If judges were strict with prerotation?

[AxelLover]

Off the top of my head, examples of current skaters whose lutzes (and flips) have minimal (at most 90 degrees) prerotation:

Men: Boyang Jin, Han Yan, He Zhang, Yuzuru Hanyu, Lucas Tsuyoshi Honda, Keegan Messing, Jason Brown, Jimmy Ma, William Annis, Ilia Malinin, Brendan Kerry, Daniel Samohin, Kevin Aymoz, Michal Březina, Matyáš Bělohradský, Deniss Vasiļjevs, Paul Fentz, Slavik Hayrapetyan, Kornel Witkowski, Nikolaj Majorov, Mikhail Kolyada, Ilya Yablokov

Women: Tomoe Kawabata, Mako Yamashita, Karen Chen, Alexia Paganini, Emmi Peltonen, Elizaveta Tuktamysheva, Veronika Zhilina

 

[yume]

Off the top of my head, examples of current skaters whose lutzes (and flips) have minimal (at most 90 degrees) prerotation:

Men: Boyang Jin, Han Yan, He Zhang, Yuzuru Hanyu, Lucas Tsuyoshi Honda, Keegan Messing, Jason Brown, Jimmy Ma, William Annis, Ilia Malinin, Brendan Kerry, Daniel Samohin, Kevin Aymoz, Michal Březina, Matyáš Bělohradský, Deniss Vasiļjevs, Paul Fentz, Slavik Hayrapetyan, Kornel Witkowski, Nikolaj Majorov, Mikhail Kolyada, Ilya Yablokov

Women: Tomoe Kawabata, Mako Yamashita, Karen Chen, Alexia Paganini, Emmi Peltonen, Elizaveta Tuktamysheva, Veronika Zhilina

Other Women: Rika Kihira, Wakaba Higuchi, Nana Araki.

 

[AxelLover]

Other Women: Rika Kihira, Wakaba Higuchi, Nana Araki.

You're definitely right about Nana Araki. I think that Rika and Wakaba prerotate their lutzes a bit more than the others I mentioned but it's still very good.

 

[CanadianSkaterGuy]

You are always so overly dramatic that it's actually ridiculous. As you can find out there are works being done to assist in judging using AI to assess jumps. Also it definitely should be possible to get additional tech panel person at least for the biggest competitions.

Seeding out AI to assess jumps would prove costly, not to mention, skaters aren't like long jumpers or a Hawkeye where it's cut and dry assessment in a "linear" sport (there's a static court, or just one runway/pit). Every skater is different, whether it's their height, their technique, the angle at which they're facing the AI bot (again, skating isn't a "linear" sport, it uses an entire ice surface and skaters approach jumps with different angles and techniques). Even things like those ice scopes that assess the height of a jump I'm skeptical of, because it uses formulas and algorithms and there is no proof as to the accuracy of it. If people want to treat that like it is accurate since it's the first shiny thing that comes along to assess jumps, great, that's on them.. but it's like using a pocketwatch to clock a 100 m dash -- it's not nearly developed or sophisticated enough to assess jump movements. Nice thought to use AI, and might be great for skaters in practice but technology isn't at a point where such a thing would be either accurate for serious competitions, let alone accessible.

 

[CanadianSkaterGuy]

I remember, after the Papadakis/Cizeron scoring incident at European Championships 2020, someone from the Technical Panel said that they had had a camera failure and that was the reason for the delay and maybe inaccuracy. This person also said, that 80% of the time of a Tech Panel consists in counting rotations in spins. Maybe the ratio is different for Single Skating, but counting rotations, either in spins or jumps, should definitely be done automatically.

For spins this is rather difficult for an AI to do because the camera would have to determine when a position has been hit, and with so many variations of positions, it would be hard for a camera to assess them (outside of basic positions).

A skater couldn't for example do arm movements in a spin position because the camera would consider it an attempt at a new position.

 

[DizzyFrenchie]

For spins this is rather difficult for an AI to do because the camera would have to determine when a position has been hit, and with so many variations of positions, it would be hard for a camera to assess them (outside of basic positions).

A skater couldn't for example do arm movements in a spin position because the camera would consider it an attempt at a new position.

Of course it needs a software and a good deal of sampling, but this has been achieved in Gymnastics: why not in Figure Skating?
We don't have the whole of Yuzuru Hanyu's Graduation Thesis, let alone in English, but I understand that he managed to conceive such a software, to determine if jumps are correct with only one camera.

