Good for you, ISU

[alyssamarie]

For number 3 yes there is bias in figure skating - so much that I did my masters thesis on it!

 

[Isk]

Personally, I think figure skating judging system before IJS was a joke and the sport was unwatchable because of it. And I didn't watch it, lol. Now it's clear what they are going for, and once AI is there, TES would be solid.

I am afraid people are not going to like this, sorry about this, but since this can of worms is opened, here goes. Perhaps you didn't watch figure skating before IJS, but it was not only watchable, it was mega popular. The way US TV handled this was they hired good commentators, among them Dick Button, Peggy Fleming and others. Dick was very good at explaining what skaters did and how it mattered. They did a review of required elements, types of jumps, compulsory dance patterns, etc. with visuals, so that the audience would understand what was going to happen. They also did previews on athletes. It was common for a reporter team to fly to other countries to film documentaries about contenders before big events, so the audience would be introduced to the skaters, their lives, training condition, struggles, etc. They would make a show out of it. There were also pro championships where retired skaters did all sorts of creative stuff, and those were quite popular for a while. There were shows that aired on TV in prime time, in fact, I'd never have survived through my first hearing of Holst's Planets suite without this movie: https://www.youtube.com/watch?v=LnQDDH7rF5o&list=LL&index=261. My first intro to the full story of Carmen was through the Carmen on Ice show. There was no youtube back then, but nowadays you can learn quite a bit about classical music and cinematography by watching historic figure skating.

One aspect of 6.0 system, which I consider important, is the overall impression. IJS decomposes skating into many components: 18 elements, for which there a is base value, 11 grades of execution, and 3 program components, for which there are a number of criteria and a range of marks. It is a lot of information to process simultaneously. In 6.0 you'd have to impress. You could come to an event with a weak program, do the jumps and not much else, and lose. John Curry and Robin Cousins certainly were weaker than the competition jump-wise, but they won in 1976 and 1980 by being much above the competition in non-jumps, choreo and skating skills. It doesn't work like this in IJS anymore. It did early on, but not anymore. There were some skaters who were not so strong jumpers or not so great skaters, but they were strong entertainers and played this strength. Overall, it was not so easy for an unspectacular skater to win a big event because they managed to get everything done and scored the points as it happened in 2010. Also was unlikely for someone with poor aesthetics and skating skills to win two WC in row as it happened the last two years. They'd have to be entertaining, have charisma, be good in choreo, have some skating skills, strong programs and stay on their feet. Having said this, there were a lot of weak skaters and crappy programs, but overall, skating was certainly watchable and gave an opportunity for creative skaters to create because rule-wise it was less restrictive. I am not saying IJS is evil, rather that it would be good to have the best of both worlds.

Of recent developments, Zagitova rule certainly took away from program variety and pushed the quad race among women. But the other innovation I question is the introduction of lyrics. I understand the reasoning behind this, and sometimes lyrics are great, they can add expression to music and certainly expand the range of music you can use. But the result is that many people stopped bothering about music at all. Many just take a popular track that will be forgotten in a season or two, oftentimes there is nothing to express in there. It's a stanza+refrain+stanza+refrain+a bit of culmination+ending, the same rhythm, the melody repeats. Same volume level, no <>, rubato, or other nuances that typically require choreographic interpretation. There is some happy or unhappy love, so we have 20 girls skating unhappy love 20 times in a row. These kids then become choreographers. Maybe I am imagining it, but it seems that music quality noticeably dropped. There are other changes I don't like: when you watched compulsory dance, you'd pretty much see who could do what, now they don't need the skill. The kids are skating to the music from the 90s this season, which doesn't mean anything to them because they didn't live through the times, didn't dance to it at discos, didn't exchange tapes at school. Music from the 90s was different in different countries, so the whole idea of finding some characteristic vibe of the 90s is wrong. IMO dance is obscure as ever and has become unwatchable. I never liked it as much as singles and pairs but it used to make some sense.

 

[Anna K.]

For number 3 yes there is bias in figure skating - so much that I did my masters thesis on it!

Can we read it online?

 

[Mathman]

My first intro to the full story of Carmen was through the Carmen on Ice show. There was no youtube back then, but nowadays you can learn quite a bit about classical music and cinematography by watching historic figure skating.

I can remember back when the only time children in the U.S. would hear any classical music at all was either a figure skating show or a Buggs Bunny cartoon. https://www.mentalfloss.com/article/66672/15-pieces-classical-music-showed-looney-tunes

 

[Mathman]

However, the PCS remains the Achilles' heel of the system.

