Here is a study that I did for Hockey Prospectus on NHL Team Architecture. Obviously since I posted it on HP, I believe they have a great site, so definitely take a look at some of the material from their other great writers too.
Front Office Focus – NHL Team Architecture
What is the best way to build a winning NHL team? There probably isn’t just one method, but every team in the league would love to know who to spend their money on, how to find bargains that outperform their salaries and which positions seem to matter most to points in the standings.
Tom Awad has some new info on the best players and Robert Vollman had some great insight for building teams in the salary cap era. What I hope to uncover is how spending and performance at different positions affects a team’s position in the final standings.
In order to gain some insight into this dilemma, I accumulated information for every team in the NHL over the last 3 seasons, which gives us 90 team seasons as data points. While this is still a small sample, at least it’s relevant in terms of today’s salary cap climate and represents the summation of 7380 team games played.
For each team I combined wins, losses, overtime losses, total points, goal differential, +/- rates for even strength, 5-on-4, 4-on-5 and a variety of positional data.
The positional data used was the cap space allocated to each team’s top 6 forwards, top 4 defensemen and goaltenders. I also included the combined GVT for those top-paid positional players as well as a combined GVT for the best-performing players on that team. Obviously, the highest paid players aren’t always the best performers on the team, so I wanted to capture both.
First and foremost, I was hoping to understand if paying higher salaries at any position helps rack up points in the standings.. I also want to know if top performers at different positions have a greater impact on wins or point totals. Are teams better off spending on a top goaltender, defensive core or elite forwards?
What Factors Matter to Winning?
Using regression analysis, the only statistically significant factor contributing to an NHL team’s point total was Goal Differential (GD). This makes a ton of sense to anyone who has read up on Pythagorean Projections. It also appeals to common sense. Teams that outscore their opponents by the most goals tend to finish highest in the standings. I’m sure you’re wondering why I needed statistics to prove that one. Call me Captain Obvious.
The more significant finding is that none of the positional data factored in. However, power play and penalty killing effectiveness definitely had an impact but to a much lesser extent that GD.
That said, when using the same regression analysis to determine the biggest contributing factors to Goal Differential, there is more insight to be gained. Rather unsurprisingly, the GVT Totals of the positional groups (fwds, def & goalies) is a significant factor to GD, which again makes perfect sense since GVT is a tallied in terms of goals. However, the cap money spent on each position has no correlation.
There also doesn’t appear to be any significant difference in the impact from different positions. The regression coefficient for each position is very similar and their confidence intervals don’t show any statistical difference. In plain terms, defensemen as a whole didn’t matter more than forwards or vice versa.
Essentially, it all boils down to performance and the money spent has no impact unless it’s related to performance. This brings us to the concept of performance at each position and more importantly performance versus salary.
Performance can be demonstrated by looking at both the GVT of the top-paid players on each team but also at purely the best performing players at each position. The reason for evaluating both is to see whether teams are getting good production from the players they identify as their best (by the contracts handed out) but also whether or not they’re finding hidden talent or trading for good players mid-season.
The table below shows the rough demographics of GVT by position:
| Avg GVT for Group | Std Dev | Min | Max | |
| Top 6 Salary Cap Fwds | 52.3 | 17.6 | 10.9 | 117.5 |
| Top 6 GVT Fwd | 64.8 | 17.5 | 27.2 | 120.5 |
| Top 4 Salary Cap Def | 24.2 | 10.5 | -2.1 | 70.2 |
| Top 4 GVT Def | 33.2 | 9.8 | 8.4 | 70.2 |
| Top Cap Goalie | 7.3 | 12.5 | -19 | 35.3 |
| Alternate GVT Goalie | 4.0 | 9.8 | -13.3 | 37.7 |
For clarity’s sake, the “Top 6 Salary Cap Fwds” is the combined GVT of every team’s six highest paid forwards. The “Top 6 GVT Fwds” is the combined GVT of every team’s six highest GVT-generating forwards. The same is true for defensemen.
For goaltenders, a slightly different method was used. The “Top Cap Goalie” is the ‘keeper who was paid the most at the start of the season. The “Alternate GVT Goalie” was the goaltender who played the most games outside of the Top Cap Goalie.
