- Joined
- Jan 1, 2013
There are very stringent regulations about what can be advertised during the Olympics. And I doubt Samsung would care unless the KSU were trying to force Kim to come back.
Use of technology in figure skating tech calls
- Thread starter jinhamasaki
- Start date
There are very stringent regulations about what can be advertised during the Olympics. And I doubt Samsung would care unless the KSU were trying to force Kim to come back.
- Joined
- Dec 29, 2013
You missed the point I failed to properly make. I was saying that maybe Samsung will develop some technology that can be used to benefit all of the events in the olympics, not just figure skating obviously, by capitalizing on some sort of slow motion 3d technology. The advertising wouldn't be advertising like radio ads or tv commercials or even giant banners but instead by simply taking advantage of product placement. In other words ....slow motion technology (#BROUGHTtoYOUbySAMSUNG )that can be used in the Olympic Games broadcasts for all events....including but not designed for figure skating. Things like this can creep into people's brains and increase the sales of Samsung's product lines in multiple areas. It's like I said though.....just an out loud thought. Just a random scenario that could unfold and by pure chance benefit figure skating and it's fans.
Last edited:
The ISU would never go for any technology that would undermine the ISU's control over the sport. And showing what the technical team is judging on to the audience could definitely cause an uproar, because there have been occasions when the rulings weren't completely in agreement with reality, and that will be the case in the future, no doubt. No way will the ISU want that dirty laundry aired in public!!
It's the same thing as having the judges' scores secret---that won't ever change, either, for the same reason.
It's the same thing as having the judges' scores secret---that won't ever change, either, for the same reason.
I'm not so sure. As long as a speed skater is ISU president I think that this would definitly have a certain appeal to him. And there's still the GOE's and PCS.
All I know is that the ISU is super behind on the times. They need to be willing to conduct more studies and try new things. They have nothing to lose.
Craig Buntin's thing he did at Skate Canada last year was pretty nifty. I remember them using it to show that Ashley Wagner came in way too slow to do her triple triple, in comparison to some of the other ladies at that event.
Craig Buntin's thing he did at Skate Canada last year was pretty nifty. I remember them using it to show that Ashley Wagner came in way too slow to do her triple triple, in comparison to some of the other ladies at that event.
I like your discussion guys, maybe I will bring it up at NT 2015, lets see what the skaters have to say about it 
@ topic
Hawk Eye definitely changed a lot in tennis no doubt. If its for the better or worse, I don't know.
If you speak to players, they are in general very positive about it, yet it also interrupts their game flow a lot, especially during serve a nasty thing and even if the challenge was successful, there is a huge difference between replay the point and actually granting / deducting said point. You also have the risk that a player takes a challenge, to disturb the opponent, something that could become a handy tool in FS as well. Lets say a skater knows the next one is a heavy competitor, lets take the challenge and let the other skater wait. Its a typical procedure at sports that have such challenge systems in place
I also wonder if its technically possible and I mean, for every event and not just the Olympics or Worlds. Who will pay for all of that? The ISU or the actual skating rinks themselves? If latter, then I seriously doubt that such an implementation is realistic. Skating events, at least those I go to depend on volunteers and wont be able to afford such an investment.
Still,
I agree that FS could need a bit of fresh air in many aspects. While I can only speak from a writers and volunteers perspective, tennis and soccer are light-years ahead of FS and this doesn't need to be like that. If the ISU would just accept the offers of those who want to help, many things could be improved.
@ topic
Hawk Eye definitely changed a lot in tennis no doubt. If its for the better or worse, I don't know.
If you speak to players, they are in general very positive about it, yet it also interrupts their game flow a lot, especially during serve a nasty thing and even if the challenge was successful, there is a huge difference between replay the point and actually granting / deducting said point. You also have the risk that a player takes a challenge, to disturb the opponent, something that could become a handy tool in FS as well. Lets say a skater knows the next one is a heavy competitor, lets take the challenge and let the other skater wait. Its a typical procedure at sports that have such challenge systems in place
I also wonder if its technically possible and I mean, for every event and not just the Olympics or Worlds. Who will pay for all of that? The ISU or the actual skating rinks themselves? If latter, then I seriously doubt that such an implementation is realistic. Skating events, at least those I go to depend on volunteers and wont be able to afford such an investment.
Still,
I agree that FS could need a bit of fresh air in many aspects. While I can only speak from a writers and volunteers perspective, tennis and soccer are light-years ahead of FS and this doesn't need to be like that. If the ISU would just accept the offers of those who want to help, many things could be improved.
Last edited:
- Joined
- Jan 13, 2014
I'd love to see this. Any idea where I can find it? Also I don't see the need for judges to review EVERY jump, they don't look at every single serve in tennis either. Kim's 3Lz for example is pretty easy to see and score.
- Joined
- Feb 27, 2012
I have copied and pasted below my post from the U.S. ladies thread in the 2014-15 archives:
The TSN link in the quoted post no longer works. I do not know whether a YouTube version of the TSN piece exists.