Is it the Achilles Heel or is it the last heroic stand of the old guard straddling the highway of progress and shouting, "STOP!"

 

[Isk]

Watch this from Euros 1996 and tell me this was not watchable:

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

www.youtube.com

Awesome jumping technique, nice Chopiniana music although sticking Paganini in the middle of Nocturne no. 2 is a bad idea

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

www.youtube.com

Here is some Disney, he had a meltdown, but awesome lutz technique, program packed with choreo. Decent music cuts considering they glued tape together according to TAT. Scored surprisingly well...

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

www.youtube.com

Here is some Paganini Webber style with pretty decent jumping

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

www.youtube.com

Shosti's Hamlet

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

www.youtube.com

And this is an entertainer, he is not a friend of aesthetics, and he had better programs frankly. Public loved this kind of stuff, but winning with this kind of skating was out of the question

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

www.youtube.com

Here are some unusual selections from von Suppe, Tchaik and Minkus. There is stuff in the ladies event from the 1970s, Danny Elfman, circus music, a Malaguena, and a Miss Saigon. You can study music watching this. Some skaters did nice extended spirals.

I just wish sometimes ISU would look back and learn from the past.

 

[bostonskaterguy86]

The way US TV handled this was they hired good commentators, among them Dick Button, Peggy Fleming and others. Dick was very good at explaining what skaters did and how it mattered. They did a review of required elements, types of jumps, etc. so that the audience would understand what was happening. They also did previews on athletes. It was common for a reporter team to fly to other countries to film documentaries about contenders before big events, so the audience would be introduced to the skaters, their lives, training condition, struggles, etc. They would make a show out of it.

I'm convinced that more people would be interested in watching skating if this was still how TV coverage of it was done. I feel like Tara and Johnny in particular try to cram all of the above into their commentary during programs - talking about skaters' backgrounds and past achievements, explaining aspects of the scoring system, discussing the relevant technique and relative difficulty of various program elements - which leads to their commentary taking up way too much air during performances.

Meanwhile, it's way too difficult for the average skating newcomer who doesn't already know all of this to absorb what they're saying while also trying to pay attention to the actual skating - which IMHO is why you hear so many people say they can't get into skating because they don't understand what the skaters are actually being judged on. I hear this a lot in particular from people who were fans during the 6.0 era, and I totally get it - two easily-understood marks on an easily-understood scale were replaced by a set of mysterious numbers calculated to two decimal places based on a highly detailed set of criteria that take a considerable amount of study to truly understand!

While I do believe that IJS has been an overall net positive for the sport, I think that for audiences here in the USA, the timing of its implementation very unfortunately coincided with several other factors that were already contributing to a decline in TV viewership - the 2002 Olympic judging scandal, the perceived lack of success of American skaters in disciplines other than women's singles, and the perceived lack of a clear successor to Michelle Kwan as it became more obvious that her elite competitive career was winding down (note that I use the word "perceived" here - we all know that there are some counterexamples to the narrative of "the USA isn't competitive outside of women's singles and there's no one coming up who can compare to Michelle," but this is true to what many casual fans thought at the time, and it was subtly reinforced by media coverage). As viewership declined, it made less and less sense for the TV networks to invest time and money in covering figure skating, and thus a lot of the dedicated educational / explanatory segments wound up on the chopping block - just as they were more necessary than ever to help viewers understand the new judging system. The comparative complexity of IJS scores, combined with a lack of content explaining them, drove even more American viewers away - setting up a negative feedback loop that would make it even less likely that networks would devote any additional time and money to covering the sport.

 

[NanaPat]

Another change the ISU has made, which I applaud, is the change in how the Olympic spots are allocated.

There was much griping in 2018 about Canada having 3 women's spots when they only had 2 competitive women. The new system has corrected that type of situation. Interestingly enough, it's biggest effect hasn't been on countries that earn 3 spots at Worlds but less (initially) at the Olympics, but on those countries that earn 2 spots at Worlds but only one (initially) at the Olympics. Some of those 2-spot countries don't even have a second team to send to the qualifier!

Of course, the biggest effect of the change is on the 1-skater countries that now earn their Olympic spots at Worlds and don't have to go through the qualifier. It saves them time and money and helps them get support for their Olympic trip!

 

[Ziotic]

Couple of thoughts.

The Zagitova rule didn’t stop people from being able to do 7 jumping passes in the last two minutes, it just stopped giving additional points for it. So you can’t blame the rule on how people distribute their jumping passes.

