FanPost

Data Mining for D-Men

This is a long post and I apologize for that right up front. I showed the draft to JJ and he suggested it needed to be a mega-post because people will want to know how I came up with my conclusions, so here it is at nearly 2,000 words. I suspect the interpretation of data and my conclusions are going to be controversial, but I wanted to put it out here because WIIM contributes to the zeitgeist that Ken Holland operates in, and maybe our thoughts can have a positive effect. Besides, I did the work so what the heck....

There seems to a consensus that the Wings need help on the back end. Most of the commentary seems focused on getting one new impact player. My analysis says they need two more top-four defensemen, and one of them should be a strong #1. The conclusion comes from about 20 hours of deep diving into data from nhl.com with an assist from http://stats.hockeyanalysis.com. Digging into the data was my way of coping with the end of the Wings’ season last Saturday.

The following is a summary of my findings about Detroit’s defensive corps and what may be needed to return to a top 10 group. It’s an attempt to be objective; I did not start with any preconceptions and the calculations were completed before I started interpreting the results.

Methods

I started with data for all NHL players who played at least 200 minutes during the regular season 2014. That was about 250 players.

It can be really hard to compare the numerical apples and oranges of statistics such as time-on-ice, goals and shot differential, so I used "non-parametric" statistics based on rank order.

Players were rated on three major attributes using several equal factors to calculate each rank:

Ironman attribute: factors include total time played in season, time on ice per game (TOI/G) even strength, TOI/G power play, TOI/G penalty kill

Offensive potency attribute: factors included shots/minute, shot percentage, goals/min 5 on 5, goals/min power play, assists/min 5 on 5, assists/min power play, other uncredited scores/min while on ice, adjusted relative offensive Fenwick/min (takes into account the weighted defensive fenwick of players he faced).

Defensive potency attribute: factors included even strength goals against/min, power play goals against/min, adjusted relative defensive Fenwick/min (takes into account the offensive Fenwick of players he faced), penalty minutes per minute of playing time (this was ranked so that a good ranking comes from few penalty minutes per unit of playing time).

Each factor was ranked in isolation, then the average rank of factors was calculated for each attribute; the average values were re-ranked to assign the overall rank for the attribute.

A "Playing Rank" was calculated to represent the "rank of ranks" of the Offensive and Defensive attributes. This indicates how well the player played without considering how much time he was able to play.

A "Grand Rank" represents the "rank of ranks" of the average of the Playing Rank and the Ironman Rank. It corresponds better to our perception of player value because we notice the guys who are always on ice.

I gave up an attempt to adjust rankings to take into account strength of goaltending, because the adjustments began to make it apparent that goalies are as much victims of bad D men as D men are victims of bad goalies. The adjusted results seemed so paradoxical that you would have to conclude nobody knows who the good defensemen are, so I decided it is best to leave goalies alone.

The methods I chose were tailored to help identify players who were underperforming against the level of competition they faced and also to identify players who are not getting enough playing time according to their abilities (for example because they have other great D men on their team).

For a reality check, the top 15 players in Grand Rank were Andrei Markov, Alex Pietrangelo, Jay Bouwmeester, John Carlson, Ryan McDonagh, Oliver Ekman-Larsson, Cam Fowler, Erik Johnson, Andrej Sekera, Ryan Suter, Duncan Keith, Roman Josi, Drew Doughty, Zdeno Chara, and Alex Goligoski.

These seem fairly realistic rankings to me. They don’t correspond to Norris Trophy votes, which seem to be heavily weighted by minutes played and offensive production. Norris candidates also are nominated by people who vote according to perceptions formed over years, while my calculations are just for 2014.

The top 7 players from each team were identified using Grand Rank. Teams were then ranked according to the average Playing Rank of their top 7 players. (Grand Rank includes a lot of ice-time weighting and that just makes team ranks cluster together more; by isolating Offensive and Defensive potency alone you get a clearer picture of the potential strength of a defensive corps.)

Results

Here are the best and worst team defense corps:

Wrjcj9_medium

Yes, without pre-conceived weigths or adjustments, Detroit’s defensive corps came out last in the NHL. After seeing the result I tried a variety of alternative methods to discover if there were mistakes or biases but several different versions of the analysis never got them above 27th out of 30.

To see where Detroit’s defensive corps problems are, its players were sorted by Grand Rank and assigned a rotation position number from 1 to 7. The relative rank of each player is compared to an expected rank for a hypothetical average player of the same position in the defensive rotation. For example, the average rank of #1 players is about 15 because there are 30 teams and therefore 30 #1 players who would be ranked from 1 to 30. If a player in a #1 position had a Grand Rank of 9, then he would rate +6 spots compared to the expected rank of 15. (The precise average rank used for comparison depends on the number of ties in the ranking calculation, but I don't care too much about decimal points here). For reference the expected rank of #2 players is about 45, #3 players = 75, # 4= 105, #5 =135, #6=165 and #7 = 195.

Here are the results:

Jpu4hc_medium

Some of the biggest issues are highlighted in the table. Niklas Kronwall ranks 23rd overall in the league but that rank was driven by his Ironman Rank. His Playing Rank is far below his position and awful for a #1; it would be appropriate for an average #4 player. The low playing rank was driven by his Defensive Rank where he was among the worst in the league. Opposition players who stepped on the ice against Kronwall played better against him than they did against their average #1 opponent. His performance would be helped by letting him settle to his level of competence; the Wings are putting too much of a burden on him.

