Data Scraping World Cup Fencing
The skills the author demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Contributed byΒ Thomoas Kolasa. He attended in the NYC Data Science Academy 12 week full time Data Science Bootcamp program taking place between January 11th to April 1st, 2016. This post is based on his third class project - Β python web scraping(due on the 6th week of the program).
As one of the five original Olympic events, fencing has a storied history. While some fencers trace their maestros' lineage back generations, data shows the globalization of fencing creates new schools that make the sport evolve more than ever. The FΓ©dΓ©ration Internationale d'Escrime (FIE) governs the sport and runs the global ranking system.
Women and men compete in three weapons: Γ©pΓ©e, foil, and sabre. Fencers improve their rankings through placement at five World Cups, three Grand Prix, the World Championships, and the Olympic Games over the previous year. Each country can send up to 12 fencers to World Cups and Grand Prix and only up to four fencers to the World Championships or Olympics.
Fencing results are available at https://fie.org/results-statistic/result from about 2002 onwards, so I collected them using URL manipulation and Beautiful Soup 4. Each tournament had a name, place, country, date, and tournament type. I made sure to record the entire tournaments list and check if they spread over multiple pages. Since I also accessed the URLs of the specific tournaments, I was also able to gather the results for each tournament, including points earned
The results for each tournament included:
- Finishing placement
- FIE points earned
- Name
- Nationality
- Birth date
- Competition
- City
- Country
- Tournament dates
- Weapon (Γ©pΓ©e, foil, or sabre)
- Male/female
- Tournament type
Investigating the data, I calculated the age of each fencer on the day of each tournament. Looking at a histogram of all tournament winners, fencers seem to get gold medalsΒ in their mid to late twenties.
Is there home piste advantage in world cup fencing?
I nextΒ investigated if there is a home-field advantage in fencing like in other sports.Β I considered a home tournament to be inΒ a fencer's home country. This means that a "home" tournament would includeΒ an American from the west coast fencing in a world cup in New York or a Russian fencer from Siberia fencing in St. Petersburg. I also only looked at world cup and grand prix results between 2010 and 2015,Β the period when the FIEΒ standardized theΒ number of world cups. Even though Grand Prix award 50% more points, the same fencers show up in nearly all theΒ world cups and grand prix. I therefore normalized the points awarded at these two events to be equal.
Senior Men's Sabre average points earned at home (blue) and away (green)
world cups and grand prix, 2010-2015:
The average number of home points earned includesΒ the sum of points earned by local fencers divided by the number of world cups in that country. The average away points are calculated analogously, but for all other tournaments.Β However, this means the graph shows the averages ofΒ five world cups in ItalyΒ and only one world cup in South Korea.
To account for this, I calculated theΒ weighted average of home-field advantage. I looked at high performing countries that scored on average at least ten FIE points abroad so as to avoid the effects of more fencers from weaker countries participating in tournaments hosted at home. I next multiplied the home percent improvement of a country times the home tournament count and divided by the total number of tournaments hosted by high-performing countries. This yielded that men's sabre fencersΒ earned 42.7% more FIE points at home than abroad.
Does this advantage exist in Γ©pΓ©e?
I next investigated men's Γ©pΓ©e world cup and grand prix results. In Γ©pΓ©e, the referee does not have to determine which fencer has the right-of-way and thusΒ makes farΒ fewer subjective calls. I similarly looked at FIE points earned by fencers in their own countries.
Senior men's Γ©pΓ©e average points earned at home (blue) and away (green)
world cups and grand prix, 2010-2015:
Looking at a similar calculation of home country advantage for countries that on average score at least ten FIE points abroad, I found that men's Γ©pΓ©e has a home-country advantage of 26.9%.
Since men's sabre fencers had a 42.7% home country advantage, does this mean that right-of-way calls in sabre unfairly goΒ to local fencers? Possibly. But with such a small sample of data, many confounding factors can account for this discrepancy. Since sabre is more fast-paced than Γ©pΓ©e, the psychological and physical advantages of fencing close to home may have a different effect. It would also be worth comparing these results with the advantage in foil and in all three women's disciplines.
ToΒ more thoroughly look into the fairness of fencing, one has to look at bout-level data.
Bout Score Data
For the tournaments with bout data, I collected direct elimination bout data including the fencers, their countries, and the rounds when the bouts took place, and the scores. The inconsistencies in the FIE's site made it a time-consuming process, especially with regards to preliminary rounds.
Results for each bout included:
- round (64, 32, etc.)
- fencer data
- name
- country
- victory/defeat
- touches
- final placement
- birthday
- height
- weight
- dominant hand
- opponent data (same features)
- tournament data
A sample bout tableau from the quarterfinals onwards:
While I originally wanted to analyze bout data going back many years, the website no longer has bout data available before 2014. Consequently, there was not enough data to perform much viable machine learning analysis on the data.
A further issue with the available data is that very few tournaments release referee data. Even though refereeing has improved drastically with the advent of video replays, consistent referee data would allow for better reviews of performance and an evidence-based certaintyΒ of impartiality among officials.