currently spends an unfathomable amount of time analyzing existing footage.
To manually record this information, I have to have the same time as the competition, twice, so when I see events like the Almada Grand Prix with over 500 athletes, I know how difficult it can be. Obviously this job gives me a different view of the competition and now I understand a lot of things and as head of performance analysis of the Italian team I can better explain to them why something is like this and try to help them. Training and understanding of behaviors they should avoid.
Emidio’s goal of efficiency in data analysis is related to Jigoro Kano’s use of the Japanese phrase “maximum efficiency with minimum effort”, albeit from a modern perspective. Explore the full dataset on the website to give judo!
Author: Tea Quinn
Involved
tweet
cable
Whatsapp
e-mail
