What is it? What can it do?
With a simple Google search, I have seen many reports indicating that the NHL will be implementing RFID technology in their buildings in the future. What I have not seen is anyone actually cover how it works, it’s limitations, and how it will assist in creating more helpful data analysis tools and new statistics. Of all things #FancyStats and what it’s future holds, RFID to me, will answer many questions and create many more tools in assessing player ability. #FancyStats, to this day, have either been taken as being Gospel or thrown completely out the window. I can tell you right now, I take neither of those approaches. I love the fact that we can dig deeper into a player’s career by the use of shot metrics, QoC, adj SV%…to just name a few. What I will never admit, is that watching a player play hockey has NO value in assessment. Without even showing charts, numbers or going into long rants; both crowds if level headed, should agree on that: It would not be very hard to find two players with like possession metrics, see their names and say, “Well that ain’t the whole story.” My hope is that RFID along with Visual Analysis will one day give the “Whole Story.”
RFID – Radio Frequency Identification
According to www.dictionary.com, it defines RFID or Radio Frequency Identification like so:
radio frequency identification: a technology that uses electronic tags placed on objects, people, or animals to relay identifying information to an electronic reader by means of radio waves
An excitement antenna does exactly what you’d expect given its name. Its purpose is to excite, or power the RFID tags. However, some tags do not require excitement, so an excitement antenna is sometimes not required.
Again, its name is very fitting. This antenna is responsible for gathering the RFID data sent by the tag. In the consumer world, it is usually a product code that would help describe a product after searching a database. In the NHL, this is how a player and his team would be identified, or how they would identify the puck and other location markers in the ice.
This is what the antennas would be powering and/or receiving information from. Within a standard passive tag, there would be a microchip and an antenna. Active or Semi-Active types of tags require a small battery as well. For this article, I will be assuming that the NHL will be using Active tags that do not require to be powered by excitement antennas. Active tags are always transmitting information when activated. I am assuming this because the NHL and its teams have a lot more disposable cash than a smaller entity such as a department store that uses RFID technology. Furthermore, the tag will most likely be transmitting a player’s personal “Code.”
A PCU by definition is a Process Control Unit. This is where all the magic happens. Within this unit, all the data logging takes place. The unit I have worked with allows you to set up a graphical grid which enables you to see your product as it sits on the storage floor and as it moves from one position to another. In the NHL’s case, it could be setup to show a graphical grid of an ice surface. In essence, you could simply sit back and enjoy watching the action happen. Well not really, it would not be much fun to watch dots travel across a computer screen…but hey, those characters in the matrix sure loved watching green ASCII code fall from the top of a screen. I should also mention that no actual data analysis will be done in the PCU itself. This will most likely be done by a computer belonging to the NHL or an NHL team that wants live analysis results.
A Simple Example of Everyday RFID Use
RFID technology is not anything new. Stores like Walmart, Best Buy, various department stores, and even your local school or public library use RFID as a loss prevention tool. Remember that time you walked out of a department store where the cashier never scanned your wrangler jeans correctly and you set the exit alarm off. That was RFID that embarrassed you ever so gracefully. Somewhere on the wrangler’s you had just purchased, there was fabric tab (in some cases) that was equipped with what looks like a flexible circuit board. This is something called an EPC (Electronic Product Code) tag. On your way out of the store, you would have noticed a usually gray, 5 foot plastic thing on either side of the doorway. These units were the RFID excitement and receiver antennas.
At the moment you walked through the RFID barrier:
- The gray “5 foot plastic thing” would excite the EPC
- Then the EPC would transmit a radio signal
- The gray “5 foot plastic thing” will receive the signal, and instantly check the database of product codes to make sure that the item(s) leaving the store were in fact purchased.
- So depending on the database search, the alarm would either sound and security would check your bag and bill (if you had one) or you would exit the store normally.
Though this a very simple and common example of RFID, that is basically how it works.
- An excitement antenna, excites/powers the EPC
- The EPC(RFID Tag) then transmits a signal
- The receiver antennas receive the signal
RFID – What it can do
Location by Triangulation
The easiest way to pinpoint the location of a radio frequency is to use triangulation. Think of a ship lost at sea. This ship is giving a “may day” radio signal calling out for help. Let’s just assume that every 10 dBMs of signal strength is easily converted to 10 kms of distance. If at Location A (Shore) the dBM signal strength is -90 dBMs, we could then assume that the ship (Location C) is somewhere within a 90km radius of Location A. This is all we can calculate with two locations; a transmit location and a receiver location. What we do know, is that the ship is lost at sea and can only be found in the blue area of the diagram posted below.