 

[CanadianSkaterGuy]

Of course it needs a software and a good deal of sampling, but this has been achieved in Gymnastics: why not in Figure Skating?
We don't have the whole of Yuzuru Hanyu's Graduation Thesis, let alone in English, but I understand that he managed to conceive such a software, to determine if jumps are correct with only one camera.

I would love to see how he conceived that as feasible. One camera would hardly be sufficient.

For example, Hawkeye used in tennis uses six, sometimes seven, cameras that work together to create a 3D projection of a ball’s trajectory. And that’s a tennis ball, which has a comparatively simpler/linear path compared to the complexity of a skater’s blade/pick position on jump takeoff and landings.

Hawkeye is also calibrated for accuracy and a system for figure skating would need to be recalibrated for every figure skater as not every skate/blade is the same size or type.

Not sure where Hanyu’s one camera would be mounted in the skating rink but it would have to be somewhere that somehow accurately captures takeoff and landing and be sophisticated and sensitive enough. If a skater is doing a jump at the other end of the rink is the camera less accurate? If a skater’s jump entry is oriented parallel to the camera on takeoff versus or perpendicular, does it make a difference in the reliability of the assessment?

Technology simply isn’t there yet. Theorizing is fine, but it's just theory and not actually applicable. Otherwise why not get tech specialists to put on a pair of special glasses that use augmented reality to 100% reliably calculate the spin rotations and jump rotations and count all the transitions and average speed just by looking at a skater. Nice thought sure, and someone could write a thesis about such a technology, but its application would not be remotely feasible for the foreseeable future.

 

[DizzyFrenchie]

I would love to see how he conceived that as feasible. One camera would hardly be sufficient.

For example, Hawkeye used in tennis uses six, sometimes seven, cameras that work together to create a 3D projection of a ball’s trajectory. And that’s a tennis ball, which has a comparatively simpler/linear path compared to the complexity of a skater’s blade/pick position on jump takeoff and landings.

Hawkeye is also calibrated for accuracy and a system for figure skating would need to be recalibrated for every figure skater as not every skate/blade is the same size or type.

Not sure where Hanyu’s one camera would be mounted in the skating rink but it would have to be somewhere that somehow accurately captures takeoff and landing and be sophisticated and sensitive enough. If a skater is doing a jump at the other end of the rink is the camera less accurate? If a skater’s jump entry is oriented parallel to the camera on takeoff versus or perpendicular, does it make a difference in the reliability of the assessment?

Technology simply isn’t there yet. Theorizing is fine, but its just scientific theory and not actually applicable. Otherwise why not get tech specialists to put on a pair of special glasses that use augmented reality to 100% reliably calculate the spin rotations and jump rotations and count all the transitions and average speed just by looking at a skater. Nice thought sure, and someone could write a thesis about such a technology, but its application would not be remotely feasible for the foreseeable future.

I see what you mean. I was very surprised by his assertion but he wasn't born with last rain as we say in France, and he has brains, and integrity. So, if he says so, it must be feasible and reasonably reliable. I believe this wouldn't be a regular camera, for instance I suppose it would have at least two lenses. You see, the need is to have something easily transportable and installable from a skating rink to another alongside other judging IT material. It doesn't need to be really cheap. I think, even if it's 1000$ it's OK, and the software would be integrated to the judging software.

 

[CanadianSkaterGuy]

I see what you mean. I was very surprised by his assertion but he wasn't born with last rain as we say in France, and he has brains, and integrity. So, if he says so, it must be feasible and reasonably reliable. I believe this wouldn't be a regular camera, for instance I suppose it would have at least two lenses. You see, the need is to have something easily transportable and installable from a skating rink to another alongside other judging IT material. It doesn't need to be really cheap. I think, even if it's 1000$ it's OK, and the software would be integrated to the judging software.

$1000?! For comparison, a Hawkeye system in tennis costs about $60000 USD, and an alternative PlaySight is about $12500 per court.

Not to mention that's to cover a pro tennis court that is 23.77 m x 8.23 m (10.97 for doubles), compared to an Olympic ice rink that is 60 m x 30 m (more than 9 times the surface area). If Hawkeye requires 6-7 cameras for accuracy, you would imagine dozens of cameras would need to be set up for figure skating to get that level of accuracy (and that's assuming the assessment of a tennis ball's trajectory involves the same level of complexity as assessing figure skating jump's takeoff — of which the latter is, logically speaking, way more complex).

