If you have not noticed, we have now entered the 21st Century. The digital and computer age is upon us after being invited in the 20th Century, it is now being refined and groomed here in the 21st.
Several sports, Soccer, Baseball, Tennis even Basketball have taken to technology to assist them in helping make their teams and sports better for the players, teams and yes the fans too. Numerous methods and approaches have been used in those sports. But so far, the only digital revolution hockey has seen is in its broadcasts, and the use of "advanced" statistics, NOTHING for the actual on-ice real time play. Hockey did at one time try the "Foxtrax" a novel concept, that in its use it was the ridicule of Hockey traditionalist and any fan with an even slight desire to follow how the game is played, not just where the puck was.
Foxtrax is now nearly 20 years in the past (seriously it is). New technologies have evolved, but are they right for Hockey? I will focus on two here, FIFA’s new Goal Line technology and Major League Baseball’s, MLB Advanced Media (MLBAM).
GOAL LINE TECHNOLOGY
Goal Line technology is not new this year, but this is the first World Cup that is using it and it is based off of the Hawk-Eye system, the same system Tennis uses. It got off to a fabulous start by awarding France a goal in its game vs Honduras that NO camera could definitively say the ball had fully crossed the goal plane.
We have all seen goals we have felt should have or should not have been awarded in hockey, so it is very reasonable to simply state "Give me that for Hockey". But it is not that simple, in fact for hockey it is very far from simple for several reasons.
First, we need to understand what the technology is and how it works. It is simply in concept, triangulation. It uses multiple lines of bearing and elevation from known locations, determines where these lines intersect and this will state where the object being observed is located. This is done with cameras, high speed cameras, the cameras are synced in time so they can accurately be compiled together to determine the location of the ball in space. When a camera is viewing the Soccer goal, the location of each camera is known, by being fixed in place and accurately referenced in range and altitude to the goal. For any given frame for any given camera, the horizontal location and vertical location of the ball in the frame will be determined. Through trigonometry these locations in frame are turned into a bearing (from the horizontal location in frame) and elevation (from the vertical location in frame) from the sensor (the camera) into a line of bearing and elevation FROM the sensor. This line is plotted onto a layout of the field in 3D (using a computer) and any other reliable camera is also applied from the same moment in time. Where these lines of Bearing and Altitudes intersect marks the balls position in time and space, making this a 4D observation of the ball (Altitude, Longitude, Latitude and Time).
Good Goal. Center of the Puck is 0.51" from Goal Line. Flat it would need to be 1.51" from center and on diagonal 1.58" from center
For both Soccer and Tennis, each frame software analyzes the pixels and determines which ones are the ball to generate the 4D location and path of the ball. This is where things start to get dicey in regards to Hockey. Soccer and Tennis use round balls, with known and symmetric deformation values, Hockey uses a short asymmetric cylinder that deforms asymmetrically. This mapping system does not give puck orientation, which is vital in determining exact puck position in space. Without knowing what the orientation of puck it is impossible determine if the puck crossed the goal line. Because of the unique shape of a hockey puck, it can be 0.51 inches from its center across the goal line on edge (photo above) and be a goal, but be 1.55 inches across the goal line from its center flatter on the long diagonal and not be a goal. This due to the fact the extreme width of the puck is 3.16" and at its narrowest it is 1". Knowing how the puck is lying and how it is rotated in space is vital for determining if a goal should be awarded, Hawk-Eye does not do this. Without markers on the puck to let the camera map the position, it is not possible yet to determine the pucks orientation, especially in lighting conditions that give no light to see the pucks edges.
Another shortcoming for the goal line technology that FIFA uses, is the number of RELIABLE cameras or camera angles available in hockey during a contested goal call. Typically when a goal is under review in hockey it is mainly because it is unseen or it is very difficult to see by more than one or two angles. Reviews for goals that are in the plain open area for the most part a much easier to determine if it was in fact a goal with existing technology being used with the overhead and in net cameras. The puck being obscured in multiple most camera angles makes it impossible for Hawk-eye technology work since it relies on at least three angles to determine the pucks point in space. If only 1 or 2 angles exist, this is insufficient for the system to determine the pucks location, and we are back at square one with how things are done today. It is also these types of goals, where the view is obscured that are the ones that cause the frustration, the very ones that could NOT be checked with Hawk-Eye.
Even though cameras used in Soccer matches may be much further away from the goals, the relative size of the ball in view, vs that of the puck is also a factor in accuracy. This is the first inherent system error (in addition to others) it is video accuracy. I do not have exact numbers, but after looking at photos of numerous Hockey and Soccer stadiums, I am estimating camera placement would be on average around 150 feet from the net for Hockey and 250 feet for Soccer. Camera placement and calibration is crucial. The cameras need to be in a fixed location that is not going to move, so their position is always known. So they cannot be placed on the stands due to crowds vibrating them and skewing the image, nor can they be placed in the nets, since these also move when the nets are struck, or even partially dislodged like hockey. This will change their position in space, introducing large errors in the ball/puck location. So the cameras need to be mounted to the firmest structures possible, like the overheads or rafters (though if the seating structure is solid enough they could be placed their if properly damped from vibrations, but I would think this is less then ideal).
With these setting you get the following with a soccer ball from 250 feet and a puck at 150 feet (using a camera that is viewing the goal plus 15%,)
|Ball Puck Width (in)||Camera Distance from Goal (ft)||Ball/Puck % of Goal Width||Ball/Puck width angle from camera||1080p/i Pixel width of Ball/Puck||Width per Pixel (in)|
- The two ball sizes are the extremes of what is allowed for a soccer ball, and the puck is the 1’’x 3" dimensions.
- Ball/Puck % of goal width is: the size (in %) of the ball/puck in relation to the size of the net (horizontally)
- Ball/Puck width angle from camera: This is the size of the angle of the ball/puck creates from the camera at the stated distance.
- 1080p/i Pixel width of Ball/Puck: How many pixels wide/tall the ball/puck occupy including the viewing area 15% larger than the goal as stated.
- Width per Pixel (in): The width of each pixel using the above conditions.
The most significant portions are the Angle, ball/puck pixel width, and width of pixels.
The angle of the object is part of the focal accuracy of the camera. The narrower the angle, the more error in determining the EXACT position of center and outer edges of an object. Think of it like this, which is harder to determine from across the room with your eyes. Where the center and left/right edges of a Sharpie is, or that of a pencil. Also, lens diffraction (how light bends in glass and appears on a sensor) becomes a larger factor the smaller the object/angles get.
This also relates to the pixel width. The more pixels and object occupies, the more accurate your measurement of where that object exactly is. As you can see the puck can display many to few pixel widths, both at the same time. However both are needed to be accurate to determine location. The way sampling occurs to determine if something is occurring in a bin (pixel), can result in the apparent width being two pixels larger than actual. The reason for this is, if enough of the object overlaps into an adjacent pixel, long enough during the sample (frame), it may cause the analysis threshold to determine the object occupies that pixel, this can happen on all edges at the adjacent pixel boundary. Because of this processing, a system can (and they do) display/state the object occupies a width greater than actual width or conversely, if sensitivity is not sensitive enough, it can reflect a width less than actual width resulting in a goal where it should not have been awarded. So the fewer pixels an object actually occupies to start off with, makes this two pixel error more significant, and in the case of a puck, that is nearly a 10% potential error on its side perspective. If the sample rate is too slow, this error increases to many more pixels, since the object will have moved in distance during the sample, filling more bins. The only way to minimize this effect is to decrease the sample time (higher speed cameras) and/or increase resolution. But this error will always be present in some fashion.
Finally, the pixel width per inch. For each potential pixel error in determining the location of the object, this is the final error that can/will be seen per pixel. 0.044" does not sound like much, but even if the error is only 1 pixel in each direction (0.04") for hockey, that would mean an inherent error on 1/11", and this would be under ideal conditions. This error would also make any attempt to mark the puck to determine orientation meaningless, since this error can mean the puck can have a rotational error of over 14 degrees at least (based on two points showing similar errors in exact position). This is significant and will effect the close calls we want to get right.
The technology is sexy and seems like it would be an "easy thing", but it is far from it. Hockey is not a sport that easily lends itself to this technology……………..yet. That may change at some time, just not now. Also the price is of it in astronomical, Soccer clubs in Europe that have bought the systems for their leagues are paying $800,000+ per team/stadium for installation, some have paid twice that. When you combine the inability to determine puck orientation the probability that the puck will not have sufficient camera coverage, the camera error, the range and pixel errors, along with the hefty price tag, all these together it is not ready for hockey.
How about an active system, like Foxtrax? As I pointed out earlier, puck orientation is as important as puck location. An active sensor may be able to tell you fairly accurately where the puck is in relation to something, but the further you get from that something the accuracy drops. You can’t you use the goal, since the goal can move 2 inches in any direction (but down) and still be considered "moored". This is an unacceptable error. So it would have to be fixed sensors around the stadium, and like with Hawk-Eye, accuracy drops with distance. That is the simple part, the hard part is determining puck orientation at any exact moment. This would require a gyroscope inside the puck, or markers on the outside of the puck, which will require a clean line of sight to see them, and at that point you are now back to the same errors as Hawk-eye in accuracy. A Gyroscope inside the puck? Considering puck can easily exceed 200G’s from a 100mph slapshot hitting the post, I do not think an imbedded gyro is going to last or be cost effective. At best, with todays technology, an Active system can give you an approximation of puck location, but when dealing with goals, that is just not good enough.
ADVANCED MEDIA (MLBAM)
Major League Baseball has developed a yet unnamed tool, right now simply called, Major League Baseball Advanced Media (MLBAM).
Now, this is not a tool for analyzing if a goal was or was not scored, but if (and hopefully when), it will be an outstanding tool for analyzing HOW a goal was or was not scored.
Watch this video for a quick idea of what it is doing.
MLB says this is NOW being done for EVERY PLAY at select MLB Stadiums and plans for it to be ready no later than next season at every stadium.
It tracks the speed and rotation of every pitch. Tracks the exit velocity of the ball from the bat. How fast the runner travels the bases. The reaction time of the fielders, the speed of the fielders, distance covered and path efficiency. Basically it tracks an entire Baseball play and all 10 to 13 players on the field (9 fielders, 1 batter and 0-3 runners). In the near future, MLB hopes for it to be able to not just track the player, but also track arm/leg position as well.
This could be a fairly easy transition to hockey. MLB has the advantage of knowing who is where when a play begin, hockey does not and that would be the biggest obstacle. But that is actually a small obstacle that can be overcome, by embedded tags in players uniforms, or player number recognition technology combined with one that can track a player and also back track as well if needed. The rest would use the existing setup like baseball, tuned to track a black puck, not a white ball.
In regards to Hockey, the possibilities are endless of how this effect the game, on the ice, in preparations for games, for the stat heads (like me) and the fans.
You can break down plays and routes, map them out and start adding values to these types of things. I am not going to get into exactly what can be done with the technology, but feel free to discuss, please! But this is certainly a technology that is ready for hockey, and hopefully makes its appearance much sooner than later.