If the “may day” signal is being received at another location, Location B, and the signal strength is -120 dBMs, then we would assume it is somewhere within a 120 km radius. With having two known locations that are receiving the signal, we can now find the third unknown location (the ship) where the two radiuses intersect. Without getting into the trigonometry(triangulation) behind finding the location, we can see with a graphical representation the location of the ship.
If we change the names on the diagram, we can see how a PCU will locate an NHL player or puck. In other words, instead of a ship, we would have a player or a puck. Instead of the sea, we would have the ice surface. With known locations of receiver antennas, we would know what side of the ice surface is relative to receiver locations. Please note that I have replaced km with feet in the next example.
Now that you have a basic understanding of how the NHL may determine where the puck or player is at any given instance, we can explore how the NHL will know how fast a player or puck is travelling. This can be calculated using many instances in time vs changes in location. If the tracking software records the 1st instance of time at zero seconds and records a 2nd instance at one second, then we now know that the object had travelled X amount of metres over 1 second. If X was an easy number like 10 metres, this would mean the object had travelled 10 m/s. This would also mean that the object was travelling at 36 kph (3.6 x Xm/s = kph). Of course the software will be comparing a lot more instances in time than two over 1 second, but for this post I want to keep the tech side easy to grasp.
A Little Background on Myself
During the day, I work as an Industrial Electrician at a finished goods processing company in Southern Ontario. Recently, we had attempted to track our “finished goods” in our storage areas. I was tasked as a maintenance liaison to the R&D team who was responsible for attempting to track our goods using RFID in a cost effective manner. The origin of the project was to reduce lost time, due to lost (unlocatable) goods. The major problem we had run into was that we were tracking metal. Metal, as some of us know, is very good at blocking and reflecting radio signals. At the time the project was attempted, we found that the only way to reduce “noise” and effectively track product location was to use, active (powered) EPC tags. This was not a cost effective solution and we have since put the project on hold. Obviously, due to the NHL’s unlimited financial resources, this should not be an issue.
During the proof of concept phase of our tracking project, I was amazed by all the data that we had created about the movement of our product. Not only could we see where the product was placed (X,Y,Z axes) most of the time, we could actually see the product movement throughout the whole storage bay over any span of time. This is when I realized, that if our company wanted to, we could rate the efficiency of the product handler based on the way he or she moved and re-located product vs. how others would move and relocate the product. Though not our end game, this is what had me intrigued when I first heard that the NHL was to use the NHL All-Star game as a so called “proof of concept.”
As a person who has worked with RFID, I can confidently say that the amount of data that can be recorded, though limited, can and will bring limitless amounts of newly compiled player, goalie and team statistics. A bold statement, yes, but it is the truth. It is the data that will be created through the use of analysis algorithms that has tickled my fancy. Furthermore, I am completely unsure what the NHL will record, but they will probably be limited to something like this:
- Player(s) Location (2 dimensional)
- Goalie(s) Location (2 and 3 dimensional)
- Player(s) Speed
- Puck Location (2 and 3 dimensional)
- Puck Speed
How will this create more data?
Those four alone, though seemingly straightforward, can create many simple to complex statistics. Some “simple” created player/team statistics could be, but not limited too, are:
- Players avg. speed
- Players top speed
- Players acceleration time
- Pass speed
- Pass Reception Rates
- Pass Accuracy
- Shot Speed
- Shot Location
- Goalies cross crease post to post movement in time
- Goalies reaction time vs shot release
- Goalies 2 dimensional position vs net location vs shot origin location
- Real Time Possession Times
Finally – A Hockey Example…Only 1 for now
RFID – Possession – In Real Time
Currently, the Hockey Analytics community is using this formula: Total Shot Attempts For over Total Shot Attempts For + Shot Attempts Against multiplied by 100 (CF/(CF+CA)*100). This percentage is used as a proxy to measure puck possession. In theory, a team that possess the puck more than their opponent has a greater chance of winning a game than a team who possess the puck less. Theoretically, the more a team has the puck, the more scoring chances should arise. With that being said, RFID should be able to give us a more accurate representation of what team possesses the puck more. What it will also tell us is how long the puck is not possessed by either team. The NHL can track this by using a triangulation comparison.
Here’s what we will know for sure:
- Where the puck is
- Where the player(s) is
- Who the player(s) is
- Who the player(s) plays for
- Where Antenna “A” is
- Where Antenna “B” is
With all of this information, why can’t we figure out who has the puck and for how long? IF Player “X” on Team Blue is moving in the same direction as the puck “P” AND is within close proximity to it, can we not assume they have possession of the puck? If the puck is passed and comes within close proximity to a Player “Y” who is also on Team Blue, can we not assume that during that whole time frame Team Blue possessed the puck? The player and puck proximity can easily be done by simply comparing their on ice locations.
Now let’s get fancy. In the scenario above, IF during the pass, the puck “P” stopped AND/OR changed direction AND is now in close proximity to Player “A” who is on Team Red, we could now assume that possession has changed over.
Let’s get even fancier by examining teams only. IF a player from Team Blue is traveling up the side boards AND a player from Team Red is getting closer to the puck, trying to squeeze the player from Team Blue off of the puck, some may assume there will be a couple instances in time where their proximities could be close enough to confuse the system. WRONG. The system knows who had the puck “P” before proximity equality and after. Either the player from Team Blue will squeeze through and still be in close proximity to the puck and the player from Team Red will be getting farther away from the puck OR another player from Team Blue will pick up the stale puck and continue possession due to proximity OR a player from Team Red will get the the puck due to proximity and perhaps change its direction of travel. Examining puck AND player direction of travel before AND after comparative instances can also be helpful in determining if possession was changed after close player-to-player puck events.
What I am trying to say is this can all be sorted out with a couple lines of code. Keep in mind that all of the above is just an assumption/example of how the NHL could track real time possession rates. Even though Corsi For % is a good proxy of possession time, there is nothing wrong with having an actual possession value to work with. RFID is something that either will now or will in the future be able to give actual possession times rather than a proxy of time.
What’s To Come
As you could have guessed, I gained a background in programming before beginning my current career. To this day, I still do a lot of hobby programming in my free time. Even within my career choice, I have many responsibilities which include programming in “ladder logic” languages. For me, it is easy to envision how I would program the comparatives above and how to cancel out noise within the system. For others, it is not that easy. The main reasoning behind this post and future ones, is not just to explain how RFID will change hockey analysis. What I really would like is to help the regular viewer understand how it works and why it will work. I want others to see how many different aspects of the game can and hopefully will be tracked. Creation of metrics similar to QB Passer Ratings are going to become a real possibility. RFID could finally put the “Shot Quality” or “Quality Shot” debate to rest. The options really are limitless and will only be limited to the amount of money the league and/or a team is willing to spend. The future of this technology in hockey looks bright as all teams will definitely want an edge over one another. In the end, I think the NHL will have their own data that they will want to be tracked and their own statistics that they will want to be released. For this reason, it is at the team level where I get the most excited. Being a fan of the richest team in the league, I would hope that they spend the most money on analyzing the new RFID data. All the RFID data that comes from the arenas will be the same and will most likely be regulated by the NHL. It is what you do with the data that counts. It is how creative you can get with analyzing it, and how far outside the box you can think.
In one of the next parts of my RFID series, I will be expanding on some of the above ideas I have posted. I will also go through some of the other more complex ones I have in mind. Fortunately for you, I won’t be delving into the math involved and/or the source code examples of how it can be done. I am trying to keep it as simple as I can because I want everyone to enjoy what RFID is and what it could bring to the table. I have also reached out to a work colleague I had met during the proof of concept phase I was involved in. She is an RFID expert working with Motorola, a hockey fan and a hockey player herself. She has agreed to do a Q&A session with us at Leafs Hub and has also agreed to help me … help you … understand RFID.
The future of RFID in hockey is still unknown, but the technology should shed light on some of the unknowns in the “Hockey Analytics” of today.
If you feel that I have misrepresented RFID in any way, please feel free to contact me via email. We can sort it out and I can make edits anytime. firstname.lastname@example.org