Agreed that it wouldn't be a regular camera either, but the reality is, if it's one camera you cannot accurately develop a 3D assessment because the camera literally captures data from one plane.

In order to accurately capture a 3D image you need multiple cameras triangulating imagery captured from several angles (six or seven cameras in the case of Hawkeye). It physically isn't possible with a single camera, and I'm not sure how Hanyu's thesis addresses this. I'm not saying he lacks brains or integrity (he obviously has both), but (like any thesis) just because he says it's feasible and reasonably reliable doesn't mean it is. A thesis is simply a proposal that's put forward to be proven as true — that doesn't mean it is, and when considering other forms of replay-tracking technology, using only one camera doesn't seem like it would give an accurate representation.

And if there were a magical, singular camera that currently existed, which could somehow track with the reliability of a multi-camera system like Hawkeye, for something requiring a greater sophistication than measuring a tennis ball's trajectory like a figure skater's blade rotation, it would be monstrously expensive and not at all accessible or feasible for any competition, let alone the ones that can somehow afford it.

 

[CanadianSkaterGuy]

Additionally, looking at the English translation of this thesis (https://drive.google.com/file/d/1svsfSPujsqLQdqziJ-BgezZeFxKl4c0b/view), it seems that jump assessment would also require not just a data-capture camera but motion capture technology strapped to a skater (which of course is incredibly impractical in a sport like figure skating where you probably wouldn't be strapping motion sensors from one skater to the next). Costs-wise he used the PERCEPTION NEURON 2.0 from NOITOM - the base system costs $6000 USD.

 

[DizzyFrenchie]

$1000?! For comparison, a Hawkeye system in tennis costs about $60000 USD, and an alternative PlaySight is about $12500 per court.

Not to mention that's to cover a pro tennis court that is 23.77 m x 8.23 m (10.97 for doubles), compared to an Olympic ice rink that is 60 m x 30 m (more than 9 times the surface area). If Hawkeye requires 6-7 cameras for accuracy, you would imagine dozens of cameras would need to be set up for figure skating to get that level of accuracy (and that's assuming the assessment of a tennis ball's trajectory involves the same level of complexity as assessing figure skating jump's takeoff — of which the latter is, logically speaking, way more complex).

Agreed that it wouldn't be a regular camera either, but the reality is, if it's one camera you cannot accurately develop a 3D assessment because the camera literally captures data from one plane.

In order to accurately capture a 3D image you need multiple cameras triangulating imagery captured from several angles (six or seven cameras in the case of Hawkeye). It physically isn't possible with a single camera, and I'm not sure how Hanyu's thesis addresses this. I'm not saying he lacks brains or integrity (he obviously has both), but (like any thesis) just because he says it's feasible and reasonably reliable doesn't mean it is. A thesis is simply a proposal that's put forward to be proven as true — that doesn't mean it is, and when considering other forms of replay-tracking technology, using only one camera doesn't seem like it would give an accurate representation.

And if there were a magical, singular camera that currently existed, which could somehow track with the reliability of a multi-camera system like Hawkeye, for something requiring a greater sophistication than measuring a tennis ball's trajectory like a figure skater's blade rotation, it would be monstrously expensive and not at all accessible or feasible for any competition, let alone the ones that can somehow afford it.

I asked my husband who told me that the camera wouldn't need two lenses and would rather be a good precise camera in the 100-200$ range? With a lot of software analysis.
If I understand correctly, motion capture is for initial sampling. There are a lot of biomechanic softwares available to implement such a program more easily.

 

[CanadianSkaterGuy]

Here is the original Japanese thesis. https://waseda.repo.nii.ac.jp/?acti...tem_id=64787&item_no=1&page_id=13&block_id=21

So... I'm wondering where the heck the actual data? I don't see any tables, graphs, or figures (beyond where the motion sensors were attached and the software showing where pressure is). He explains the mechanics of the jumps but there aren't any actual numbers. He omits anything quantitative: e.g. he says things like the "measurement of rotation was excellent" - what were the actual measurements? excellent compared to what? He describes the loop jump as having a red indicator on for a long time. For how long was it on? What was the sample size of loop jumps tested? Etc. There isn't any data surrounding the motion capture when analyzing the axel/flip/loop jumps, and in something that is supposed to be regarded as quantifiable, it's astonishing that there isn't a single set of stats coming out of the motion capture analysis. Also, why were the salchow and toe loop omitted (I understand not a lutz because the flip was there), but for jump rotation, especially combination jumps and in the context of pre-rotation at least the toe loop is significant.