Also, if you have ever skated you would know doing the jumps spaced throughout a 4 minute program is actually harder than a 2 minute jumping drill. However, when that 2 minute jumping drill goes wrong it goes REALLY wrong (worlds 18).

As for the PCS it could use some work, but will always have an inherent biased because we all don’t view certain skills the same.

I would love AI to qualify jumps, spins, and ice coverage. For the average viewer some people don’t see UR and think they don’t matter since it doesn’t detract from the skating, but it is a sports and how many revolutions you do in the AIR matters.

 

[Mathman]

Couple of thoughts.

I would love AI to qualify jumps, spins, and ice coverage. For the average viewer some people don’t see UR and think they don’t matter since it doesn’t detract from the skating, but it is a sports and how many revolutions you do in the AIR matters.

That's an interesting point. In American football, for instance, there is instant replay to clarify whether the runner stpped out of bounds or not, whether the defender illegally interfered with someone trying to catch the ball, whether the player's knee touched down before he lunged forward for a first down, etc., etc. Sometimes this technology sheds light, sometimes it is still unclear after many slow-motion replays from different angles.

The TV commentators offer their opinions as do the fans. and all in all I think this is a positive and fun exercise for the audience. It has become part of the game.

In basketball it is a common practice that if the referees call a ticky-racky foul on one team, then on the next play they call an equally questionable foul on the other to make up for it.

 

[Mathman]

*Wh0le post*

Thank you for a thoughtful perspective.

 

[Isk]

That's an interesting point. In American football, for instance there is instant replay to clarify whether the runner stpped out of bounds or not, whether the defender illegally interfered with someone trying to catch the ball, whether the player's knee touched down before he lunged forward for a first down, etc., etc. Sometimes this technology sheds light, sometimes it is still unclear after many slow-motion replays from different angle.

This is not AI, it's slow-motion replay, as far as I understand you. AI is something you give an input, e.g. a phrase in one language or a footage of a jump, and receive an output, e.g. the same phrase in another language or the jump label. It's designed to emulate neuronal connections in the brain, in other words identify some patterns in the data the way brains supposedly do it (from math point of view it is just a function approximation, well, "just a function" is too simple, it is a complicated non-linear function approximator). This is done by training AI, i.e. solving a large optimization problem, in which parameters of a neural net (usually millions of them) are optimized and tuned so that for certain inputs for which the outputs are known, AI produces those known outputs. These inputs with known outputs are called training data, and typically one needs a lot of training data. AI is only as good as the training data it learns from. If there is one camera, and tech.panels mislabel the footage on that one camera because they can't properly see, and these data are then used to train an AI, mislabelling will propagate into the AI. In other words, if the data have mistakes, AI will learn them. Garbage in, garbage out. AI is effective at solving certain problems, but it has limitations, and those are relatively well understood and formulated, although there is no satisfactory general theory. For something as complex as a spin or a step sequence, for example, one would need a lot of good quality training data and smart net architectures, perhaps recurrent, because it's more of a time series. Each time the rules are changed and broken into, the data will have to be regenerated, relabelled, the net retrained, retested, etc. I am not sure if it's worth it. It is not an easy problem.

 

[Anna K.]

Watch this from Euros 1996 and tell me this was not watchable:

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

www.youtube.com

Awesome jumping technique, nice Chopiniana music although sticking Paganini in the middle of Nocturne no. 2 is a bad idea

Thank you for a flashback!
Figure skating has never been covered this way in Latvia, so it was very interesting for me to discover how it worked and that this is the same format that is today used for talent shows. actually.
It also reminded me how much the entertainment has changed. How many talent shows and reality shows were there in the eighties and early nineties? And how many were there at 2000? So, maybe figure skating had to become a sport, have IJS, and live on IOC money from then on because the ratings had fallen already and it could not compete with more successful talent shows?
In either case there is some material to think about

 

[4everchan]

I agree that the rules do not prevent anyone from front loading or back loading. Roman doesn't use the bonus in the sp and one of his programs had the three spins as the final elements. Pairs still can do quads though they are worth very little more than a triple with the base value changes for twist and throw. Since that change only two teams have done the quad twist.... So, fair enough, rules have not formally prevented some of the things to happen but the lack of incentive certainly has.

 

[Mathman]

How many talent shows and reality shows were there in the eighties and early nineties? And how many were there at 2000?

Everything old is new again.

In the U.S. there was a long-running and very popular radio show called Major Bowes' Amateur Hour later moved to television as Ted Mack's Amater hour. Listeners/viewers on New York called in on the telephone to vote for their faves in real time with the show in progress.

