Daily fantasy NHL is relatively new compared to other DFS sports like MLB, NFL, and NBA. The NHL is also the least popular of the four major sports leagues. These two factors contribute to the large knowledge gap between NHL DFS and the more popular sports.

Stacking is an important tactic in guaranteed prize pools. In the more popular DFS sports, the benefits and uses of stacking (for the most part) are known. In football, for example, we know stacking a quarterback with one or more of his receivers can result in amplified production through the passing game. Stacking a team’s running back or kick returner with his defense/special teams is less common but also viable. But NHL stacking is largely uncharted territory.

It’s common for DFS players to stack two or three players on the same even strength line, but some DFS players seek correlation more creatively by stacking multiple forwards and a defenseman on the same power play or two defensemen on the same even strength line. NHL stacking is complex because the game is so fluid. By analyzing past games and how lines and positions correlate with each other we can better understand which stacks are most likely to be productive in the future.

The Data

Here’s some information on the correlation matrix below. Our data consists of 549 team samples for games played on individual dates in 2016. Our sample isn’t massive, but it is large enough to provide insight into how positions correlate. All data is based on DraftKings scoring and consists of only skaters.

For simplicity, we used the four ‘forward line’/three ‘defensive pair’ structure deployed by most teams. A small number of teams were missing data for one or two positions on specific nights. Those empty data points were not factored into our correlation numbers.

To designate between wingers on the same line and defensemen in the same pair we used each player’s DK salary at the time of the game. The more expensive of the two is labeled ‘a’ and the cheaper ‘b.’ For example, a second-line winger who was more expensive than the other wing would be labeled ‘2Wa.’ All line designations refer to even strength lines.

Power play lines were not factored into the data, but I intend to sort through that data in the near future.

The Matrix

I’ll let you draw your own conclusions from the matrix, but there are a few points I want to highlight:

  • Right away we can see how fluid hockey is. We don’t have many significant positive correlations.
  • If two players don’t play at the same time at even strength they wont correlate highly. Playing two or more forwards on separate even strength lines doesn’t make much sense.
  • The highest correlations come from forwards on the same lines. Most of you probably predicted that.
  • For the most part, centers correlate with their wingers more than same-line wingers correlate with each other. If you’re looking for a two-man forward stack, playing the center with one of his wingers probably makes the most sense.
  • Rostering two defensemen on the same even strength line doesn’t result in much value. While they do tend to correlate positively, the correlation is not significant.
  • The 1C-1Da pairing has the matrix’s 11th-highest correlation while the 1C-1Db pairing has the 10th-lowest. On the surface that doesn’t make much sense, but the higher-priced defenseman is likelier to be on the top power play line more often than his cheaper line mate. The difference in correlation with the first-line center might have to do with power play opportunity.
  • While there are certainly situations in which stacking defensemen with forwards makes sense, the majority of the time forward-only stacks offer the most production.

In the near future I’ll be looking into how power play lines affect correlations as well as how players correlate with their opponents, which should give us an edge when considering game stacks.

For now, keep these historical correlations in mind when building tournament rosters via the Lineup Builder in our Player Models. Also, it would be worthwhile to experiment with the Trends tool to see what you can learn in terms of correlations and Plus/Minus.