In a recent article for NHL.com, writer Nick Cotsonika describes the Detroit Red Wings’ use of scoring chance analysis.
Jeff Blashill receives a report the morning after each Detroit Red Wings game. The information ranges from raw data to sophisticated statistics, and what the coach cares about most is scoring chances, a metric that is more complex than it might seem.
The first place his eyes go is the total scoring chances for each team. If he's looking at players, the first place his eyes go is the even-strength plus-minus in terms of scoring chances for each individual. Digging deeper, he can see breakdowns of scoring chances by situation (such as 5-on-5, power play and penalty kill), type (such as forecheck, rush and face-off) and grade (A or B).
"It is the No. 1 thing I go back and look at," Blashill said. "Did we out-chance them, or did they out-chance us? Did a certain player create chances or give up chances, and to what degree? Ultimately, that's really what the game is about."
Blashill’s use of data is not so different than that of many other coaches I’ve worked with. Looking at objective information after the fact is a great check against personal bias.
However, how DET records scoring chance data creates certain problems.
It isn't easy to track scoring chances, though, because it isn't clear what a scoring chance is. You must come up with criteria that cover a wide range of scenarios in a fast, free-flowing game and apply them consistently. It's part science, part art. It's always evolving, and each team does it differently…
"Scoring chances can cause lots of debate in every coaching room in the NHL, or really in hockey, because there's a lot of different opinion on what's a scoring chance and what's not a scoring chance," Blashill said. "Over time, I've tried to educate myself on what's the best way to measure”…
For consistency, the same staff member tracks scoring chances for the Red Wings, and the staff constantly goes over the scoring chances afterward to make sure they are done the way Blashill wants them to be…
A shot (scored, saved, blocked or missed) is very easy for a human being - even one without a hockey background - to track.
Whether a shot, or even a non-shot, is dangerous enough to qualify as a chance, however, is a whole other debate. And debates there are, in the coaching rooms I’ve set foot in.
"I'm ultimately not trying to measure the goalie; I'm trying to measure the story of the game," Blashill said. "If we get a breakaway and miss the net, that's a scoring chance for us. If we're on a 2-on-1 and get a great shot off and miss the net, that's still a scoring chance for us. I might not give the guy that missed the net a chance, but I'm giving the team a scoring chance.
"We say that if a player makes a defensive play to negate the shot from hitting the net, then it's not a scoring chance. So if you give up a 2-on-1 and get a stick on it, even though you gave up a dangerous rush, it's not a scoring chance…"
Another significant point: When the Red Wings track scoring chances for players, they aren't "on ice," meaning a player does not get a plus or minus simply because he was on the ice when a scoring chance occurred. He has to have been involved in the play. Defensively, he is absolved if he wasn't responsible in Detroit's system.
The other wrinkle in DET’s methodology is that, unlike publically-available stats such as Corsi or Expected Goals (xG), DET’s scoring shances is not an “on-ice” stat. This means the person manually tracking chances needs to constantly assignment credit or blame to skaters on a given play.
Corsi and xG get around that complication by being indifferent to a skater’s involvement - if you’re on the ice, you get 1/5 of the credit by default, at 5v5.
DET’s analysis process is sophisticated. The raw data is tracked by an expert statskeeper, and then refined by a group of seasoned coaches.
But I think it’s a terrible process for two reasons.
1. The Statistical Factor
The debate between on-ice shot attempts (Corsi) and manually-assigned scoring chances (DET’s method, pioneered by Roger Neilson in the 1970s) is a persistent one.
Proponents of the Neilson Numbers such as David Staple point to the additional precision offered by assigning credit and blame to individuals, but analysts such as Eric Tulsky (now with CAR), Garret Hohl and others have consistently argued on behalf of on-ice statistics due to the following factors:
It is unknowable to know what actions actually cause a scoring chance, so it is best to assign equal credit to all skaters and then use statistical regression over a large sample to figure out who is driving the play
There are simply more shots in a game than chances, so it is easier to build up a large sample size of Corsi rather than chance events - and statistical regression works much better on a larger sample size
Despite the added precision proportedly offered by looking at only scoring chances as opposed to all shots, chances are largely correlated to shots - broadly speaking, hockey cares less about “how” than “how often”
2. The Human Factor
Aside from the compelling statistical argument, the main reason I dislike DET’s methodology is due to the human factors that come into play.
Having trained a number of data trackers during my time with the McGill University women’s program, I know that a person with a passing interest in hockey and an eye for detail can learn to track Corsi with minimum documentation and a few hours’ of practice.
Conversely, to be a reliable Neilson Number tracker, one needs to memorize a six-page document filled with subjective guidelines rather than firm rules.
Not only are scoring chances difficult to count, but each tracking decision (chance/no chance, player credited or blamed) can lead to a prolonged debate in the coaching room, which takes the staff off-task and affects the entire team’s work flow.
During the second half of the 2019-20 season, the new head coach of the AHL’s Toronto Marlies decided to adopt Neilson Numbers as a way to quantify the team’s performance.
Rather than rely on our full-time analyst’s Corsi-based methodology, the team’s three assistant coaches (me included) now invested 10 to 20 hours per week dissecting chances, assigning credit and organizing clips for the head coach to review.
By and large our findings lined up with our analyst’s, but now we were diverting valuable man-hours to what amounted to busy work.
Instead of connecting with our players, working with them one-on-one or taking a much-needed nap, we were hunkered down in our office, duplicating someone else’s work.
Our team went under 0.500 down the stretch and finished the COVID-shortened season outside the playoffs.
Support free content on the Hockey Tactics Newsletter
If you enjoy this newsletter and want to see more free content, then join a select group of professional players, coaches and managers and read about hockey from a unique perspective. Buy the ebooks now