Are Advanced Stats Smarter Than Mike Babcock?

When you ask an NHL coach about why they switch up their lines so frequently, the answer is rarely based on statistics. Coaches cite finding chemistry, promoting cohesiveness, and igniting a spark as the major reasons why they change up their lines. Finding the right line combinations can be like trying to guess the number of gumballs in a glass fishbowl. Coaches often mix and match forwards, seldom having patience with each combination. Red Wings’ coach Mike Babcock is especially notorious for frequently mixing up his line combinations. Should we evaluate lines based on these measures listed by Babcock and other coaches or is there a statistical measure that proves better?

On the surface, the answer seems simple. Look at how many goals that line scores and look at how many they give up. Digging a little deeper, one can track the number of shots the line produces against the number the line gives up. However, in recent years, Corsi For% has become increasingly popular in measuring an individual player’s impact on the game. For those unfamiliar with Corsi For%, it is the number of shot attempts (shots on goal + blocked shots + missed shots) a team generates when a particular player is on the ice divided by the total number of shot attempts that occur while the player is on the ice. A percentage above 50 indicates that more than 50% of the shots taken when a player is on the ice are in the offensive zone. However, this statistic is currently only being tracked for individual players and for the team as a whole. While both those numbers are informative, it does not provide information on how a particular line combination works together. I decided to take a look at the Detroit Red Wings’ line combinations to see which combinations had the best Corsi For%. With Corsi For% being a surrogate for “puck possession”, we can essentially determine which lines controlled the puck the best for Detroit. Coach Babcock, you were looking for information….here it is.

(Brace Hemmelgarn-USA TODAY Sports)
Mike Babcock has requested more information on advanced stats. Can they aid him in picking the correct line combinations? (Brace Hemmelgarn-USA TODAY Sports)

Corsi For% Study Parameters

The study was started by first going to Extra Skater, and reviewing the game log for each game in the 2013-2014 season. From the Corsi Events section, the period, time, situation, Corsi event, and team generating the Corsi event were charted. From there, the NHL play-by-play event recaps were examined for each game. From those recaps, the players on the ice for each event were copied down. After obtaining all the data, the data was pared down to focus only on 5-on-5 Corsi events that occurred when three Red Wings’ forwards were on the ice. After matching all the forwards with the events, each line combination was copied down into a separate Excel sheet where the number of events for each line combination was tracked.

There are a few limitations to this study. First off, no analysis of quality of competition or usage was made. There currently are not measures of quality of competition or usage for  line combinations. This information would have allowed for a better analysis of the Corsi events. In addition, there was no analysis made in terms of time on ice spent. Again, this information is currently not readily available but would have allowed for an examination of the ratio between Corsi events and ice time. This could provide information on how different line combinations generate Corsi events per 60 minutes.

What are the Optimal Line Combinations?

The Detroit Red Wings employed 480 forward line combinations that were on the ice for 6,457 5-on-5 Corsi events. Below you will see a chart showing every line combination that was on the ice for a minimum of 50 Corsi events. The minimum of 50 was selected to ensure that we have an adequate sample size for that line. A typical line is involved in four to eight Corsi events a game so a minimum of 50 events gives us roughly six to ten games of data for that line.

Forward 1 Forward 2 Forward 3 5-on-5 Corsi For 5-on-5 Corsi Against 5-on-5 Total Corsi Events 5-on-5 Corsi For%
Andersson Abdelkader Miller 41 16 57 71.9
Nyquist Andersson Tatar 33 20 53 62.3
Datsyuk Alfredsson Abdelkader 37 25 62 59.7
Sheahan Tatar Bertuzzi 34 23 57 59.6
Sheahan Tatar Jurco 222 151 373 59.5
Nyquist Sheahan Tatar 90 63 153 58.8
Datsyuk Alfredsson Zetterberg 41 29 70 58.6
Helm Alfredsson Cleary 32 23 55 58.2
Glendening Helm Miller 43 31 74 58.1
Nyquist Legwand Franzen 108 85 193 56.0
Andersson Tatar Abdelkader 43 34 77 55.8
Datsyuk Abdelkader Zetterberg 89 71 160 55.6
Datsyuk Bertuzzi Zetterberg 157 126 283 55.5
Glendening Eaves Miller 48 39 87 55.2
Nyquist Abdelkader Zetterberg 157 128 285 55.1
Nyquist Alfredsson Zetterberg 28 24 52 53.8
Legwand Glendening Miller 30 26 56 53.6
Andersson Tatar Miller 86 77 163 52.8
Nyquist Franzen Zetterberg 64 58 122 52.5
Weiss Franzen Alfredsson 31 28 59 52.5
Datsyuk Franzen Abdelkader 27 25 52 51.9
Andersson Helm Alfredsson 34 32 66 51.5
Legwand Franzen Alfredsson 54 60 114 47.4
Helm Alfredsson Abdelkader 65 75 140 46.4
Glendening Bertuzzi Miller 47 56 103 45.6
Glendening Eaves Cleary 89 107 196 45.4
Nyquist Franzen Bertuzzi 23 28 51 45.1
Andersson Bertuzzi Cleary 37 47 84 44.0
Andersson Alfredsson Cleary 24 31 55 43.6
Weiss Alfredsson Cleary 37 49 86 43.0
Weiss Franzen Abdelkader 40 60 100 40.0
Glendening Abdelkader Miller 20 37 57 35.1