The teams that had the most significant performance from their highest paid players at each position are below:
| Best Top 6 Salary Cap Fwds | 3-yr Avg GVT for Group |
| Washington Capitals | 82.0 |
| Detroit Red Wings | 74.2 |
| Pittsburgh Penguins | 70.7 |
| San Jose Sharks | 67.6 |
| Vancouver Canucks | 67.4 |
| New Jersey Devils | 66.5 |
| Ottawa Senators | 65.2 |
| Best Top 4 Salary Cap Def | 3-yr Avg GVT for Group |
| Detroit Red Wings | 55.5 |
| San Jose Sharks | 34.7 |
| Best Top Cap Goalies | 3-yr Avg GVT for Group |
| Florida Panthers | 26.6 |
| New York Rangers | 21.7 |
| Buffalo Sabres | 20.1 |
| Vancouver Canucks | 19.8 |
| New Jersey Devils | 19.0 |
The most notable thing in these tables is how few teams stood out from the crowd with defense performance. Very few teams were more than one standard deviation from the mean. We could conclude that either the talent distribution of players on defense is very tight or the talent is spread quite equally throughout the league.
The goaltender table also illustrates the ridiculously impressive impact that Tomas Vokoun has had for Florida. No other team was within 5 GVT of the Panthers and the drop-off after this 1st tier of goaltending is steep and quick. The Toronto Maple Leafs, for instance, had a 3-year average GVT for their top-paid goaltender of -16.2!
Using GVS to Create More Clarity
Robert Vollman introduced the statistic of Goals Versus Salary in a column last summer. I have frequently used GVT per $M of salary as a measure of value. Now that I’ve had time to compare both, I’m much happier with GVS as it is a much more intuitive number and follows the spirit of GVT much closer considering that it compares to threshold output for the same salary.
With that in mind, I decided to use GVS to further evaluate the architecture of the NHL’s teams over the last 3 seasons.
GVS for all 3 positions correlated strongly with Goal Differential for the team in the regression analysis.
At the positional level, here is how each team’s top paid players faired versus the cap hit they took up.
| Avg GVS for Group | Std Dev | Min | Max | |
| Top 6 Salary Cap Fwds | -3.8 | 17.9 | -36.2 | 57.8 |
| Top 4 Salary Cap Def | -10.3 | 10.7 | -35.7 | 22.8 |
| Top Cap Goalie | 4.6 | 12.0 | -22.0 | 30.8 |
| Alternate GVT Goalie | 9.9 | 9.8 | -11.0 | 43.4 |
It’s not surprising to see that the average top-paid player in the NHL has trouble living up to his salary. With higher dollars come much higher expectations. It’s also easier to under-perform a large contract than a smaller one.
What was surprising, however, was the gap between defense and forwards. Either GVT is undervaluing defensemen, or GM’s are spending too much on defenders for the value they get. Given that the total for forwards accounts for six players and the total for defensemen only accounts for four, meaning that the GVS per individual defenseman was 4 times that of an individual forward.
That said, it’s surprising to see how many goalies outperform their salaries. The average cost for both the top-paid goalie and his alternate over the last 3 years was $4.9m, which was divided into $3.9m for the expected starter and $1m for the backup. At this wage, it’s much easier to outperform. That said, the Alternate Goalie is a position of tremendous upside for many a team.
Just in case you’re curious, here are the best and worst teams by GVS over the 3 year period at each position.
| Best Top 6 Salary Cap Fwds | 3-yr Total GVS |
| Washington Capitals | 71.9 |
| Detroit Red Wings | 51.8 |
| Pittsburgh Penguins | 34.7 |
| Vancouver Canucks | 29.4 |
| Phoenix Coyotes | 25.2 |
| Worst Top 6 Salary Cap Fwds | 3-yr Total GVS |
| New York Rangers | -66.1 |
| Edmonton Oilers | -62.7 |
| New York Islanders | -61.0 |
| Dallas Stars | -60.9 |
| Florida Panthers | -49.2 |
The Phoenix Coyotes are the most pleasant surprise in the top of the forward list. Phoenix didn’t make the list the same way the other teams did. The Capitals, Wings, Penguins and Canucks did it by signing powerhouse forwards to decent contracts and not overpaying for good complementary talent. Of course, they were bad enough in previous seasons that they got to draft those powerhouse forwards in the first place. The Coyotes did it by stacking two lines with low-cost talent getting decent, but unspectacular performance from those players.
| Best Top 4 Salary Cap Def | 3-yr Total GVS |
| Nashville Predators | 14.5 |
| Detroit Red Wings | 9.4 |
| Dallas Stars | -0.3 |
| Washington Capitals | -2.2 |
| San Jose Sharks | -2.7 |
| Worst Top 4 Salary Cap Def | 3-yr Total GVS |
| St. Louis Blues | -69.0 |
| Calgary Flames | -68.8 |
| Edmonton Oilers | -67.9 |
| Philadelphia Flyers | -66.6 |
| Atlanta Thrashers | -58.8 |
On defense it’s a very different story. The Nashville Predators topped the charts with the same model Phoenix uses for forwards. They spent frugally on their top 4 blueliners and got above-average production compared to other teams.
On the bottom end of the table, the Blues spent average cap space on their top 4 defensemen, but got very poor performance. The Flames on the other hand spent 2 standard deviations higher than the average team on their top 4 defensemen the last two seasons, and got slightly below average production in return. No wonder no one in Calgary mourned the departure of Dion Phaneuf. The Flyers could be put in the same category as the Flames.
| Best Goalie Tandem | 3-yr Total GVS |
| Florida Panthers | 103.2 |
| Montreal Canadiens | 96.8 |
| Boston Bruins | 90.7 |
| Phoenix Coyotes | 78.0 |
| New Jersey Devils | 72.5 |
| Worst Goalie Tandem | 3-yr Total GVS |
| Toronto Maple Leafs | -43.0 |
| Chicago Blackhawks | -2.5 |
| Dallas Stars | -2.2 |
| New York Islanders | -1.9 |
| Ottawa Senators | 1.6 |
The Tomas Vokoun factor is again apparent, but 45% of the GVS for Florida comes from the backup goalie. Craig Anderson was a huge impact on Florida’s goaltending success at an absolute steal with only $600k in cap money.
The Bruins on the other hand got their entire GVS benefit from the “Backup” role. Tim Thomas for two seasons and Tuukka Rask last season outperformed their higher-paid counterparts.
The Maple Leafs’ woes in net dwarf those of any other team in the league. The incredible deficit shown on the table is either due to a stimulus package or just two horrendous seasons by Vesa Toskala. Come to think of it, they could both be considered as throwing money blindly at a problem.
Conclusions
In the end, there’s no new discovery in relation to how a team finishes the regular season. The more a team can outscore the opposition, the more points they’ll rack up. Brilliant!
Also there isn’t a specific link between money spent on any given position relating to points in the standings. However, the performance (measured by GVT) of a team’s top players at each position, all contributed equally and directly to a team’s goal differential.
Through the eyes of GVS, it’s more likely for top forwards to outperform their contracts than top defensemen. In fact, only two teams outperformed the contracts they handed out on the blueline over the last 3 seasons. Given that a lot of large contracts were handed out to defenders this summer, there are probably a lot more teams set up for disappointment.
Teams with forwards that outperformed tended to use one of two models; spend big on superstars that perform even better than the big spending would indicate or spend average money on slightly above-average players and get good value for the money spent. Also, given that teams like Pittsburgh and Washington have had the benefit of top draft picks becoming stars, great GVS seasons could be on the horizon for the Lightning, Islanders or Oilers in the not-so-distant future.
Using this information, it may be prudent for teams to stock their blueline with three sets of second-pairing defensemen and use the extra money to either lure a superstar forward (like those on Detroit, Washington, Vancouver, Pittsburgh or San Jose) or use the Phoenix model and stock 4 lines worth of 2nd-line quality guys.
As for goalies, they represent a team’s best chance to outperform the contract money spent on them. In particular, having a starter paid league-average starter money and a backup with a lot of potential seems to be the best setup. Paying top-end money for goalies can ensure great GVT from the position (see NYR, VAN, BUF) but rarely gets great value for the money. In this case, I would expect to see Tampa Bay, Atlanta and San Jose benefit from some shrewd goalie signings this summer.
Related posts: