Offside involves tracking the position of the skater at the moment the puck crosses the line, puck tracking technology does not really help that at all.
Based on your sentence, it actually solves 1/2 the problem. Let's bar code the skaters and track them too.
Based on your sentence, it actually solves 1/2 the problem. Let's bar code the skaters and track them too.
Wow, that is a gross oversimplification, isn’t it? No it really doesn’t. A puck is a fixed shape object with a simple shape. Many moving parts in a player.
Bar codes? Imagine all of the out of work stock clerks who can skate and that could do the scanning.
I don't think this tech is going to be used for either of those things. It's going to be for analytics.
For the first few years, for sure. But I'll bet this is where we're headed with this. Once the puck has accurate and proven tracking, I would be surprised if they didn't leverage it to make more accurate calls.
Microsecond accurate wouldn't be a problem so long as the call is near-instantaneous.
It is the offside being called 30 seconds later after a goal that pisses everyone off.
If you are 0.12s offside and it is called immediately, no one would have a problem with it.
That's a really good point. If it's blown right away, then we just have the frustration of a play developing that was blown down. Which we have now, and it's only frustrating when it's wrong.
Offside involves tracking the position of the skater at the moment the puck crosses the line, puck tracking technology does not really help that at all.
The thought is more 'what will this become' than 'what do we have right now'. It's a pretty wise decision for the NHL to start testing this tech to gain game insights before relying on it to make calls.
The hardest part to track is the puck, so when that's reliable tracking the players becomes much easier.
TBH, I thought some of the tech from the Olympics were going to be used when they announced cameras determining offside calls. Apparently they decided used cellphones duct-taped to the boards were sufficient...
Anyways, here's tech that exists today that would help determine offside calls:
The Following User Says Thank You to Split98 For This Useful Post:
In terms of puck across the line, if they put in 2 sensors couldn't they always know the orientation of the puck? If each was a half an inch from the edge on opposite sides, they could then draw the overall shape using those 2 locations?
I wonder what the potential tolerance would be? With the hawk-eye system used in tennis and other sports, it claims a 3.6mm tolerance. But when trying to make those discrete decision of goal or no goal, even 3.6mm could be valuable. With that said though, the current video system has lots of faults and can often flub up even greater distances due to obstructed view. It could still be an improvement while being imperfect.
The thought is more 'what will this become' than 'what do we have right now'. It's a pretty wise decision for the NHL to start testing this tech to gain game insights before relying on it to make calls.
The hardest part to track is the puck, so when that's reliable tracking the players becomes much easier.
TBH, I thought some of the tech from the Olympics were going to be used when they announced cameras determining offside calls. Apparently they decided used cellphones duct-taped to the boards were sufficient...
Anyways, here's tech that exists today that would help determine offside calls:
It is a starting point.
First, you need to change the rule of offside from having a skate on the ice, on or before the line, to breaking the plane.
Sprinters are breaking the plane going one direction. Probably have to disallow tagging up.
False positives with stick breaking the plane.
Even that video shows that there is manual intervention to finalize the race result.
Isolating body parts from multiple skaters, that may be going different directions, and are on different teams, it seems a lot more complex.
I suspect we are a good way from solving this simple problem.
* and I think automating tracking of the players is much harder than doing so for the puck simply due to the complexity of the shape of the player and all of the moving parts.
Last edited by DeluxeMoustache; 05-23-2018 at 09:39 AM.
First, you need to change the rule of offside from having a skate on the ice, on or before the line, to breaking the plane.
Sprinters are breaking the plane going one direction. Probably have to disallow tagging up.
False positives with stick breaking the plane.
Even that video shows that there is manual intervention to finalize the race result.
Isolating body parts from multiple skaters, that may be going different directions, and are on different teams, it seems a lot more complex.
I suspect we are a good way from solving this simple problem.
* and I think automating tracking of the players is much harder than doing so for the puck simply due to the complexity of the shape of the player and all of the moving parts.
Very true. I think my brain switched from automating the process to making the process much better with this tech. None of that I'm suggesting really is automated, is it.
It's likely not feasible, but I guess the same sensors in the puck could be used on skates?
Besides that, I think we're looking at relying on the same image recognition tech we are seeing in security. A coloured mark on each players skates to determine which team they're on would tell the camera that a Flames skate has crossed the line. If that correlates with the puck incorrectly = offside?