What was cool about the radio version was that a very popular;lar form of radio entertainment was a ventriloquist act. Think about it. A ventriloquist act on the radio. Now that's entertainment!

 

[Mathman]

Roman doesn't use the bonus in the sp and one of his programs had the three spins as the final element.

The good old days when spins were valued and spotlighted as the principal crowd pleasers.

 

[Mathman]

This is not AI, it's slow-motion replay, as far as I understand you. AI is ...

Thank you for that detailed explanation. I think that the primary obstacle for figure skating would be the research and development costs. The second would be the challenge in addressing esthetic considerations.

I actually have a little experience with chess-playing programs. Back in the glory days of the 1990s the race was on to produce a computer program that could beat the human world champion (Gary Kasparov at that time). Their strategy was to work on speeding up the hardware until it could generate and evaluate 100,000,000 or so positions per second. No luck. Finally, if you can't beat 'em, loin 'em. IBM invited Kasparov to join their team and the first thing he told them was that they didn't know anything about chess and so Deep Blue was evaluating certain features as negative when they were actually posirive, etc. Very quickly Dep Blue was baeting up on Kasparov and succeeding world champions with regularity.

Mu favorite artificial gamester came out last year when the University of Tokyo announced Janken, a robot that plays Rock, Paper, Scissors and has a 100% winning percentage against humans. It works by scanning the human opponent and analyzing tiny muscular twitches that enable it to anticipate what sign the human is preparing to throw, and then responds with it's robot hand with the winning response in a thousandth of a second.

(Not exact;tly, AI, though.)

 

[Anna K.]

My favorite artificial gamester came out last year when the University of Tokyo announced Janken, a robot that plays Rock, Paper, Scissors and has a 100% winning percentage against humans. It works by scanning the human opponent and analyzing tiny muscular twitches that enable it to anticipate what sign the human is preparing to throw, and then responds with it's robot hand with the winning response in a thousandth of a second.

A robot PCS judge sounds like a tool for a primitive children's game when compared to this one

 

[Isk]

Thank you for that detailed explanation. I think that the primary obstacle for figure skating would be the research and development costs. The second would be the challenge in addressing esthetic considerations.

I actually have a little experience with chess-playing programs. Back in the glory days of the 1990s the race was on to produce a computer program that could beat the human world champion (Gary Kasparov at that time). Their strategy was to work on speeding up the hardware until it could generate and evaluate 100,000,000 or so positions per second. No luck. Finally, if you can't beat 'em, loin 'em. IBM invited Kasparov to join their team and the first thing he told them was that they didn't know anything about chess and so Deep Blue was evaluating certain features as negative when they were actually posirive, etc. Very quickly Dep Blue was baeting up on Kasparov and succeeding world champions with regularity.

Mu favorite artificial gamester came out last year when the University of Tokyo announced Janken, a robot that plays Rock, Paper, Scissors and has a 100% winning percentage against humans. It works by scanning the human opponent and analyzing tiny muscular twitches that enable it to anticipate what sign the human is preparing to throw, and then responds with it's robot hand with the winning response in a thousandth of a second.

(Not exact;tly, AI, though.)

I am actually not sure how they teach chess programs, but I think Janken is AI. It's not such a hard task to train a robot to scan hand muscles and predict what the hand is going to do next, and it's very fast when trained. This doesn't mean the robot will be able to sit an anatomy exam: I think it's based on image analysis. If one wants it to predict another response, e.g. related to leg muscles, they may have to retrain it.

 

[Mathman]

I am actually not sure how they teach chess programs...

I believe that there are several different approaches that people experiment with. The easiest method would be just to let the computer make the move that it likes best, then if it wins the game that type of move in that type of position is advanced, and if the computer loses, then that move is demoted.

On the amateur level, the most popular chess program currently on the market allows the human player to select the level that he or she wants the computer to play at. If you set it at. say, the "expert" level (2000-2200 rating points), then when it is the computer's turn to move it searches its huge data base of actual games played bt experts and sects the move the the computer sees most frequently. If you set the computer at "master" level (222-- - 2400 points) then it searches its data file of games played by human masters instead. It is actuallylly quite instructive to play over the same game at different levels and learn how (human) players of different strengths react to the same position.

I agree that this is not AI -- it is the human player that is learning rather than the computer.

[Aside: To chess fanatics there is currently a big crises because the newest vesrsion of the Apple operating system does not recognize this "select the level" feature on the default program that comes free with Apple computers. (The chess software hasn't been updated in years, and the OS has moved on. Plus, I think that there is a problem involving tariffs or copyrights or omething of that sort that hinders revision of this program, developed in China in happier times.)