Previous bits of discussion of VeriSkate are scattered among other GS threads. PJ Kwong talked to Craig Buntin about it in a podcast. Skate Canada and IceNetwork have written about it.
Non-exhaustive collection of links related to VeriSkate:
http://goldenskate.com/forum/showth...estions-Thread&p=995150&viewfull=1#post995150 (posts #1017 and #1019)
http://www.goldenskate.com/forum/sh...Web-Coverage&p=1002858&viewfull=1#post1002858 (post #14)
http://www.goldenskate.com/forum/showthread.php?p=1004743#post1004743 (post #403)
http://www.goldenskate.com/forum/sh...ens-SP-10-31&p=1006036&viewfull=1#post1006036 (post #450)
http://www.goldenskate.com/forum/sh...ens-SP-10-31&p=1006173&viewfull=1#post1006173 (post #454)
https://twitter.com/veriskate/with_replies
https://www.facebook.com/veriskate
http://www.goldenskate.com/forum/sh...Web-Coverage&p=1002858&viewfull=1#post1002858 (post #14)
http://www.goldenskate.com/forum/showthread.php?p=1004743#post1004743 (post #403)
http://www.goldenskate.com/forum/sh...ens-SP-10-31&p=1006036&viewfull=1#post1006036 (post #450)
http://www.goldenskate.com/forum/sh...ens-SP-10-31&p=1006173&viewfull=1#post1006173 (post #454)
https://twitter.com/veriskate/with_replies
https://www.facebook.com/veriskate
- Joined
- Feb 27, 2012
For those of you saying the cost to implement any system would be too great please consider that as of December 31, 2014 the ISU has over $270,000,000 in assests and $230,000,000 in equity...yes, that's million. Cost isn't the real issue.
- Joined
- Jun 21, 2003
Here are the details, if anyone wants to check them out. Most of the ISU's money is in long term investments held by Swiss banks. (A Swiss Franc is about equal in value to a U.S. dollar.)
If you scroll down there are a lot of other interest facts. Fo instance, the ISU made 9,000,000 off the Olympics and paid out 17,000,000 in expenses for ISU events, including prize money.
http://static.isu.org/media/211012/isu-bdo-report-financial-statement-2014.pdf
As far as I can tell from the little I can find online, as of last year, the VeriSkate software still seems to be in development, not ready for widespread use.
I'm not clear how it works. It's software, but where does it get the data from? That may be the biggest determinant of affordability and accuracy.
In this article, Buntin talks about how it could be used for training, and the author (Beverley Smith) speculates that it could also be useful for media and judging officials.
The ISU has no direct control over what skaters and coaches do in training or what the media do for broadcasts. They could make information available to them.
Someday, maybe next month, maybe in 10-20 years, this or other technology may be readily enough available that it could be used in official scoring. Could there be a purely objective part of the score that measures details like speed and height of jumps, correct edges for jump takeoffs or amount of in-air rotation, speed of spin rotations and accuracy of centering, average speed across the ice throughout the program, depth of edge or cleanness of turns, etc. (some but not all of which has been claimed for VeriSkate), and factor those data into the scores for individual elements or for Skating Skills or things like the pattern and ice coverage aspects of choreography?
There would still need to be someone to identify the elements and basically do everything the technical panel already does -- except the edge and rotation calls if the technology could do that more accurately than the human eye, but including calling level features on non-jump elements unless there were a decision not to give extra points for those kinds of added difficulty after all.
To the extent that everything that is not quantitative about a program is still considered to contribute to evaluations of good skating, judges would still be needed to do what they do now. E.g., judges might still consider the neatness and control, attractiveness of positions, etc., of a jump or spin, or their relation to the music, when awarding GOE, but the height and distance aspects could be determined by the technology instead.
I'm not clear how it works. It's software, but where does it get the data from? That may be the biggest determinant of affordability and accuracy.
In this article, Buntin talks about how it could be used for training, and the author (Beverley Smith) speculates that it could also be useful for media and judging officials.
The ISU has no direct control over what skaters and coaches do in training or what the media do for broadcasts. They could make information available to them.
Someday, maybe next month, maybe in 10-20 years, this or other technology may be readily enough available that it could be used in official scoring. Could there be a purely objective part of the score that measures details like speed and height of jumps, correct edges for jump takeoffs or amount of in-air rotation, speed of spin rotations and accuracy of centering, average speed across the ice throughout the program, depth of edge or cleanness of turns, etc. (some but not all of which has been claimed for VeriSkate), and factor those data into the scores for individual elements or for Skating Skills or things like the pattern and ice coverage aspects of choreography?
There would still need to be someone to identify the elements and basically do everything the technical panel already does -- except the edge and rotation calls if the technology could do that more accurately than the human eye, but including calling level features on non-jump elements unless there were a decision not to give extra points for those kinds of added difficulty after all.