EDIT:
Totally forgot about what Google and Tesla are doing right now. It would require a collaboration between someone like that and the NHL, but I guess the tech for recognition is here today. Having the dataset of those restricted to recognizing a skate only seems pretty reasonable.
To limit the cost and not need insane compute, it could be set to analyze images only when a puck crosses the line. Scan the 10,000 or so (most of what I've seen is calculating images/second well above 10,000) images that match the second the puck crossed and I think we could see results pretty quickly.
Very true. I think my brain switched from automating the process to making the process much better with this tech. None of that I'm suggesting really is automated, is it.
It's likely not feasible, but I guess the same sensors in the puck could be used on skates?
Besides that, I think we're looking at relying on the same image recognition tech we are seeing in security. A coloured mark on each players skates to determine which team they're on would tell the camera that a Flames skate has crossed the line. If that correlates with the puck incorrectly = offside?
I think RFID chips are pretty small and could be embedded in the heel and toe of skate boots/ blade holders. There would have to be a database with the chips from every skate in behind the scenes.
I think RFID chips are pretty small and could be embedded in the heel and toe of skate boots/ blade holders. There would have to be a database with the chips from every skate in behind the scenes.
And knowing the NHL, SAP would handle that database.
The Following User Says Thank You to Split98 For This Useful Post:
Not to derail, but the reason I was thinking 2 is that the puck is symmetrical. It wouldn't matter if the puck was right-side up or not as they would be centered vertically, and spin would not be a problem as the 2 sensors would give that axis.
Its totally arbitrary and I am not trying to be argumentative, but just trying to conceptualize the physics of it.
Not to derail, but the reason I was thinking 2 is that the puck is symmetrical. It wouldn't matter if the puck was right-side up or not as they would be centered vertically, and spin would not be a problem as the 2 sensors would give that axis.
Its totally arbitrary and I am not trying to be argumentative, but just trying to conceptualize the physics of it.
I think his argument for three was so that you know where the puck is on the ice (similar to how you need 3 gps satellite signals to determine where your car is)
The Following User Says Thank You to Canehdianman For This Useful Post:
Would putting a sensor in the puck change the overall physics of a normal pick? If so would it be drastic?
I'm no math guy but in theory if you put a sensor in the puck and then two sensors on opposite sides, inside the net.
Then place them behind the line approximately the radius of the puck (1.5 inches) + the space needed to techically call it a goal (let's say 0.1 inches) wouldn't that be an easy indicator for a goal? I guess that would only be if the puck is laying flat.
Also how much would a Chara snapshot affect the functional properties of the sensor in the puck. Will it still work 100% fine if it's rocketed 105 mph into the boards?
I wonder if they have a few sensors at ice level if they would be able to easily determine if a puck is knocked in by a high stick. Should be easy enough to at least do that.
The Following User Says Thank You to Robbob For This Useful Post:
Not to derail, but the reason I was thinking 2 is that the puck is symmetrical. It wouldn't matter if the puck was right-side up or not as they would be centered vertically, and spin would not be a problem as the 2 sensors would give that axis.
Its totally arbitrary and I am not trying to be argumentative, but just trying to conceptualize the physics of it.
The trouble with two sensors:
Hold a puck between your thumb and forefinger, so that your two sensors are along the line between them. You can still spin the puck in the third dimension without moving the sensors. With only two sensors, the puck could just barely enter the net standing on end, or it could be horizontal with part of the disc still outside, and the data from the sensors would not tell you that. If you put three sensors in a triangular arrangement, you can always tell the angle as well as position of the puck.
HTH.
__________________
WARNING: The preceding message may not have been processed in a sarcasm-free facility.
The Following 3 Users Say Thank You to Jay Random For This Useful Post:
If you draw a line right through the Center of the puck, the distance from that line to the top of the puck is 0.5”. The distance from that line to the edge of the puck is 1.5”
The Following User Says Thank You to DeluxeMoustache For This Useful Post:
I wonder if they have a few sensors at ice level if they would be able to easily determine if a puck is knocked in by a high stick. Should be easy enough to at least do that.
It might even be as easy as looking at the puck data as it is shot over. If there was no more data of impact after it was initially shot, it's out. If there's another data point of impact, that would be the glass or another player.