He asserts that it's possible to capture data about the movements up to the moment of takeoff to a high degree of precision... but there are literally no numbers in his thesis that back it up. I think as a starting point, it's an interesting thought to use motion capture, but there is nothing remotely scientific about this thesis looking at it. He also doesn't elaborate on things like how a camera would differentiate a skater's take off, jumps that are executed in various parts of the rink or the need to triangulate a 3D image using multiple cameras. Or the feasibility of having to apply motion capture sensors to a figure skater for the system to even work.

At the end he does mention some limitations, like how in motion capture there is no data for things like figure skating jumps (only walking, talking on the phone) and acknowledges that skeletal frame tracking is limited to only 2D for the time being, and 3D is a possibility in the future. But these very acknowledgements should tell everyone that it's very much a theory (the 3D technology hasn't even been developed yet) and such motion capture, is a way's away if it's even possible/feasible/accessible.

It is definitely not "realistic" and moreover he hasn't even presented any concrete quantitative results/numbers/data to support the motion capture analysis, which was the basis of his method.

 

[CanadianSkaterGuy]

I asked my husband who told me that the camera wouldn't need two lenses and would rather be a good precise camera in the 100-200$ range? With a lot of software analysis.
If I understand correctly, motion capture is for initial sampling. There are a lot of biomechanic softwares available to implement such a program more easily.

I'm doubtful that a $200 camera would be able to capture. However, cameras are definitely getting more sophisticated, so by the time an actual 3D modeling software that could support Hanyu's hypothesis comes out, perhaps cameras will be more affordable. I still think, however, that it would be inaccurate with the size of a rink as well as having only one camera. Right now, a single camera (certainly not a $200) is sophisticated enough to produce a 3D model, let alone calibrated to every size and type of skater/skate blade.

Cameras are fun for stats and viewer play by plays, but if I were a skater I would refuse to allow such technology to assess my jumps or determine the outcome of a competition. Even video replay software like Hawkeye that is calibrated to within 2 mm accuracy are prone to errors. And figure skating jump assessment software/hardware would have to be far more sophisticated in order to be accurate enough to legitimize its use in assessing jump rotation. Every technology has its flaws or work arounds.

 

[CanadianSkaterGuy]

Regarding the use of just one camera instead of many, while obviously not the same, anyone ever seen the red-light-green light challenge from Squid Game? (SPOILER )

Where if people moved they were out of the game? But if you moved while behind somebody else who was motionless, you were okay because the doll could only detect people moving directly in front of her? Yes, it's fictitious motion-sensor tech written for the purpose of a show, but even that acknowledges the limitations of having technology that detects motion only from one plane (versus having multiple motion sensors around the room to detect movement from all angles). The doll might be super accurate in detecting motion when the players were moving in front of it or at a certain angle (with some leeway of course, it didn't take out people who were breathing or blinking or shuddering), but if there were multiple cameras, people wouldn't have been able to take advantage of someone moving when behind somebody in front of them, and thus they would actually have to maintain stillness from all angles.

That would be the equivalent of having only one camera that detects motion/prerotation... if skaters knew where this singular camera was placed on the ice rink, they would also know how to choreograph/execute their jumps to give them the best chance of gaming the camera (the same way skaters strategically place their jumps far from the judging panel, or obscure each other in twizzles like S/K to hide technical flaws or get the benefit of the doubt on tech calls). You would still get the same issues with just one camera. That's why Hawkeye employs multiple cameras triangulating a 3D image of the ball because one camera's angle might not be able to adequately assess where the ball lands depending on its placement so it's an aggregate of multiple angles.

 

[Jadeice]

Off the top of my head, examples of current skaters whose lutzes (and flips) have minimal (at most 90 degrees) prerotation:

Men: Boyang Jin, Han Yan, He Zhang, Yuzuru Hanyu, Lucas Tsuyoshi Honda, Keegan Messing, Jason Brown, Jimmy Ma, William Annis, Ilia Malinin, Brendan Kerry, Daniel Samohin, Kevin Aymoz, Michal Březina, Matyáš Bělohradský, Deniss Vasiļjevs, Paul Fentz, Slavik Hayrapetyan, Kornel Witkowski, Nikolaj Majorov, Mikhail Kolyada, Ilya Yablokov

Women: Tomoe Kawabata, Mako Yamashita, Karen Chen, Alexia Paganini, Emmi Peltonen, Elizaveta Tuktamysheva, Veronika Zhilina

LOL. Oh please, don't make me laugh... Jason doesn't even have a Lutz. Also "interesting" you leave out Nathan Chen who's known for his excellent Lutz take-off, but I guess the recent OGM wouldn't be on your mind, not current enough I guess...

 

[jenaj]

Using a super-sophisticated camera is a bad idea. Would coaches and training centers have to have them, too, to assess whether their skaters were pre-rotating? There is really no good solution to the issue of pre-rotations. People can't even agree on how much pre-rotation is too much. And the "acceptable" amount is different for different jumps. This video discusses some of the problems. https://www.youtube.com/watch?v=HkcAmGCkjtA

 

[CanadianSkaterGuy]

LOL. Oh please, don't make me laugh... Jason doesn't even have a Lutz. Also "interesting" you leave out Nathan Chen who's known for his excellent Lutz take-off, but I guess the recent OGM wouldn't be on your mind, not current enough I guess...

Very true that Chen should have been on the list.

Jason doesn't have the best outside edge on his lutz (although it's good enough to never get called, but there is minimal pre-rotation on his lutz pick. https://www.youtube.com/watch?v=W44QDl16pK0#t=5m10s

 

[yume]

LOL. Oh please, don't make me laugh... Jason doesn't even have a Lutz. Also "interesting" you leave out Nathan Chen who's known for his excellent Lutz take-off, but I guess the recent OGM wouldn't be on your mind, not current enough I guess...

They were strictly talking about pre-rotation.

 

[AxelLover]

LOL. Oh please, don't make me laugh... Jason doesn't even have a Lutz. Also "interesting" you leave out Nathan Chen who's known for his excellent Lutz take-off, but I guess the recent OGM wouldn't be on your mind, not current enough I guess...

Jason's Lutz edge is not very good, but I wasn't talking about the edge quality here, only about prerotation.
Nathan Chen definitely doesn't prerotate much, but he does prerotate a bit more than 90 degrees and that's why I didn't include him in the list.

By the way, Nathan is actually one of my all time favorite skaters, but I'm not going to pretend he doesn't prerotate his lutz a bit more than the skaters I mentioned in the list.

 

[DizzyFrenchie]

Here is the original Japanese thesis. https://waseda.repo.nii.ac.jp/?acti...tem_id=64787&item_no=1&page_id=13&block_id=21

So... I'm wondering where the heck the actual data? I don't see any tables, graphs, or figures (beyond where the motion sensors were attached and the software showing where pressure is). He explains the mechanics of the jumps but there aren't any actual numbers. He omits anything quantitative: e.g. he says things like the "measurement of rotation was excellent" - what were the actual measurements? excellent compared to what? He describes the loop jump as having a red indicator on for a long time. For how long was it on? What was the sample size of loop jumps tested? Etc. There isn't any data surrounding the motion capture when analyzing the axel/flip/loop jumps, and in something that is supposed to be regarded as quantifiable, it's astonishing that there isn't a single set of stats coming out of the motion capture analysis. Also, why were the salchow and toe loop omitted (I understand not a lutz because the flip was there), but for jump rotation, especially combination jumps and in the context of pre-rotation at least the toe loop is significant.

He asserts that it's possible to capture data about the movements up to the moment of takeoff to a high degree of precision... but there are literally no numbers in his thesis that back it up. I think as a starting point, it's an interesting thought to use motion capture, but there is nothing remotely scientific about this thesis looking at it. He also doesn't elaborate on things like how a camera would differentiate a skater's take off, jumps that are executed in various parts of the rink or the need to triangulate a 3D image using multiple cameras. Or the feasibility of having to apply motion capture sensors to a figure skater for the system to even work.

At the end he does mention some limitations, like how in motion capture there is no data for things like figure skating jumps (only walking, talking on the phone) and acknowledges that skeletal frame tracking is limited to only 2D for the time being, and 3D is a possibility in the future. But these very acknowledgements should tell everyone that it's very much a theory (the 3D technology hasn't even been developed yet) and such motion capture, is a way's away if it's even possible/feasible/accessible.

It is definitely not "realistic" and moreover he hasn't even presented any concrete quantitative results/numbers/data to support the motion capture analysis, which was the basis of his method.

No, this is not the thesis itself, it's an article to introduce his thesis. I have in memory that his thesis is 300 pages but I'm really not certain of this number. The professor who supervised it believes there's enough to make several theses.