July 2, 2016

Why Stats and Scouts Disagree on Danny DeKeyser

Danny DeKeyser is about to receive a new contract from Ken Holland and the Detroit Red Wings. Preordained as the future of the Red Wing blue line it will probably contain a hefty raise spread out over many years. Scouts and executives believe he is worthy of such a deal and that the big smooth skating defender is a new age defensive defenseman. However the statistics are not so kind to DeKeyser, some making him appear to be more qualified for the press box than the top pairing. With the Red Wings already spending $9 million on aging defenders Ericsson and Kronwall, Holland cannot afford to make another mistake with this defense.


One of the typical reasons that scouts and statisticians disagree on the assessment of a player is that scouts have difficulty properly weighing a player’s attributes. Dan Girardi or Jonathan Ericsson were believed to be good defenders because scouts see that they play physically, clear the crease and block shots, which are all positive attributes. However it is more challenging for a scout to see that these positive attributes do not compensate for, or are a manifestation of, the player’s inability to exit the zone with control and that the players have a negative impact on their teams as a result. Although there is disagreement on their assessment of him, Danny DeKeyser does not fit this typical example. Despite the fact that he possesses great size, positioning, skating and passing ability his individual possession numbers are similar to someone who belongs in the AHL.


Without a clear explanation I took a deeper look at the numbers to try and determine the cause of this disagreement.

My first potential reason was his usage. Despite being better suited as a second or third defenseman, Kronwall's decline has forced DeKeyser to take over the role as number one defenseman. He started more shifts in the defensive zone and played more minutes against the opposition's best players than any other Red Wing defender last season.


Previously it was shown that roles have a negligible impact on a player’s possession, but additional research has opened the possibility that this may not be true. Personally I think player roles do have an impact, particularly when a player is playing significantly above or below their capabilities. DeKeyser provides some evidence for this, as his numbers when playing on the 2nd pairing last year (rel. CF% = .45) are significantly higher than when he was used as the team's number one defender this season (rel. CF% = -5.26). To compound the problem for DeKeyser, he has received very little assistance to this point. The following is a list of his most common partners since joining the league.


Over seventy percent of his minutes have been spent with either Kyle Quincey, a fringe top four defender, or Jonathan Ericsson, a well-known possession anchor. Previously, the idea that these two factors negatively impact DeKeyser's possession numbers was my only defense. However when digging even deeper into his numbers I found some interesting results. 

As we know, the point of the game is to score more goals than the opposition, so ideally we would evaluate players based on their Goals For%. However goals are so rare that we are typically unable to accumulate a large enough sample for the data to accurately represent a player’s ability. Determining how much the number of goals scored can be attributed to the player and how much should be attributed to other factors such as luck is very challenging. To get around this problem we can use the number of shot attempts or Corsi For%. However this statistic treats all shots equal and there are differences in the quality of shots. Newer metrics such as expected goals for percentage (xGF%) partially accounts for this impact by weighing the shot attempts based on the likelihood of each shot becoming a goal.

When I say that DeKeyser's numbers are poor I am referring to his Corsi For %, the numbers referenced earlier are his relative CF% (The CF% when he's on the ice compared to when he is not). In agreement with the chart, I found that his relative Corsi For % during the past season was last among the Red Wing's seven defenders. However I found that his numbers improve as you focus on the most dangerous shots. Despite being last in relative CF%, he was 6th on the team in relative Fenwick For % (CF% without blocked shots), 4th in relative Shots For % (shots on goal) and 3rd in relative xGF%. To ensure this wasn’t a fluke, I included every season since he came into the league and the pattern persisted. He was 6th out of the seven in relative CF% but 4th in relative FF%, 3rd in relative SF% and 2nd in relative xGF%. This result was extremely encouraging so I widened the pool to include every defender who played at least 2000 minutes since DeKeyser came into the league.

Of the 165 defenseman who met the criteria, DeKeyser was 142nd in relative CF%, between Dennis Wideman and Jonathan Ericsson, and 132nd in relative FF%. Both of these results agree with the earlier statistical assessment that DeKeyser is a below average player. However DeKeyser was 108th in relative SF% and when sorting by relative xGF% Danny DeKeyser is 54th; just below Doughty, Letang and Burns and above players like Chara and Carlson. I believe this is the main cause of the disagreement between what we see and what the numbers say. Fans typically have a better sense of a player's xGF% than CF%, as we intuitively know what shots are more dangerous than others and will likely ignore shots we view as harmless.

UPDATE [Initially I had assumed the difference between CF% and xGF% was due to DeKeyser impacting the quality of shots the Red Wings were giving up. More digging found that it is the Red Wing's offense that is driving this difference, which I would expect DeKeyser to have less impact over, so I am still searching for an explanation.]

When doing an analysis of any player it is important to gather all possible evidence. When the numbers and scouting reports agree in their assessment of a player you should have a great deal of confidence that it is the correct assessment. However they will disagree on occasion and when this occurs we need to dig deeper and identify potential explanations. In the case of Danny DeKeyser, I believe he is a talented player and there are two reasons his numbers do not accurately represent his ability. First he has been placed in an overly demanding role and received limited assistance from his playing partners. Second, despite a low CF%, the Red Wings have been expected to score more goals when DeKeyser is on the ice then when he is off it.

These numbers do not give me complete confidence in DeKeyser’s ability to be a number one defender and carry the Red Wing defense singlehandedly. Although his xGF% is more impressive, expected goals has been shown to be less predictive of future success than CF%. However I believe the combination of factors presented here supports the idea that DeKeyser is a second or third defender and although the team should continue to look for a true number one, he has absolutely earned a long-term contract from the Detroit Red Wings.

First chart from http://ownthepuck.blogspot.ca
All other statistics and charts are from http://www.corsica.hockey

No comments:

Post a Comment