All of the top six D-men for the Wings played below position; on average they played about 16 rank levels below the expected rank. Kyle Quincey’s Playing Rank was dead last among the 210 top-7 players on the thirty teams. His Grand Rank was 121 but that was achieved by being on ice for #3 minutes, and so his Ironman Ranking is equal to a #2 player. In my opinion his current salary is too high for that level of performance and as a UFA I suspect he will want/get even more. He probably should go because solid #6 or #7 defensemen are a lot cheaper than he is.

If you can believe the results, this table shows that the team needs to pick up two top-four D-men. Of course, they could hope that Jonathan Ericsson will improve next year. He missed a lot of time and that hurt his Iron Man rank, but his Playing Rank of 204th does not include ice time, and it says he really did have a poor year. With good health, perhaps he can jump up significantly.

Another possibility is that one of the rookies will jump in like Danny DeKeyser did this year. With a Grand Rank in his first full year of 55, DeKeyser is a weak 2/strong 3 and I suspect he will continue to get much better. I wouldn’t be surprised if he moves past Kronwall next year. I won’t try to analyze AHL stats to figure out who that great rookie might be- I’m done digging for a while.

I took a look at this year’s UFA class and decided to pick two players not over 30 who could really help the overall D-corps rank. I chose Matt Niskanen of Pittsburgh and Andre Benoit of Colorado. The latter might surprise you but on the Wings his scoring would have been second and his +/- also would be second. In 2014 he was a strong #3 who might be a big boost as a 3/4 for the Wings. Another good choice might be Anton Stralman of the Rangers who plays as a solid #4, is 27 years old and a right-hander. The best of the old-timers are Markov (age 35) and Boyle (age 37); both are top #1 players (Markov’s Grand Rank rank is 1 and Playing Rank is 8; Boyle’s Grand Rank is 28 and Playing Rank is 4).

Here is a hypothetical 2015 table for the Wings, plugging in the numbers for Niskanen and Benoit, letting Quincey and Lashoff find work elsewhere:

Htuwiq_medium

The result is that by adding two players of this level the Wings would go from dead last in the league (in terms of overall strength of D-Corps) to a spot in the top 10. Getting a solid #4 is just as important as a #1. With this lineup I think Kronwall would eventually fall back to his level at #3.

Of course, it isn't this simple in real life. Niskanen and Benoit probably would play differently in Detroit. But Niskanen’s right-hand shot would make Detroit’s D-corps better overall so his rating might be even higher. And Benoit’s experience and solid play would surely be a big improvement over any of the #4 through #7 Detroit D-men.

Just in case this all seems a bit too much like "playing with numbers," here are tables comparing the conventional stats of the Wings’ top three pairs five years ago to those of this year. The striking thing to me is that Kronwall has very similar numbers five years apart, yet in 2009 he was a #3 and this year he was a #1. And look at how much stronger both Lidstrom and Rafalski were than anybody on this year’s Wings. I think this supports the idea that the team really is short two top four defensemen.

27xkkmd_medium

Of course, I’m aware of the deficiencies of this approach and I’m not suggesting that all of the Wing’s problems can be solved by plunging into the free-agent market. One of the biggest issues I see is that many of the best options are players on teams that are currently stronger than the Wings. That must play a strong role in their decisions, and prying Niskanen from the Penguins or Benoit from Colorado or Stralman from the Rangers will be tough. Does Boyle really want to leave the Sharks for the Wings? Would Markov leave Montreal for Detroit?

I’m also aware that other factors can have a huge impact on a D-Corps. Getting bigger, getting a couple of right-handers, and getting some snarl could help a lot too. Also getting some calm, veteran experience would help the rookies that I hope will make next year’s team. But I do believe that data can provide a strong foundation for decision-making. Anyway, I’m just a data digger and don’t know the art of deal-making, but I do know that when the Wings make a deal I’ll be checking to see whether it looks like an improvement.

As a final note, here are the top UFA D-Men ranked according to my analysis:

25qrz9k_medium

This is a fanpost written by a WIIM community member. The views and opinions expressed here are that member's and do not necessarily reflect the views of the site itself.

X
Log In Sign Up

forgot?
Log In Sign Up

Forgot password?

We'll email you a reset link.

If you signed up using a 3rd party account like Facebook or Twitter, please login with it instead.

Forgot password?

Try another email?

Almost done,

By becoming a registered user, you are also agreeing to our Terms and confirming that you have read our Privacy Policy.

Join Winging It In Motown

You must be a member of Winging It In Motown to participate.

We have our own Community Guidelines at Winging It In Motown. You should read them.

Join Winging It In Motown

You must be a member of Winging It In Motown to participate.

We have our own Community Guidelines at Winging It In Motown. You should read them.

Spinner.vc97ec6e

Authenticating

Great!

Choose an available username to complete sign up.

In order to provide our users with a better overall experience, we ask for more information from Facebook when using it to login so that we can learn more about our audience and provide you with the best possible experience. We do not store specific user data and the sharing of it is not required to login with Facebook.

tracking_pixel_9355_tracker