To the extent that everything that is not quantitative about a program is still considered to contribute to evaluations of good skating, judges would still be needed to do what they do now. E.g., judges might still consider the neatness and control, attractiveness of positions, etc., of a jump or spin, or their relation to the music, when awarding GOE, but the height and distance aspects could be determined by the technology instead.
- Joined
- Feb 27, 2012
Last year, the website www.veriskate.com was alive and well and full of promise, but now it is dead. And the VeriSkate social media accounts have been silent for many, many months. Thus my fear that development of VeriSkate now is dormant. But I would love to be proven wrong.
Craig Buntin's company for hockey analytics is http://sportlogiq.com/ -- based on a proprietary computer vision system (which I think was the basis also for VeriSkate).
"Backed by over seven years of university research, our technology generates advanced hockey analytics using standard, single-camera game footage."
Of its small number of employees, many are computer vision researchers.
There's several issues here:
1. Using cameras to evaluate skating isn't the same as for a tennis ball or even a baseball strike zone. For whether a tennis ball was in or out, it's easy; the software just needs to figure out when the ball hit the ground and align it with the line marking the edge of the court as a reference. For a strike zone, it's a bit more difficult because there's no direct reference point (i.e. the ball is in midair rather than touching the ground), but it's fairly straightforward to interpolate between the pitcher's and catcher's positions. For skating, the challenge is much more difficult. Under-rotations are based on the skater's path of travel over the ice, so it's not just a single measurement. Instead, you'd have to figure out the path of the skater, and then use that as a reference for the skater's landing. You'd also have to figure out not just the position of the skater's skate, but also its orientation. Edge calls also depend on determining the skate's orientation. So it's not just a single frame that you can show the viewer, the way it is for seeing if a tennis ball was in or out.
2. Furthermore, a lot of the rules are not as specific as you'd expect. For example, how do you define a skater's path of travel in the air during a jump? Ideally we may want something like the path of the skater's center of mass, but good luck figuring that out when the skater not only has an irregular shape (i.e. not a ball, cube, etc.) but also an irregular mass distribution, plus clothing on top of that. So instead, you may use something like the tracing on the ice. But what is the line of motion that you would use? Taking the (counter-clockwise) flip as an example, it's pretty clear that the landing spot would be one point of the line of travel. But do you count the starting point as the right toe-pick point? The end of the left foot glide? Maybe an average of the two? It's actually not so clear. Additionally, when measuring the orientation of the skate on landing, at what point do you make the measurement? Is it right when the toe-pick touches the ice? Is it when the (non-toe-pick) blade first touches the ice? If you say toe-pick, then it encourages skaters to land more flat-footed to avoid under-rotation calls (since there's some rotation between when the toe-pick touches and when the blade touches), which would lead to more wear-and-tear on the body and less control coming out of the landing; as already mentioned, it would disfavor some skaters like Mirai who tends to point their toes a lot on landing.
3. A corollary to this is, what defines a good jump? For example, I'm guessing that what a judge looks at to determine under-rotation is not actually the skate's orientation at the moment of landing (which is the basis for its definition), but at how that landing affects the skater's body, in terms of speed going out of the jump, if the free leg swings wildly, the path of the landing leg on the ice, etc. Focus on these more technical aspects (i.e. under-rotation) may lead to less emphasis on other aspects of a jump that makes the sport aesthetically attractive (such as carriage, flow, etc.).
4. How much would the system be adopted? I'm not talking about whether or not the ISU would use this kind of thing to replace judges at the Olympics. I'm talking more about at what level of competition would people be expected to use this type of system. Do you use it on basic skills tests? Juvenile competitions? Etc. Although I know both IJS and 6.0 are still used, it's a bit of a bureaucratic mess to manage different scoring systems -- one that's done via software, and one that's done via the human eyeball.
Having said that, my thoughts are (these are similar to a similar thread from just after the 2014 Olympics):
There's a lot of data in videos. The problem is generally not whether or not there's enough data, but whether or not we can generate good enough algorithms to process that data and extract the information that we want. Conceptually, however, it's not that difficult to do. For example, the path that each skater took in their Olympic programs is already worked out; I remember seeing Youtube videos at the time, so either those guys used an algorithm or they drew it by hand (I don't know which). Even if they drew it by hand, if you have a video that captures the entire rink, it's actually not that difficult to calculate the skater's position in any video frame relatively accurately (i.e. to say within 10-20 cm). I demonstrated how this could be done in an earlier thread; I would guess the Veriskate guys basically do the same thing. After calculating the skater's position for multiple video frames, it's fairly straightforward to work out the skater's velocity. For a jump, it's just a matter of taking the skater's position at the beginning and at the end of the jump, and dividing the difference between them with the time elapsed (calculated via number of video frames elapsed) to figure out the (horizontal) velocity, air time (and thus jump height), etc.
