Skating Statistics Project (Revived and In Action!)

[sk8indel]

Hey there everyone! I started a thread sometime last year about doing a skating statistics project for my IB Internal Math Studies Assesment. The time has come that I'm actually putting this thing into action. After much consideration, I have decided to focus on quads in men's figure skating: if they're worth the risk, and how necessary they are to win.

I've started by compiling every quad that was attempted in international competition (Japan Open, Grand Prix Events, Grand Prix Final, Four Continents, and Worlds) from the 2010/2011, 2011/2012, and 2012/2013 (basically this olympic cycle) seasons. I'm taking down the name of the competitior, the quads attempted, their base values, and the total net points that were awarded for the quad/quad attempt (so GOE plus falls if applicable). I'm then going to sort by country, event based on its time in the season, and by individual competitior to find out what percentage of their average total scores in competition was made up of quads/quad attempts.

If anyone has more suggestions on specific things to investigate with this data, they would be highly appreciated.

Right now I have things on paper, but I'm going to make some excel charts after I have all my data. I'll update this thread with my findings and keep you all posted.

[lulu]

Don't have anything substantial to add, but I look forward to reading the results of your study when you finish! Sounds like an ambitious project, hope you have fun with it.

[RFOS]

Sounds fun! I'm a skating and stats geek too.

[TwizzlerS]

You could look at quad attempts in relation to base values. Haven't these values changed over this time period you chose?

[Belinda]

Your project will offer a great deal of food for thought to all of us here, and I wish I were a statistician so I could help you more!
I am just thinking aloud here but, given your objective, ie to establish how necessary quads are to win, it may be worth considering the following:
1) how would you remove the effect of "home country advantage" and "co-nationality of judge and skater"? You might need to note, as regards the event, whether it was held in home country of the skater, as well as time of the event in the season.
2) The ratio of points earned by a quad to total scores. Are you going to separate TES and PCS? Are you planning to test the theory that the success or failure of an element (say a quad) should not affect PCS (other things being equal)?
3) The use of "average" - I am always a little uncertain about this. For example, when Mao won over Akiko at NHK trophy this season, there was a view that somehow the judges or JSF or whoever willed this to happen. But the final scores the skaters got are only sum totals (except max and min) of individual judges' scores (divided by the number of the judges except max/min). Not one of the judges actually voted for the final PCS for each component. One might need to focus on the behaviour of individual judges for example in terms of how varied their scores are when errors are made in elements. Because of judge anonymity we can't trace individual judges' scores but for each skate we can measure the variance for example. A failed quad of some skaters may divide judges, those whose PCS scores are not affected and those who may give lower PCS when failure happens, for example.
Sorry if I am not as lucid as you would like me to be. These are just my initial thoughts.

[sk8indel]

TwisslerS- I've been slaving over protocols, and I'm pretty sure the base values changed prior to the 2010/2011 season. A 4T in the first half of a program has had a base value of 10.3 since then, and still does now.

Belinda- I'm keeping track of the country each competition was held in, I just don't know how to ultimately factor that in in the end.

So far I've made a pretty extensive table (175 rows by 13 columns). I decided to throw out the Japan Opens because only having a long program score was messing up my variables. The data I've collected and recorded is the name of the competition, the location of the competition, the date of the competition (so I can later see early season vs. late season results), the name of the skater, their home country (to look at country advantages and see which country holds the skaters that try the most quads - just because I'm curious), and then I have a bunch of numbers. I started out by recording the base values of all teh quads attempted, added or subtracted GOE, and took off for any falls that were a direct result of the quad attempt. This gave me a raw score for the quad. I then added the raw scores for the quads in a single competition and added them to get a total competition raw quad score, which I then divided by the number of quad attempts there were by that person in that competition to get the average points garnered per quad attempt. I also made note of if the quad was done in combination, but didn't include the other jump in the combination in the quad raw score. I then put this side by side with the total competition score, and the placement (ordinal) in the competition. I think next I'm going to look at the percentage of the total score that was made up by points garnered from quad attempts and sort everything out by individual skater.

Sorry if that sounds super confusing. I guess I'm stuck on how to specifically evaluate if quads are 'worth it'.

[gkelly]

TwisslerS- I've been slaving over protocols, and I'm pretty sure the base values changed prior to the 2010/2011 season. A 4T in the first half of a program has had a base value of 10.3 since then, and still does now.

I think what changed is that for the 2010 season (was it only one year?), the value of the negative GOEs on quads and triple axels was 1.5 points for each minus, vs. 1.0 for other triples. Then that was changed so that the quads and 3A had 1.0 per minus and the other triples had less. I.e., the penalty for mistakes on all difficult jumps was higher in 2010 than it is now.

The other major change was the added distinction between downgraded jumps (base values of jump with same takeoff and one less revolution) and "underrotated" jumps (70% of initial base value).

Even more than the GOE penalties, that is now a less harsh penalty on moderate underrotation (91-180 degrees) and doesn't even require a negative GOE if a judge sees the jump as good in real time.

Both those reasons now make it more worthwhile to attempt a jump in competition that is likely to be close but slightly flawed. Under the old rules, an apparently clean 91-degree cheated quad was worth less than a decent triple; now it's worth significantly more, though not as much as a fully rotated clean quad.

I think both these changes took effect with the 2010-11 season, but that's just from memory, I could be off.

[sk8indel]

So this is the spreadsheet that I have so far. This is sorted by country, skater within the country, and then scores in chronological order for the skater.

https://docs.google.com/open?id=0B_r...jY1NXJrNFBLcFE

I've also added a Summary sheet at the end. It's organized by skater and has the number of competitions the data is from, the number of quad attempts they made, the average score per quad attempt, the average of all their total score, and the average of all their placements just for fun.

If anyone has any suggestions on what data to use to bring about some kind of conclusion, that would be awesome. Thanks!