NHL Team Salary Visualization

In an earlier post I presented charts to display Blackhawks’ salaries. This was pretty popular so I took it a step further and made an interactive site. This tool can display any team for both the 2014-15 and 2015-16 seasons. The data was scraped from NHLnumbers.com. The field used is Cap Hit. A green box represents the cap space for teams below the cap. The site is hosted by github.

http://frutoper.github.io/Hockey-Viz/

Let me know if you have any questions or comments.

Chicago Blackhawks Salaries Visualized

The implosion of the 2015 Stanley Cup winning Blackhawks has begun. I don't want anything in this post to be considered a criticism of the Hawks front office. Winning 3 Cups in 6 years doesn't happen by accident. The front office has rightfully earned the trust of the fans.

The Blackhawks are currently in a very difficult financial situation. The biggest change from last year to this year is that the Kane and Toews deals kick in. The cap hit for each of them has gone from 6.3 MM to 10.5 MM. Kane and Toews became collectively 8.4 MM more expensive. Another issue is that the salary cap didn't go up at much as expected. In December 2014 The salary cap was expected to be around $73 MM. Before then it was projected to be even higher. However, the cap for the 2015-2016 is only 71.4 MM.

As a result of the above the Hawks are going to look very different next year. Below is an attempt to visualize the differences between seasons. Each box represents a player. The area of the box is the percent of the salary cap allocated to that player. You can see that the Kane and Toews boxes are become bigger. Some names have also changed. Sharp and Saad have been traded and new players added. Some current Hawks still need to sign deals.  Red boxes are forwards, yellow are defensemen, and purple are goalies.

2014-2015 salaries:

2015-2016 salaries (as of July 12):

Expect the 2015-2016 salary chart to change as players are signed or traded.

The data comes from nhlnumbers.com.  The figure used is "Cap Hit".

 Let me know if you see anything interesting.

NHL’s January Birthday Pattern Begins in Juniors

Fifteen-year-old hockey players born in January are seven and a half times more likely to be drafted to Juniors than players born in October, November, or December. This insane statistic came out of the recent Western Hockey League (WHL) Bantam Draft (recently featured on Deadspin). Of the 228 players drafted, 53 (23%) were born in January. If birth month had no effect on whether a player was drafted or not then each month would have about 19 players drafted (the red line in the graph below). Instead the results are heavily skewed toward the beginning of the year.

Malcom Gladwell popularized a theory which explains why NHL players are more likely to be born early in the year (Jan, Feb, and Mar) than later in the year (Oct, Nov, and Dec).  The theory is that players born early in the year tend to be bigger and more skilled than relatively younger player born later in the year.  This effect is largest for young players.  Fifteen-year-olds born in January are 5% older* than players born in October, November, or December.  That means they are 5% bigger, strong, faster, and theoretically 5% more experienced.  The relative age effect will be exasperated by the recent draft.  Players drafted to the WHL will have access to better coaching and better competition than players who were not drafted.  This will give them a better chance of later making the NHL.  This is only one draft in one league.  For more sides to the story check out: QuantHockey, Wired, and This Paper.

*Days alive: 365*15 = 5475

Days older if born in January:  30.5*9 = 275

275/5475 = 5%

NHL 2014-2015 Regular Season Power Rankings

Below are power ranking for the 2014-2015 NHL regular season.  The method I used was based on the Simple Ranking System outlined HERE by Football Reference.  The ranking system is based on goal differential and the goal differential of your opponents.  The New York Ranger have the best goal differential (60), but lose 3 “Goals” due to their relatively weak schedule.  The St. Louis Blues have a goal differential of 47 and gain 1.6 “Goals” for their relatively tough schedule.  Recently fivethirtyeight.com used this system to rate college basketball teams.  Just like the NHL scoring system I included a goal for a shootout win. Let me know if you think these goals should be counted or not.

Notes:

-The Central is good.  Both wildcards in the West came out of the Central so it shouldn’t be a surprise to see these teams at the top of the list. 

- LA was the best team to not make the playoffs.  This ranking system had them better than 3 playoff teams.

-Anaheim is the worst playoff team, which is surprising given their 109 points.  The Ducks average 2.78 goals per game and give up 2.70 goals per game.  Look for another early exit for Anaheim.

-Buffalo and Arizona are really bad.  The spread in goal differential is larger than points. This ranking shows how bad these teams really are.  Buffalo did have a tough schedule, but that is partially because they never got to play themselves. 

Technical Notes:

- I used 30 iterations of the system to get to the Adjusted Goal Differential.  This may have been excessive given how little changed after even 5 iterations.  More iterations might be needed in leagues like the NFL or college sports where ever team doesn't play each other

Data: http://www.nhl.com/stats/

Shot Attempts vs Ice Time: What Coaches Know That SAT Rel60 Doesn’t

Hypothesis:  Coaches should allocate ice time based on SAT Rel 60. (shot attempts for a player relative to shot attempts for team when that player is not on the ice per 60 minutes). Or, coaches should give more ice time to players with high SAT Rel-60.
Data: Scraped from nhl.com. The statistics used are: SAT Rel60, Time On Ice Per Game Played (TOI), Position, and Goals.  Only players with 30 games played and at least 10 minutes of ice time are used in the analysis.
Analysis: The first step was to plot the data.  I split the data into forwards and defense because often defensemen are asked to play more than forwards.  The expected relationship holds for forwards, but is less prevalent in defenseman. Players with high SAT Rel% tend to get more ice time.

The black line is a linear regression fit line.  Generally, forwards with 0 SAT Rel60 should get around 15 minutes per game.  Players with plus10 SAT Rel60 should play about 17 minutes (or two2 minutes more per game).   The blue lines represent four minutes above or below the estimation.  Forwards above  he top blue line are getting played too much, while players below the bottom blue line are getting played too little.

This is where the analysis falls apart:

Below is a list of forwards who are playing four or more minutes than their SAT Rel60 suggests they should.  These names are easy to recognize and have scored tons of goals this season.  These 17 players average 22 goals (~78 games into the season).  The rest of the forwards in this analysis average 12.5 goals. These 17 players have an average SAT Rel60 of -0.32.  (MPG = Minutes Per Game).

Below is a list of players the SAT Rel60 predicts should be playing more minutes.  These names don’t pop out at you like the list above.  This group has an average of five goals (~78 games).  Their average SAT Rel60 is plus 3.34.

Conclusion: Clearly coaches are making the right choice by playing guys like Kane, JVR, and Getzlaf 20 minutes a game. The results of this analysis don’t pass the eye test. Maximizing SAT Rel60 doesn’t appear to be an optimal strategy when allocating ice time.  Check out hockeyprospectus.com for a look at Shea Weber’s RelativeCorsi stats.

2013 - 2104 NHL Radar charts

Below are some summary statistics from the 2013-2014 season.  Teams are organized by regular season record. All variables are positive (more is better) except D_zone and Neutral_Zone.  These are the percentage of face offs that took place in each respective zone.  Shooting is shooting percentage and save is save percentage.  For a full glossary check out the source for this data: Stats.HockeyAnalysis.com. All variables are standardized.  Let me know if you find anything interesting.