Next, figuring out under-rotations. Regardless of whatever rule you adopt, you need to have a way to accurately measure the skate's position on landing. This can be done via slow-motion video of the landing skate. However, we have to be cognizant of visual distortions (basically, it's zoomed in from far away), and the human eyeball is pretty bad at accounting for this, but algorithms can compensate for it. Basically you just need to determine the skate's orientation with respect to the camera, and if the slow-motion video is clear enough (i.e. not blurry), this should be doable. You can also check for correct edges in this way.
In short, I don't think it's all that difficult to actually measure, relatively accurately, various parameters of interest of a skater's jump such as the height (calculated from air time), distance traveled, landing rotation (i.e. amount of under-rotation), and take-off edge, from video. In fact, I think the only videos that you really need are:
1. A "wide" view of the entire rink while the skater is skating. This will give the skater's position in the rink at all times. If the video has good enough resolution, it can also be used to determine the position of each skate (for things like jump take-off and landing locations of each skate).
2. A "close-up" slow-motion view of the skater's skates at the times of interest, i.e. take-off and landing of a jump. This will give the 3-D orientation of the skate relative to the camera. If the camera's position within the rink is known (such as if it's within view of the wide-view camera), then the orientation of the skate relative to the rink, and hence of the jump's path of travel, can be calculated. From this, how much under-rotation occurred on landing can be calculated, as well as the take-off edge.
That's it. The rest is just developing the algorithms to make this work. The question then is what incentive is there for the ISU to bother adopting this type of stuff.
In fact, my claim is that this can be done largely from the videos that are already out there, and so it's possible to give a good analysis just from available Youtube videos and such. My evidence is to actually do it from past videos, specifically, videos from the Olympics, mostly because they're relatively plentiful (it's hard to find fancam videos of joe schmo competitions, and even of many of the Olympic performances). A rundown of the process is:
1. The entire rink is visible in the "wide" camera. The rink's dimensions are known (at least, I hope it is, I assume the Olympic rink was an "Olympic-sized" 60 m X 30 m). From perspective geometry, it's possible to calculate the X-Y position of the skater's skates relative to the rink to within the pixel size of the video. From the Olympic fancams I found on Youtube, this error was generally around 20 cm or so I think. (Don't have the spreadsheet in front of me right now, might be fuzzy on some of the details.) This is just from ordinary handheld cameras, a more specialized camera would give better precision. From this, the skater's X-Y coordinates at jump take-off and landing can be calculated. I posted a demonstration of this a while back, even giving a grid of points across the rink calculated from the measurements.
2. Figure out the position of the "close-up" camera. It turns out that the camera used for the network close-ups was in some of the fancam videos that I found. Therefore, I can determine its location relative to the rink pretty accurately (again, within 20 cm or so).
3. Figure out the orientation of the skate relative to the "close-up" camera on landing a jump. This is probably the most difficult part. However, I can get a good approximation of this by comparing images of the skate as it's in the air prior to landing (as it's rotating). From the orientation of the skate relative to the "close-up" camera, combined with the skater's X-Y position at the beginning and the end of a jump, how much under-rotation occurred can be calculated. It's more difficult to determine the inside/outside edge on take-off but that can be done as well once the orientation is determined.
During the post-Olympic discussion I did this manually for some of the most-discussed (i.e. controversial) jumps, but never got around to doing it for all of them nor writing up the procedure so it never got posted (yet). IIRC (spoiler alert), though I don't have the spreadsheet in front of me so it's just from memory right now, the punchline was that Yuna's LP second 3Lz was around 120 degrees under-rotated, Mao's LP 3A was over 135 degrees under-rotated, and Adelina's LP 3T was around 160 degrees under-rotated. None of those were marked with a UR call.
Of course, officially, anything over 90 degrees is supposed to be marked with a UR "<", but this hints at an even deeper rule that was often-ignored in the judge debates: The rules say something like if it's questionable, it should be resolved in favor of the skater, and the judges (IIRC) are evaluating the jumps in real-time so they're not using slow-mo frame-by-frame video for UR calls the way the Internet does. For example, if the skate was 100 degrees under-rotated, unless the judge can differentiate between this and 90 degrees under-rotated, it wouldn't get called. (Keep in mind that if a skater is rotating 5 times per second in the air, he rotates 10 degrees every 5.56 ms.) This makes all the Internet caterwauling about why some jumps weren't called UR kind of pointless if they were relying on super-slow-mo and hand-drawn angles (remember, the human eyeball is very bad at estimating angles from an image when it's been distorted due to zoom). The question then is *how* much leeway beyond 90 degrees the judges give in practice, and whether or not this leeway tends to be lower for some skaters and higher for others. My original plan then was to calculate the amount of under-rotation for the jumps that *were* called UR and compare those to ones that weren't, to figure out this leeway (noting that it may vary by jump), but it was very time-consuming and eventually the threads were archived anyway.