 

There are a couple of important things to consider with this table. First off, the Red Wings’ team average for 5-on-5 Corsi For% was 51.4%. That means that of the 32 line combinations that qualified for this study, 22 of them were above the team average. However, if you increase the minimum amount of Corsi events to 100, five of the 13 line combinations that would qualify have a Corsi For% of 47.4% or less. That statistic is a bit troubling. The ability to identify unproductive lines is a necessity for a coach. Tracking the Corsi For% for line combinations would have allowed Babcock and his staff to identify and avoid these five unproductive line combinations. For example, we know that Stephen Weiss received a lot of flak this past season for his play when he was healthy. This table shows how his utilization could have affected his play. Weiss’ name appears on this list three times. When paired with Johan Franzen and Daniel Alfredsson, their combination posted an above average 52.5% Corsi For. This should be taken with a grain of salt as their line combination barely met the cutoff of 50 Corsi events, but still it is a sign that maybe that combination should have been utilized more.

On the flip side, one piece of information that stood out to me was how well the “Kids Line” played. The line consisting of Riley Sheahan, Tomas Tatar, and either Gustav Nyquist or Tomas Jurco was quietly one of the most dominant lines for the team. When the line utilized Jurco, they were involved in a team-high 373 Corsi events, but came away with the 5th highest Corsi For% in this study at a fantastic 59.5%. If Jurco was substituted out for Nyquist, the line still performed at an outstanding level, controlling 58.8% of the Corsi events. People have constantly said that the future of Detroit’s playoff streak will be determined by how quickly the kids grow up. This data suggests that they are growing up real fast. Both “Kids Lines” performed better than any line combination that involved captain Henrik Zetterberg. As the 2014-2015 season approaches, coach Babcock should feel more comfortable relying on the kids for big minutes. This will allow him to conserve both Zetterberg and Pavel Datsyuk a little bit more throughout the season to keep them injury-free and fresher for the playoffs. However, some of you may be reluctant to turn the reigns over to the kids, citing that while they do control the play, they don’t put goals on the board at the same rate as Datsyuk and Zetterberg.

The "Kids Line" was phenomenal last season in terms of puck possession. Can they keep it up? (Rick Osentoski-USA TODAY Sports)
The “Kids Line” was phenomenal last season in terms of puck possession. Can they keep it up? (Rick Osentoski-USA TODAY Sports)

Which Line Combinations Produce the Most Goals?

The Red Wings as a team gave up 145 goals at even strength, with one of those goals coming when the Wings only had two forwards on the ice. Below you will see a table showing every line combination that was on the ice for at least five 5-on-5 goals. The number five was selected in order to ensure that the line was together long enough to make an adequate assumption about their performance.

Forward 1 Forward 2 Forward 3 5-on-5 Goals For 5-on-5 Goals Against 5-on-5 Total Goals on Ice For 5-on-5 Goals For%
Nyquist Sheahan Tatar 8 0 8 100.0
Nyquist Franzen Zetterberg 5 1 6 83.3
Nyquist Abdelkader Zetterberg 10 4 14 71.4
Sheahan Tatar Jurco 12 5 17 70.6
Helm Alfredsson Abdelkader 4 2 6 66.7
Datsyuk Alfredsson Zetterberg 4 2 6 66.7
Andersson Tatar Miller 5 3 8 62.5
Datsyuk Bertuzzi Zetterberg 11 7 18 61.1
Datsyuk Abdelkader Zetterberg 3 2 5 60.0
Andersson Tatar Abdelkader 3 2 5 60.0
Nyquist Legwand Franzen 7 7 14 50.0
Legwand Franzen Alfredsson 2 3 5 40.0
Andersson Helm Alfredsson 2 3 5 40.0
Andersson Bertuzzi Cleary 2 3 5 40.0

 

Fourteen lines met the five-goal minimum and they are sorted by Goals For%. We can pick up some very important information from this table. We can see that the kids line of Sheahan, Tatar, and Jurco scored the most goals of any line combination, narrowly edging Datsyuk, Zetterberg, and Todd Bertuzzi. However, the most impressive statistic comes from the Nyquist, Tatar, and Sheahan combination. The combination scored eight goals on 90 Corsi For attempts and yielded 0 goals against on 63 Corsi attempts against. Not one goal was scored against the Nyquist-Tatar-Sheahan combination. Talk about impressive. As we can see, the kids are more than capable of picking up the slack both offensively and defensively and Babcock should have no reservations about decreasing the minutes for Datsyuk and Zetterberg ever so slightly so that they can be fresher for the playoffs.

