My NHL Research Process

While being able to see the projections daily is an excellent way to gauge who are the studs and the duds on a particular slate, it’s also essential to realize why players might being good or bad matchups. These concepts will be the point of this article, to inform you, the reader, as to what my process for researching NHL is. Here, I will break down all of the different stats that I look at, plus some roster construction tips and how I go about building multiple lineups. 


Different Corsi Stats – If you’ve been reading the rundowns this season, then you should realize at this point that Corsi is a frequently referenced statistic in each article. For those that aren’t familiar, Corsi is any attempt at a shot that a skater performs toward the goalie. These include everything, including missed shots, block shots, and of course, actual shots on goal. I weight this stat very heavily because Corsi can tell us a variety of different things about a player or a team. For example, a team’s Corsi for percentage can say to us the rate of Corsi attempts a team has compared to their opponents. This information is vital because it can tell us who should win the puck possession battle throughout a game. This information leads to knowing who will shoot the puck more, and thus, have more opportunities for scoring goals. We can also look at Corsi for and Corsi against / 60 minutes, as I often cite in the article. For these stats, Corsi against, more importantly, lets us know how many of these shot attempts are being allowed by a team per game. The more Corsi attempts a team allows per game, the more likely I am to target their defense. The New York Rangers so far are the prime example of this. At the time of this writing, they are allowing 66.06 Corsi attempts to their opponents per game. Which is far more opportunities than most teams allow. In theory, a team that plays the New York Rangers will have a lot more goal scoring chances been a team that plays the Pittsburgh Penguins, the team allowing the fewest Corsi attempts per game to their opponents at 49.25. There is a similar stat to Corsi, called Fenwick. The only difference between the two is that Fenwick does not include block shots. It’s a matter of personal preference, but I prefer including blocked shots as it then encompasses all of the data instead of just most of it. 

Expected Goals – The expected goals metric derives from people much smarter than I, who can calculate a value for goals on each shot. Whether or not they were actually goals. This goes down to the quality of the shot itself. So, for example, a shot right in front of the goalie would be worth more to a skater’s expected goals been a blue line shot for a defenseman would be. Giving up the former is much easier to score on the latter. The usage for such a statistic is quite apparent, as we would want to know who is getting high-quality shots, and who might be allowing high-quality shots to their opponents.  Using the New York Rangers again as an example, they are allowing 2.84 expected goals per game to their opponents. Combining that with the opportunities that they allow on a game-by-game basis, they are an extremely advantageous team to target. As opposed to the Dallas Stars, who we might want to avoid, as they allow only 1.86 expected goals per game as of this writing. When I cite these metrics in my article, I usually do so excluding the goalie from the equation. While understanding goalie statistics is essential in evaluating a matchup, we have to look at the team that a skater is going up against as a whole. And looking at expected goals allowed from the skaters that are playing in the game is the best way to do so. Another statistic that can tell you a similar story is high danger shots. High danger shots come from the classification of different shots based on where they were shot, which goes back to the quality of shot discussion that I brought up earlier. I do not use this a ton, as I believe the expected goals metric already takes in to account a lot of what high danger shots do. 

Zone Starts – All zone starts does is look at the percentage of the time a player or a team begins the possession, and they’re offensive, defensive, or the neutral zone. The reason why I value these metrics so much is that the more a team starts in their offensive zone, the more opportunities that they are going to get to score a goal. Not only that, but it lowers the probability that they allow a goal because it is challenging for a team to score from their defensive zone. Looking at a player like Brad Marchand, for example, he currently has a 69.61 offensive zone start percentage, which means that he starts almost 70% of his faceoffs in his offensive zone, which is extremely impressive. The way I would interpret this is that I would be hard-pressed to use top-line players from opposing teams that we’re going up against a Brad Marchand lead line — while possibly targeting players, for example, that we’re going up against Esa Lindell. Who currently only has a 43.1% offensive zone start percentage. 

These stats can be factored in from even strength and power-play perspectives. These are stats that are recorded throughout the game and can tell you which players and teams the stack. Whether it be their even-strength units or their power plays. 

Lineup Construction

While I have gone over lineup construction in a previous strategy article this season, I think it deserves an update based on what I have learned so far in a little over a month and a half of the NHL season. DraftKings updated the scoring for NHL over the offseason to include a multitude of different bonuses for skaters and goalies, as well as increase the value a pretty much every stat. Many in the daily fantasy hockey community debated on how this would change lineup construction, and how winning lineups would look this season and beyond. For the most part, I don’t see much of a difference in winning lineups from this year as compared to previous years. However, that’s not to say that nothing has changed. I find that the more games on the schedule, the less that full-line stacking becomes the optimal way of building lineups. On a hypothetical slate that includes seven games, we might see an optimal lineup that has two different three-man stacks, two defensemen from different teams, and the goalie. Which may or may not be stacked with one of the lines, which is something that we’ve seen for quite a while already. But when I look at slates that included ten or more games, I see that this type of stacking doesn’t seem to work as well. 

The winning lineups on these nights do include at least one three-man stack. Still, instead of adding a second three-man stack, the line-ups either use pairings of different players from different teams or a cavalcade of different players that just had a lot of upside. These are just things that I’ve seen from winning lineups and may seem like more of an exception and a trend. We have to remember that we’re still not even two months into the new season, so expect further analysis on this topic later on in the season.

For me, I still use traditional stacking methods, as I stated before. The only exception to this is on smaller slates of fewer than four games, where I tend to go with a five or six-man stack that always includes the goalie and probably stacking a power-play unit as opposed to one or two even-strength lines. In these lineups, I prefer stacking the power play units because these skaters tend to have more upside then some even-strength lines do. No matter what, though, I only play defensemen that are on their respective team’s power-play lines. You’re leaving a lot of upside on the table by consistently playing defensemen that are not on their team’s power-play lines. 

The last topic that I want to cover here is how to multi-enter these contests. There is a lot of debate by many in the daily fantasy community as to what the best method of doing so would be. Of course, making sure you’re only including players that you want to roster, but there’s certainly plenty of questions to ask before hitting that optimize button. For example, what are the ideal exposures for players? Luckily, some of that work has already been done for us by some brilliant MIT students that were asking themselves very similar questions at any daily fantasy hockey player would. When it comes to hockey, I prefer to diversify as much as possible. As opposed to a sport like basketball where I would have a core and stick with it. Of course, we know that hockey is a lot more volatile than basketball on a nightly basis. In the students’ research, they found that the optimal diversification of lineups is based on the size of the slate. So, for example, a night where there are less than four games on the schedule, you would want to have a relatively small amount of diversification between lineups. As opposed to a night with more than nine games where you would only one and overlap of four players at most per lineup that you build. I do not mess with exposures too much, as I trust my projections and diversification methods to create ideal exposures for me. Most of the time, I will not end up with more than 35% of any one player. The exception to this would be a defenseman, where I’ll frequently have at least 80% where the top value is a defenseman on the slate.