But TL;DR -- it's doable, and the question is really whether or not the ISU would be interested in investing in this kind of thing.
1. Using cameras to evaluate skating isn't the same as for a tennis ball or even a baseball strike zone. For whether a tennis ball was in or out, it's easy; the software just needs to figure out when the ball hit the ground and align it with the line marking the edge of the court as a reference. For a strike zone, it's a bit more difficult because there's no direct reference point (i.e. the ball is in midair rather than touching the ground), but it's fairly straightforward to interpolate between the pitcher's and catcher's positions. For skating, the challenge is much more difficult. Under-rotations are based on the skater's path of travel over the ice, so it's not just a single measurement. Instead, you'd have to figure out the path of the skater, and then use that as a reference for the skater's landing. You'd also have to figure out not just the position of the skater's skate, but also its orientation. Edge calls also depend on determining the skate's orientation. So it's not just a single frame that you can show the viewer, the way it is for seeing if a tennis ball was in or out.
2. Furthermore, a lot of the rules are not as specific as you'd expect. For example, how do you define a skater's path of travel in the air during a jump? Ideally we may want something like the path of the skater's center of mass, but good luck figuring that out when the skater not only has an irregular shape (i.e. not a ball, cube, etc.) but also an irregular mass distribution, plus clothing on top of that. So instead, you may use something like the tracing on the ice. But what is the line of motion that you would use? Taking the (counter-clockwise) flip as an example, it's pretty clear that the landing spot would be one point of the line of travel. But do you count the starting point as the right toe-pick point? The end of the left foot glide? Maybe an average of the two? It's actually not so clear. Additionally, when measuring the orientation of the skate on landing, at what point do you make the measurement? Is it right when the toe-pick touches the ice? Is it when the (non-toe-pick) blade first touches the ice? If you say toe-pick, then it encourages skaters to land more flat-footed to avoid under-rotation calls (since there's some rotation between when the toe-pick touches and when the blade touches), which would lead to more wear-and-tear on the body and less control coming out of the landing; as already mentioned, it would disfavor some skaters like Mirai who tends to point their toes a lot on landing.
3. A corollary to this is, what defines a good jump? For example, I'm guessing that what a judge looks at to determine under-rotation is not actually the skate's orientation at the moment of landing (which is the basis for its definition), but at how that landing affects the skater's body, in terms of speed going out of the jump, if the free leg swings wildly, the path of the landing leg on the ice, etc. Focus on these more technical aspects (i.e. under-rotation) may lead to less emphasis on other aspects of a jump that makes the sport aesthetically attractive (such as carriage, flow, etc.).
4. How much would the system be adopted? I'm not talking about whether or not the ISU would use this kind of thing to replace judges at the Olympics. I'm talking more about at what level of competition would people be expected to use this type of system. Do you use it on basic skills tests? Juvenile competitions? Etc. Although I know both IJS and 6.0 are still used, it's a bit of a bureaucratic mess to manage different scoring systems -- one that's done via software, and one that's done via the human eyeball.
Having said that, my thoughts are (these are similar to a similar thread from just after the 2014 Olympics):
There's a lot of data in videos. The problem is generally not whether or not there's enough data, but whether or not we can generate good enough algorithms to process that data and extract the information that we want. Conceptually, however, it's not that difficult to do. For example, the path that each skater took in their Olympic programs is already worked out; I remember seeing Youtube videos at the time, so either those guys used an algorithm or they drew it by hand (I don't know which). Even if they drew it by hand, if you have a video that captures the entire rink, it's actually not that difficult to calculate the skater's position in any video frame relatively accurately (i.e. to say within 10-20 cm). I demonstrated how this could be done in an earlier thread; I would guess the Veriskate guys basically do the same thing. After calculating the skater's position for multiple video frames, it's fairly straightforward to work out the skater's velocity. For a jump, it's just a matter of taking the skater's position at the beginning and at the end of the jump, and dividing the difference between them with the time elapsed (calculated via number of video frames elapsed) to figure out the (horizontal) velocity, air time (and thus jump height), etc.
Next, figuring out under-rotations. Regardless of whatever rule you adopt, you need to have a way to accurately measure the skate's position on landing. This can be done via slow-motion video of the landing skate. However, we have to be cognizant of visual distortions (basically, it's zoomed in from far away), and the human eyeball is pretty bad at accounting for this, but algorithms can compensate for it. Basically you just need to determine the skate's orientation with respect to the camera, and if the slow-motion video is clear enough (i.e. not blurry), this should be doable. You can also check for correct edges in this way.
In short, I don't think it's all that difficult to actually measure, relatively accurately, various parameters of interest of a skater's jump such as the height (calculated from air time), distance traveled, landing rotation (i.e. amount of under-rotation), and take-off edge, from video. In fact, I think the only videos that you really need are:
1. A "wide" view of the entire rink while the skater is skating. This will give the skater's position in the rink at all times. If the video has good enough resolution, it can also be used to determine the position of each skate (for things like jump take-off and landing locations of each skate).