Projecting the 2014-2015 Line Combinations

After collecting all of this information, the best thing we can do is use it to figure out possible line combinations for next year. Gone are David Legwand, Todd Bertuzzi, Mikael Samuelsson, Cory Emmerton, Jordin Tootoo, and Patrick Eaves. When projecting these line combinations, I will project them with and without Daniel Alfredsson. I will also project these line combinations assuming that Danny Cleary makes the team and that Tomas Jurco will have to be sent down to Grand Rapids if Alfredsson stays. I will not include Anthony Mantha as I currently have no data to project his possible line partners.

Whether or not Daniel Alfredsson returns, the Wings have plenty of options up front (Marc DesRosiers-USA TODAY Sports)
Whether or not Daniel Alfredsson returns, the Mike Babcock and the Wings have plenty of options up front (Marc DesRosiers-USA TODAY Sports)

With Daniel Alfredsson

Forward 1 Forward 2 Forward 3 2013-2014 Corsi For%
Line 1 Pavel Datsyuk Henrik Zetterberg Justin Abdelkader 55.6%
Line 2 Riley Sheahan Tomas Tatar Gustav Nyquist 58.8%
Line 3 Stephen Weiss Daniel Alfredsson Johan Franzen 52.5%
Line 4 Darren Helm Drew Miller Danny Cleary/Luke Glendening N.A./58.1%

 

Every line combination here had a Corsi For% greater than 52.5%. Helm, Miller, and Cleary never played together last season, but the combination of Helm, Glendening, and Miller had a Corsi For% of 58.1%. If Glendening eventually takes the full-time spot away from Cleary, the Wings will have four lethal possession lines.

Without Daniel Alfredsson

Forward 1 Forward 2 Forward 3 2013-2014 Corsi For%
Line 1 Pavel Datsyuk Henrik Zetterberg Justin Abdelkader 55.6%
Line 2 Riley Sheahan Tomas Tatar Tomas Jurco 59.5%
Line 3 Gustav Nyquist Stephen Weiss Johan Franzen 52.6%*
Line 4 Darren Helm Drew Miller Danny Cleary/Luke Glendening N.A./58.1%

*based on 19 Corsi Events

If Daniel Alfredsson does decide to retire, the Wings still have four great line combinations. Tomas Jurco can step in and take Nyquist’s spot on the “Kid Line”. Nyquist can slide down and play with Weiss and Franzen. That combination was on the ice for 19 Corsi events which isn’t a significant number, but the line did record a goal together. The first and fourth lines don’t change to give the Wings four solid possession lines. Coaches change line combinations more frequently than you change your underwear. However, a majority of the time, the change is not based on any sort of statistical backing. In the case of the Detroit Red Wings, the use of Corsi For% can aid Babcock in his staff in the identification of strong line combinations. To take this to the next step, teams can begin to track situational Corsi For% based on time remaining in the game or what the score is to determine which liens perform best in these situations. Imagine the wealth of information that could be presented to the coach. We are only scraping the surface when it comes to the usage of Corsi and line combinations is one place where coaches should start utilizing it more often.

 

For anyone interested in seeing the full Excel spreadsheet of all line combinations, please contact me via Twitter or email.

9 thoughts on “Are Advanced Stats Smarter Than Mike Babcock?”

  1. Thanks for putting this together. It really begs the question – would DRW be more productive if the lines were kept together longer? Sure seems that younger, faster players would benefit from ‘knowing’ their line mates better. We may not find out until a coaching change.

  2. Prashanth, I’m impressed! You certainly are punishing Franzen, but that stats don’t lie.

  3. Top Corsi line was Anderson, Abdelkader, Miller. Worst was Glendenning, Abdelkader, Miller. Obviously, Anderson was Detroit’s best player, or Glendenning the worst.

  4. I’m pretty sure this can be assumed and is such a minimal error, but should it be Tomas Tatar and not Jurco on the second line where the scenario is that Alfredsson stays?

    Fantastic read nonetheless.

    • Hi Michael. You are absolutely correct. My mistake, thanks for catching it. I will update the post to reflect the change. Thanks for the read.

    • Hey Johnny,

      By the parameters I looked at, the Kid Line was fantastic. However, all the information here should be taken with a grain of salt. I did not have a way to measure quality of competition that each of the lines faced nor did I have the ability to get the time on ice for each of these lines. Without that information we can’t make a conclusive statement that the Kid Line was the best, but the data here definitely suggests that they should receive more playing time together next year.

Comments are closed.