I wrote a post the other day about the logical fallacy in most arguments for paying Phaneuf the $7mm a season over 7 seasons, and while my intention was never to pinpoint his exact value I took some criticism (justly so) for not backing up my hypothesis with facts. So here goes an attempt at valuing Phaneuf based solely on his 2012-2013 season.
First, I will address the fact that looking at just one season is problematic: it is. Looking over a longer time period may benefit or hurt Phaneuf, but frankly I don't get paid to analyze this stuff and doing one season was enough work. That said if anyone has the advanced stats for the past 3 to 5 seasons in an excel spreadsheet, I'd be happy to perform this same analysis over a longer time period. I also tend to think that the Leafs strong performance (relative to previous years) of last season helped Phaneuf more than hurt him (i.e. if I included more seasons in the analysis).
Second, I will go over methodology. I believe players (defencemen in particular) get paid on three things: offensive ability, defensive ability, and ability to eat up minutes. Having strong offensive (or defensive) ability will obviously get you paid, my theory on minutes is that if you could find a guy who played 75% of the game it would free up a lot of salary room for you... so you'd naturally pay him more (i.e. if you could have 4D instead of 6 you would spend more on getting 4 awesome ones).
I took the top 50 highest paid players in the NHL (defence only) and compared their stats, I removed the three players that did not play 10 games in the 2012-2013 season (Jovo, Meszaros, Pronger). The reason I used only the top 47 (as opposed to all defencemen) is because it would give me a better idea of Phaneuf's peer group (not comparing his stats to say Mark Fraser) and how they do. The consequences of using just the top 47 are: average salary is very high ($5.6mm adjusting for UFA/RFA status) and so are average stats (like ice time, points, etc). If a player is better than average in this group, then he will get paid more than average ($5.6mm) by the number of standard deviations he is above average, obviously the same goes if he is worse than average.
To determine offensive ability I would normally just look at even strength points versus even strength ice time, however last season was short, and getting 11 versus 12 even strength points would make a huge difference to some of the stats... even though the actual difference wouldn't necessarily be noticeable to teams. So what I did was look at total points scored, divided that number by total ice time (that gets points per minute) and then multiplied that number by 20 (points per 20 minutes of ice time). The reason I use 20 is because that's approximately what you'd expect a defenceman to play in a game. This gives me P/20. Unfortunately that isn't a completely fair measure in that some players with much fewer games played could have a very high P/20 because they went on a hot streak. So for any players who played under the average number of 20 minute chunks (49.6, or just under 1000 total minutes) I adjusted their scoring.
P/20 for players over 49.6 20 minute chunks (~1000 minutes) = total points/total ice time * 20.
adjusted P/20 for players under 49.6 20 minute chunks = (total points + (total points/total ice time)*0.75*(~1000-total ice time))/~1000 * 20
Basically it just means that a guy who played 800 minutes and scored 1 P/20 (40 points) would have an adjusted total of 47.5 points for an adjusted P/20 of 0.95.
The average adjusted P/20 for the top 47 (by salary) was .4078, Dion Phaneuf had 0.4685, or 0.4816 std deviations ABOVE the mean. By this methodology Dion Phaneuf is a slightly above average (for the top 47 by salary) offensive defenseman.
Defense is a much harder thing to figure out and while I'm not completely sold on 'advanced' stats, I think Corsi For is a fair measure of a team or individual's defensive ability. Naturally Corsi For relies heavily on your teammates and your team's strategy, so luckily extraskater.com provides relative Corsi For. That is the Corsi of an individual player relative to his team's Corsi. I looked at 5v5 figures only (excluding PP and PK for obvious reasons). You would expect that a strong defensive player would have a positive relative Corsi For % and a weaker one to have a negative Corsi For %.
The average relative Corsi For % was +0.80%. Dion Phaneuf was -3.80% or -1.4260 std deviations below the mean. By this methodology Dion Phaneuf is a well below average (for the top 47 by salary) defensive defenseman.
This is straight forward. I looked at total icetime, which I had already divided into 20 minute chunks (to see the equivalent of full games played each player). No fuss no muss.
The average 20 minute chunks played per player was 49.5738 in the shortened 48 game season. Dion Phaneuf played 59.76 20 minute chunks or 0.9563 std deviations above the mean. By this methodology Dion Phaneuf is above average (for the top 47 defensemen by salary) in terms of workload.
So I'm aware that no system is perfect. I'm aware, for example, that I didn't factor in things like penalty kill time in P/20, or Quality of Competition in relative Corsi For %. I'm also aware that if someone were to be elite in any particular category they may get paid more than their standard deviations would indicate. The reason I didn't use quality of competition as a measuring stick, is it doesn't necessarily tell us anything. I'm sure Brad Marsh had ridiculous QoC numbers back in Ottawa's first season. It doesn't mean he was a great defenseman (at the time), it can be measure of the coach not having a particular great alternative. Also most of the top 47 had a very tight distribution on QoC numbers playing against competition that, on average, played 29.1% of the game (versus Phaneuf and his 30.2%).
Using an equal weight of the three measures Dion Phaneuf ended up as the 29th most valuable (of the 47) defenseman, with a "worth" of $5.6mm based on last season.
Statistical problems that worked in his favour in determining this value:
- Only the top 47 by salary were included, so naturally the average salary was very high. Including all defencemen would have dropped the average salary substantially.
- Only the top 47 by salary were included, so some of the better offensive performers were excluded (i.e. Norris finalists Letang and Subban)
- Only the top 47 by salary were included, so the the relative Corsi For % had a high standard deviation. His minus 3.8% would be abysmal by total league standards.
Statistical problems that worked against him in determining his value:
- Including more players would have made his offensive stats a higher standard deviation above normal, although in order to justify the much higher salary relative to average, that would be expected/required.
- Ditto for his ice time stats.
- It was a short season, so there's not much you can do. There will be more statistical outliers, injuries had a greater impact, etc.
|Rank||Player||CF% Rel||T20||Adj P/20||Worth|
|1||Timonen, Kimmo "||6.30%||48.60||0.5967||$ 7,567,521|
|2||Ehrhoff, Christian "||8.10%||57.81||0.3806||$ 7,383,536|
|3||Campbell, Brian "||4.10%||62.64||0.4310||$ 7,138,748|
|4||Yandle, Keith "||3.00%||52.32||0.5734||$ 7,024,000|
|5||Goligoski, Alex "||3.40%||52.17||0.5175||$ 6,813,393|
|6||Boyle, Dan "||6.40%||51.29||0.3899||$ 6,709,725|
|7||Shattenkirk, Kevin "||4.70%||50.40||0.4563||$ 6,656,403|
|8||Suter, Ryan "||-0.70%||64.56||0.4957||$ 6,635,107|
|9||Hamhuis, Dan "||3.70%||53.82||0.4460||$ 6,610,978|
|10||Byfuglien, Dustin "||1.70%||51.82||0.5404||$ 6,573,320|
|11||Doughty, Drew "||2.60%||62.64||0.3512||$ 6,444,003|
|12||Kronwall, Niklas "||0.20%||57.60||0.5035||$ 6,437,454|
|13||Markov, Andrei "||-0.80%||57.12||0.5252||$ 6,321,759|
|14||Karlsson, Erik "||7.30%||22.87||0.6123||$ 6,313,230|
|15||Weber, Shea "||-1.10%||61.20||0.4575||$ 6,165,676|
|16||McDonagh, Ryan "||3.40%||56.40||0.3369||$ 6,158,421|
|17||Visnovsky, Lubomir "||8.00%||39.20||0.3571||$ 6,141,013|
|18||Gonchar, Sergei "||-0.50%||52.88||0.5106||$ 6,055,374|
|19||Streit, Mark "||-0.80%||55.20||0.4891||$ 6,026,724|
|20||Tyutin, Fedor "||1.30%||56.40||0.3901||$ 6,013,170|
|21||Ekman-Larsson, Oliver "||0.10%||58.80||0.4082||$ 6,011,017|
|22||Chara, Zdeno "||2.20%||59.04||0.3218||$ 6,004,204|
|23||Pietrangelo, Alex "||0.10%||58.28||0.4118||$ 5,998,396|
|24||Wideman, Dennis "||0.30%||57.04||0.3857||$ 5,833,005|
|25||Hedman, Victor "||1.30%||49.50||0.4040||$ 5,673,379|
|26||Keith, Duncan "||-2.80%||55.46||0.4868||$ 5,638,214|
|27||Garrison, Jason "||3.00%||49.82||0.3212||$ 5,610,201|
|28||Green, Mike "||-2.20%||42.88||0.6064||$ 5,608,220|
|29||Phaneuf, Dion "||-3.80%||59.76||0.4685||$ 5,605,631|
|30||Edler, Alexander "||-0.40%||52.43||0.4196||$ 5,591,830|
|31||Carle, Matt "||-1.70%||56.40||0.3901||$ 5,424,492|
|32||Martin, Paul "||-2.00%||42.84||0.5369||$ 5,296,591|
|33||Bouwmeester, Jay "||-2.80%||57.34||0.3837||$ 5,232,403|
|34||Voynov, Slava "||-3.90%||52.80||0.4735||$ 5,197,379|
|35||Giordano, Mark "||-0.30%||54.05||0.2775||$ 4,995,034|
|36||Wisniewski, James "||1.70%||33.75||0.4148||$ 4,870,354|
|37||Vlasic, Marc-Edouard "||3.90%||48.72||0.1437||$ 4,831,184|
|38||Seabrook, Brent "||-3.50%||50.53||0.3958||$ 4,751,263|
|39||Johnson, Jack "||-4.00%||55.88||0.3400||$ 4,691,186|
|40||Enstrom, Tobias "||-1.60%||24.53||0.6115||$ 4,661,741|
|41||Bogosian, Zach "||-2.80%||37.62||0.3721||$ 4,003,162|
|42||Bieksa, Kevin "||-2.20%||38.70||0.3101||$ 3,873,716|
|43||Myers, Tyler "||0.10%||40.56||0.1972||$ 3,869,490|
|44||Pitkanen, Joni "||0.20%||24.97||0.3604||$ 3,781,680|
|45||Volchenkov, Anton "||1.30%||29.05||0.1721||$ 3,295,092|
|46||Michalek, Zbynek "||0.00%||35.36||0.0566||$ 2,835,305|
|47||Coburn, Braydon "||-3.00%||36.96||0.1353||$ 2,736,552|