2. A "close-up" slow-motion view of the skater's skates at the times of interest, i.e. take-off and landing of a jump. This will give the 3-D orientation of the skate relative to the camera. If the camera's position within the rink is known (such as if it's within view of the wide-view camera), then the orientation of the skate relative to the rink, and hence of the jump's path of travel, can be calculated. From this, how much under-rotation occurred on landing can be calculated, as well as the take-off edge.
That's it. The rest is just developing the algorithms to make this work. The question then is what incentive is there for the ISU to bother adopting this type of stuff.
In fact, my claim is that this can be done largely from the videos that are already out there, and so it's possible to give a good analysis just from available Youtube videos and such. My evidence is to actually do it from past videos, specifically, videos from the Olympics, mostly because they're relatively plentiful (it's hard to find fancam videos of joe schmo competitions, and even of many of the Olympic performances). A rundown of the process is:
1. The entire rink is visible in the "wide" camera. The rink's dimensions are known (at least, I hope it is, I assume the Olympic rink was an "Olympic-sized" 60 m X 30 m). From perspective geometry, it's possible to calculate the X-Y position of the skater's skates relative to the rink to within the pixel size of the video. From the Olympic fancams I found on Youtube, this error was generally around 20 cm or so I think. (Don't have the spreadsheet in front of me right now, might be fuzzy on some of the details.) This is just from ordinary handheld cameras, a more specialized camera would give better precision. From this, the skater's X-Y coordinates at jump take-off and landing can be calculated. I posted a demonstration of this a while back, even giving a grid of points across the rink calculated from the measurements.
2. Figure out the position of the "close-up" camera. It turns out that the camera used for the network close-ups was in some of the fancam videos that I found. Therefore, I can determine its location relative to the rink pretty accurately (again, within 20 cm or so).
3. Figure out the orientation of the skate relative to the "close-up" camera on landing a jump. This is probably the most difficult part. However, I can get a good approximation of this by comparing images of the skate as it's in the air prior to landing (as it's rotating). From the orientation of the skate relative to the "close-up" camera, combined with the skater's X-Y position at the beginning and the end of a jump, how much under-rotation occurred can be calculated. It's more difficult to determine the inside/outside edge on take-off but that can be done as well once the orientation is determined.
During the post-Olympic discussion I did this manually for some of the most-discussed (i.e. controversial) jumps, but never got around to doing it for all of them nor writing up the procedure so it never got posted (yet). IIRC (spoiler alert), though I don't have the spreadsheet in front of me so it's just from memory right now, the punchline was that Yuna's LP second 3Lz was around 120 degrees under-rotated, Mao's LP 3A was over 135 degrees under-rotated, and Adelina's LP 3T was around 160 degrees under-rotated. None of those were marked with a UR call.
Of course, officially, anything over 90 degrees is supposed to be marked with a UR "<", but this hints at an even deeper rule that was often-ignored in the judge debates: The rules say something like if it's questionable, it should be resolved in favor of the skater, and the judges (IIRC) are evaluating the jumps in real-time so they're not using slow-mo frame-by-frame video for UR calls the way the Internet does. For example, if the skate was 100 degrees under-rotated, unless the judge can differentiate between this and 90 degrees under-rotated, it wouldn't get called. (Keep in mind that if a skater is rotating 5 times per second in the air, he rotates 10 degrees every 5.56 ms.) This makes all the Internet caterwauling about why some jumps weren't called UR kind of pointless if they were relying on super-slow-mo and hand-drawn angles (remember, the human eyeball is very bad at estimating angles from an image when it's been distorted due to zoom). The question then is *how* much leeway beyond 90 degrees the judges give in practice, and whether or not this leeway tends to be lower for some skaters and higher for others. My original plan then was to calculate the amount of under-rotation for the jumps that *were* called UR and compare those to ones that weren't, to figure out this leeway (noting that it may vary by jump), but it was very time-consuming and eventually the threads were archived anyway.
But TL;DR -- it's doable, and the question is really whether or not the ISU would be interested in investing in this kind of thing.
- Joined
- Jan 1, 2013
Seriously, how many people here have ever sat on a tech panel? I strongly suggest you join your local club and volunteer as Data Entry. It would seem that what people might learn from that would SHOCK a lot of you.
No. It is where the blade lands. Pure and simple. Mirai gets pinged not because of her "toe point" but because when her toe touches down, it's facing the wrong way. End of story. The body is not even considered, and nor should it be, because a lot of skaters develop methods of "cheating" jumps to make them look rotated to the naked eye.
The technical panel at large events has its own video feed. When they are reviewing elements, they can slo-mo, super-slo-mo, reverse-slo-mo, full speed, FRAME-BY-FRAME analyse elements - including jumps.
Technical panels aren't there because they're evil people who want to tear skaters apart. They will often given benefit of the doubt - but only if it's really close.
It really ticks me off that so many posters here are prepared to tear the technical panels to shreds as though they're a bunch of evil, corrupt, boneheaded, uneducated morons who are just out to get favourite skaters when the posters themselves don't even seem to realise how the system works.
No. It is where the blade lands. Pure and simple. Mirai gets pinged not because of her "toe point" but because when her toe touches down, it's facing the wrong way. End of story. The body is not even considered, and nor should it be, because a lot of skaters develop methods of "cheating" jumps to make them look rotated to the naked eye.
The technical panel at large events has its own video feed. When they are reviewing elements, they can slo-mo, super-slo-mo, reverse-slo-mo, full speed, FRAME-BY-FRAME analyse elements - including jumps.
Technical panels aren't there because they're evil people who want to tear skaters apart. They will often given benefit of the doubt - but only if it's really close.
It really ticks me off that so many posters here are prepared to tear the technical panels to shreds as though they're a bunch of evil, corrupt, boneheaded, uneducated morons who are just out to get favourite skaters when the posters themselves don't even seem to realise how the system works.
- Joined
- Dec 29, 2013
I'd be interested in reading the ISU guidelines on what the tech panel can and should review when replaying jumps. I've been under the impression that certain aspects such as pre-rotation are not allowed to be reviewed but isn't that just asking for a moral decision from the judges. If the jump gets reviewed and the PR is spotted is the panel supposed to ignore it?
I would just like to see the official guideline and conduct of the panel in regards to reviews but can't find it in the handbook. Any help would be greatly appreciated.
Thank you so much Vanshilar for such a wonderful and informative post
Karne:
I haven't seen anyone in this thread saying such things and in fact find the discussion to be quite informative and free of such statements. I think most people just are interested in discussing the possibility of seeing such technology to assist in scoring jumps fairly.
I don't think it's necessary for anyone to qualify themselves to engage in a discussion here. We don't know everyone's involvement with the sport and frankly it's not anyone's business. All opinions are valid regardless of someone's qualifications or experience. I agree it would be nice if more people could volunteer to help with events but I don't think it should be expected. We're talking some long hours here on what is likely a well earned day off.
I would just like to see the official guideline and conduct of the panel in regards to reviews but can't find it in the handbook. Any help would be greatly appreciated.
Thank you so much Vanshilar for such a wonderful and informative post
Karne:
I haven't seen anyone in this thread saying such things and in fact find the discussion to be quite informative and free of such statements. I think most people just are interested in discussing the possibility of seeing such technology to assist in scoring jumps fairly.
I don't think it's necessary for anyone to qualify themselves to engage in a discussion here. We don't know everyone's involvement with the sport and frankly it's not anyone's business. All opinions are valid regardless of someone's qualifications or experience. I agree it would be nice if more people could volunteer to help with events but I don't think it should be expected. We're talking some long hours here on what is likely a well earned day off.
Last edited:
Are you claiming then that you've personally sat on a tech panel? Had personal experience with how a tech panel does things? I mean since you're berating other posters for not having personal experience with tech panels right?
And yet you didn't even address neither my point about Mirai nor answer my question regarding toe-pick vs blade landing (unless you mean your position is that the toe pick doesn't count and that it's when the blade proper touches the ice) -- that pointing the toes means contacting the ice earlier, means less rotation, means more likely to under-rotate, means incentivizing skaters to land more flat-footed, means harsher landings, means the jump won't look as aesthetically pleasing.
So then from your personal experience, how often are these features used to determine if an under-rotation should be called? I'm not looking for subjective speculation here, I'm looking for empirical experience/knowledge of these features being used.
The reason is because I remember specifically asking this question (whether or not slow-mo video is used to judge under-rotation) on this forum before, and my memory was the response was that it is not -- that what's used for slow-mo is judging correct edge, but not for pre-rotation nor under-rotation. But maybe my memory is faulty, and I couldn't find this question in a cursory check of the archives. So I'm looking for if someone can personally vouch (not just random speculation) that slow-motion video is used or not by the tech panel to judge under-rotation in jumps. Since you seem to imply that you have firsthand experience with these things, can you vouch for it ("I have personally seen it being used")?
And again you failed to address my point, which was "how much leeway is usually given in practice"? You say "if it's really close". Well, how much is "really close" to you? I pointed to 3 examples from the Olympics of under-rotation beyond 90 degrees that weren't called as such -- Yuna with 120 degrees on her solo 3Lz, Mao with 135 degrees on her 3A, Adelina with 160 degrees on her 3T. Do you consider them all to be "really close", especially since the tech panel has "slo-mo, super-slo-mo, reverse-slo-mo", etc. according to you?
The ISU Technical Handbook says:
("Cheated take-off" section) "The TP may only watch the replay in regular speed to determine the cheat and downgrade on the take off (more often in combinations or sequences)."
("Taking off from wrong edge (Flip/Lutz)" section) "The TP may watch the replay in slow motion."
However, in discussing under-rotated jumps, it doesn't say whether or not slow motion can be used, just that "The camera angle is important to consider when deciding upon a cheated jump particularly when the jump is at the opposite end of the rink than the camera." and "In all doubtful cases the Technical Panel should act to the benefit of the skater."
More germane to the broader discussion, the OP is broadly discussing using technology to make the judging calls more transparent (i.e. so the lay audience can better understand the calls) and if it would increase confidence in the judging. So there's basically two options here with regards to under-rotated jumps. Either my recollection is right and the tech panel doesn't use slow motion to gauge under-rotations, in which case software analysis of slow-motion video would lead to more accurate calls and thus more even judging, or my recollection is wrong and the tech panel does use slow motion to gauge under-rotations, which leads to the more uncomfortable questions of, why do they get it wrong so much and what processes can be used to improve their accuracy?
- Joined
- Feb 27, 2012
I'm sure that karne will want to answer for herself.
But now and then in previous GS posts, she has discussed her personal experiences as data entry operator at competitions.
So her firsthand knowledge of how tech panels review things is legit :yes:.
The easiest way to measure a skater's technical elements would be something close to the motion capturing technique. Dress them in black, attach colored dots and then the fun can begin! Count different positions in spins? No problem! This would also work for the footwork sequences. Everything else could be very very inaccurate. And we don't want to get it wrong, imagine a skater being short of rotation on the 3Lz, recognized and then there's half a rotation missing in a spin, so it's shouldn't be a level 4, but the skater gets level 4? Unfair! When you want to measure you have to measure everything accurately.
We only have to ask ourselves if we want this. I think the technology isn't quite there yet to do it without colored dots. Remember, a tennis ball is yellow and the lines are lines and white.
We only have to ask ourselves if we want this. I think the technology isn't quite there yet to do it without colored dots. Remember, a tennis ball is yellow and the lines are lines and white.
- Joined
- Jan 1, 2013
If you're going to call me a liar, do it outright, please.
Yes, I have sat on panels. I have been both Data and Replay Operator. I have had the privilege to listen to ISU-level Technical Specialists and Controllers. I have sat on competitions both large and small - large competitions where replay is used, and tiny club-level comps where there simply aren't enough people to have replay (or even a full panel). I have probably sat on panels at more competitions than I've competed in. I enjoy it enormously - both as a means to watch skating through a different pair of eyes and as a learning experience.
First, every skater who stands up on a jump has landed on their toe to some degree. If you land on a perfectly flat blade, you're going down without some serious luck and strength. But even a very "toe heavy" skater still does not land solely on their toepick. And yes, it is the point of contact with the ice that is examined. You would be amazed how many people do "hook turns".
It is not impossible to rotate a jump if you point your toes. Jason Brown rotates the majority of his and he has very pointed toes. Plushenko points his toes well too. He's never had a problem with URs. It is purely a technique problem if you cannot rotate jumps - it has nothing to do with how pointed your toes are.
At a competition with replay - it would be a rare occasion where a skater would not have at least one element reviewed, especially when multi-revolution jumps are in play, and more difficult features are attempted. BUT - it has to be called for review. The elements are called in real time during the program. The element is ONLY reviewed IF one of the technical panel calls for one. You would hear down the earphones, for example, "double flip under-rotated" and someone who disagrees or wants to check might say "review". The element is then flagged, and during the review period after the skater has finished, replays will be examined, yes, in slo-mo, super-slo-mo or frame-by-frame if needs be. The call will then either be changed or left dependent on the outcome of said review.
HOWEVER, if IN THE OPINION of all of the tech panel, the double flip was really under-rotated, I would hear only silence and the pre-call for the next element. The element would then NOT be looked at in the review period.
Slo mo is also very handy when trying to give a skater the feature for 8-revs in a spin, or to see if they made two revs in a difficult position.
Fine. I have personally seen it being used. I have been the one in control of the buttons. I have sat on a flight of 40 skaters and you could probably count on one hand the number of times we didn't use it.
It varies from panel to panel. TECHNICAL PANELS ARE MADE OF HUMAN BEINGS. Some are naturally more lenient than others - even when they think they're being harsh. It also varies from level to level. At a lower level, a technical panel might be a little more lenient. If, after multiple looks on the slo-mo, the panel can't agree, benefit of the doubt usually goes to the skater.
Remember, too, the panel is on a time limit. After 10 minutes the remaining skaters in the warmup group have to go on for another six-minute warmup. Spectators, coaches, skaters and their families all get quite cross with you if it takes too long. The record, I think, for the longest review period I have sat on was eight minutes - a horrible affair in which a nine-element program had eight reviews. The only thing not reviewed was the step sequence. On the flip side, if a caller has called a skater in real time and there are no reviews, it's quite possible to have it done in as little as 20 seconds.
arty:
