Monday, February 4, 2008

NHL 08: Introducing... The Clutch Factor

Clutch Goals, Soft Points, and the Overrated Superstar...


In today’s NHL, not all goals and assists are created equal.


There are big goals and not-so-big goals.


The value of an overtime shorthanded away goal against a strong divisional rival is undoubtedly greater than the value of an empty net goal against a weak non-conference team, scored on a two-man advantage.


For years, there has been a belief that certain NHL players elevate their games to score important goals in crucial situations in big games.


Clutch” players.


Meanwhile, other players pad their stats with garbage points in meaningless contests.


Soft” players.


Until now, this belief has been unsubstantiated.


We have created The Clutch Factor, a method by which the value of goals and assists may be objectively calculated and standardized.


By assessing the “Clutch Elements” which are at play on every goal, ranging from goal differential to manpower situation to opposition strength, and by analyzing every goal scored in the NHL 2007-08 season through the recent All-Star break, we have figured out who is, in fact, scoring clutch goals. And who is padding their stats with soft points.


Clutch Elements


Goal Elements

  •       Goal Differential at the time the goal is scored
  •      Manpower 
  •      Period of game
  •      Empty Net
  •       Overtime

Game Elements:

  • Home versus Away
  • Divisional, conference, or inter-conference opponent
  • Strength of opposing team
  • Defense of opposing team
  • Time of season


Terminology


Goals (G)

Assists (A)

Points (P)


Clutch Factor (CF)


This is the calculated factor which takes into account the Clutch Elements. 


Some goals are more clutch, and they will have a value of greater than one

Some goals are less clutch, and they have a value of less than one


(If you look at all of the goals scored in the league, they average out to be 1.) 


If a player consistently scores Clutch Goals (or assists), their overall Clutch Factor will end up greater than one.


Conversely, if a player scores soft goals (or assists), their overall Clutch Factor will be less than one.


However, looking at just the Clutch Factor is only part of the story. Ideally, we should also factor in how many goals the player scores. 


That is, the Clutch Factor tells you the quality of the goals, and the total goals tells you the quantity.


Therefore, if you multiply the quantity (total goals) by the quality (Clutch Factor), you end up with an adjusted goal (and assist and point) value, we call Clutch Goals (and Clutch Assists and Clutch Points) which takes both elements into account.


This gives us a good indication of the overall value of a player, taking into account the number of goals he scores as well as their quality.


Clutch Goals (CG): G X CF = CG.


Clutch Assists (CA): A X CF = CA


Clutch Points (CP): P X CF = CP

 


Results


We found a number of interesting, and sometimes counter-intuitive, results.

For example...


Question: Of all the Hart Trophy Candidates, who is scoring Clutch points? 


                                 P           CF        CP

Alfredsson                    67            .79             52

Lecavalier                     66            .93            60

Ovechkin                       65            1.06          67

Kovalchuk                     63            .98            60

Iginla                              63            1.09          67

Crosby                            63            .98            62

Zetterberg                     61            .86            52   

Thornton                       60            1.07          66

St. Louis                         59           .80            47

Getzlaf                            58            1.05          62

                                                                     

If the players were to be ranked by their Clutch Factor, the top ten order would look quite different...


  1. Iginla 1.09
  2. Thornton 1.07
  3. Ovechkin 1.06
  4. Getzlaf 1.05
  5. Kovalchuk .98
  6. Crosby .98
  7. Lecavalier .93
  8. Zetterberg .86
  9. St. Louis .80
  10. Alfredsson .79

 

Similarly, if the players were to be ranked in terms of their Clutch Points (CP)...


  1. Ovechkin 67
  2. Iginla 67
  3. Thornton 66
  4. Crosby 62
  5. Getzlaf 62
  6. Kovalchuk 60
  7. Lecavalier 60
  8. Alfredsson 52
  9. Zetterberg 52
  10. St. Louis 47
Conclusions: 

  • Ovechkin,  Iginla, and Thornton are scoring Clutch Points and should be considered for the Hart.

  • Alfredsson, Zetterberg, and St. Louis are scoring Soft Points.



Question:  Among the Rocket Richard Trophy Candidates, who is scoring Clutch Goals?

                                                            G                CF             CG          

  1. Ovechkin                           39            .95            37            
  2. Kovalchuk                         37            .96            36           
  3. Alfredsson                         32            .61            20            
  4. Iginla                                  32            1.06          34           
  5. Lecavalier                          30            .93            28            
  6. Boyes                                29 1.09         31
  7. Zetterberg                         28            .88            24           
  8. Malkin 27 .76           21
  9. Nash 26 .90           23
  10. Perry 26 .93           23
  11. Gaborik 25 .95            24
  12. E. Staal 25 .54            14


If the players were to be ranked in terms of their Clutch Factor (CF)...


  1. Boyes 1.09
  2. Iginla 1.06
  3. Kovalchuk .96
  4. Ovechkin .95
  5. Gaborik .95
  6. Lecavalier .93
  7. Perry .93
  8. Nash .90
  9. Zetterberg .88
  10. Malkin .76
  11. Alfredsson .61
  12. E. Staal .54


Or multiplied by their total goals, gives us their Clutch Goal ranking...


  1. Ovechkin 37
  2. Kovalchuk 36
  3. Iginla 34
  4. Boyes 31
  5. Lecavalier 28
  6. Zetterberg 24
  7. Gaborik 24
  8. Nash 23
  9. Perry 23
  10. Malkin 21
  11. Alfredsson 20
  12. Staal 14

Conclusions: 

  • Of the Rocket Richard candidates, only Boyes and Iginla have a Clutch Factor above the league-wide average.

  • Alfredsson and Staal are among the lowest in the league in terms of their Clutch Factor.

  • Ovechkin, Kovalchuk, and Iginla have the highest Clutch Goals among the top scorers.



Question: Who is scoring Clutch Assists?


                               A      CF      CA

  1. Thornton  45   1.06    48
  2. Savard      44 .94     41
  3. Crosby     43 1.01   43
  4. H. Sedin   42 1.03   43
  5. Datsyuk    41 .85     35
  6. Lidstrom   41 .74     30
  7. St. Louis   40 .81     32
  8. Getzlaf      39 .97     38
  9. Spezza      39 .81    32
  10. Lecavalier  36 .89    32
  11. Alfredsson 35  .92     32


Conclusions:


  • Thornton, Sedin, and Crosby are scoring Clutch Assists

  • Lidstrom, Spezza, St. Louis, and Datsyuk rank well below in terms of Clutch Assists.



Question: Which Norris Trophy Candidate are scoring Clutch Points?  


Total Points Ranking

  1. Lidstrom 46
  2. Rafalski 43
  3. Gonchar 40
  4. Pronger 36
  5. Zubov 35
  6. Chara 34
  7. Markov 34
  8. Campbell 33
  9. Jovanovski 33
  10. Phaneuf 32
  11. Kaberle 31


If the players were to be ranked in terms of their Clutch Factor (CF - Points)...


  1. Jovanovski 1.03
  2. Pronger     1.01
  3. Phaneuf     .90
  4. Chara        .88
  5. Kaberle     .88
  6. Rafalski     .81
  7. Zubov       .78
  8. Gonchar    .76
  9. Markov     .76
  10. Lidstrom    .76
  11. Campbell   .71 



If the players were to be ranked in terms of their Clutch Points

  1. Pronger 37
  2. Lidstrom 37
  3. Rafalski 36
  4. Jovanovski 35
  5. Gonchar 31
  6. Chara 31
  7. Phaneuf 30
  8. Zubov 29
  9. Kaberle 28
  10. Markov 26
  11. Campbell 25

Conclusions:

  • Jovanovski and Pronger have the highest Clutch Factors among offensive defensemen.

  • Campbell, Gonchar, Lidstrom, and Markov have among the lowest Clutch Factors.



Question: Which projected 30 goal scorers are among the league's best in terms of Clutch Factor (Goals)?


                G CF CG


Getzlaf 19 1.23 23

Kopitar 19 1.21 23

Hagman 18 1.29 22

Zherdev 20 1.12 23

Svatos 20 1.28 26

Horcoff 21 1.13 24


Conclusion


These players may be undervalued based on conventional statistics.



Question: Which projected 30 goal scorers are among the league's worst in terms of Clutch Factor (Goals)?


                  G CF CG


E. Staal 27 .54 14

Alfredsson 32 .61 20

Brown 20 .59 14

D. Roy 18 .60 11

Langkow 20 .70 14

Prospal 19 .70 13

Whitney 21 .74 15

Antropov 18 .74 13

Malkin 27 .76 21


Conclusion


These players may be overrated, based on conventional statistics.



Which Upcoming Unrestricted Free Agents are Clutch Scorers?  

Clutch Factor (Points)


M. Hossa .92

Campbell .71

Conroy 1.08

Huselius .90

Brunette .97

Fedorov 1.10

Cleary .95

Nagy .50

Demitra 1.04

Satan 1.26

Comrie .88

Redden .77

M. Naslund 1.04

Stillman .84


Conclusions:


  • Satan, Fedorov, and Conroy may pay dividends.

  • Nagy may not.



Other interesting numbers...


Above average:


Sundin CF (Points) = 1.01

Dumont CF (Points) = 1.03

M. Michalek CF (Points) = 1.14

Doan  CF (Goals) = 1.18

Cammalleri CF (Goals) = 1.06

Penner  CF (Points = 1.08

Modano CF (Goals) = 1.45

Hemsky CF (Goals) = 1.12

Kariya CF (Goals) = 1.18

Tanguay CF (Goals) = 1.38

S. Koivu  CF (Goals) = 1.07

Satan CF (Goals) = 1.37

Toews CF (Goals) = 1.10



Below average:


Stillman CF (Points) = .84

Briere CF (Points) = .85

Kovalev CF (Points) = .76

Heatley CF (Points) = .86

Jagr CF (Points) = .80

B. Richards  CF (Points) = .76

Kessel CF (Goals) = .56

Kotalik CF (Goals) = .68

O. Jokinen CF (Goals) = .83

Guerin CF (Goals) = .85

Drury CF (Points) = .78

Kane CF (Goals) = .77


These are just some of the results obtained so far.

We believe this is a useful objective tool, whereby the value of a goal or assist, and the players who score them, may be measured more accurately. 


This may be helpful in player scouting, contract negotiations, arbitration, signings, and player recognition.

19 comments:

A poorly organized band of hockey fanatics said...

Nice job on a interesting concept.
Any thoughts on doing any research on who might be the all-time clutch performer??

BTW; I have added a link to your site on our blog.

Jeff J said...

You need to read this, this and this.

Anonymous said...

I think having the strength of opponent stat throws this off. A strong team is strong due to it's strong players. However their contribution is watered down because... they're on a strong team.

We don't know how well they would do if they weren't on a strong team; however we see that all the leaders in your categories are from weaker teams. I think you may be introducing a bias.

The Puck Stops Here said...

How well does this work from year to year? If there is anything to this method, the same players will appear near the top and bottom of the clutch standings year in and year out. If there is nothing to it, last season you would find a vey different group of clutch players and non-clutch players and the year before that would be yet another random assortment of players who are clutch and non-clutch that year.

Can you give clutch standings year by year for a few seasons into the past?

Unknown said...

I don't really have the time to go thru and total the assists for Thornton through the All-Star break... but Thorntons assists you have listed at 45, and his clutch assists at 48... just an error? if so, what are the real numbers?

Anonymous said...

Thanks for the comments...

Poorly organized band: I agree, it might be interesting to figure out the alltime clutch performer, but crunching every goal for all eternity might be challenging.

It would be useful, though, to see if the year-to-year stats hold up.

Regarding strength of opposition, the point is a reasonable one, but I'm not sure we're saying the same thing. We're not talking about the strength of the playing team, but the strength of the opposing team.

Jeff, I agree that there are fundamental difficulties with labelling a player "Clutch". However, when I first embarked on this project, I wasn't even convinced that there would be a significant difference between players. Over the course of the year, and thousands of goals, most stats regress to the mean. However, it is certainly apparent that using these criteria, there are consistent differences between players.

Now you can, and do, argue that the original criteria are subjective. This is indisputable. But while I'm not saying Alfredsson's points should be disregarded, because he scores so many of his points in blowouts against weak teams, I am saying that perhaps his points should be discounted.

Even the stats which you like are subjective. For example, game-winning and tying goal stats are misleading and in my opinion, of minimal value. In fact, my frustration with the infamous GWG was the impetus for the development of this concept.

Appreciate your input.

Anonymous said...

Michael,

Thanks for the comment.

The goals, assists, and points are the official ones. The Clutch Goals, Assists, and Points are the ones calculated based on the Clutch Factor.

Anonymous said...

Yea, but Jeff J, the clutch factor is about sports, those links are about baseball. ;)

The Pest said...

Hey Nathan,

Good points. A few comments.

[quote]
Furthermore, unlike turn-based sports such as baseball or golf, hockey is free-flowing, and as such, specific situations that would be considered "clutch situations" do not arise on a consistent basis.
[/quote]

I'd argue that these situations occur on a constant basis, but are just harder to identify. All in all, theres way more events that take place in scoring a goal then in a hit, from a hitter's perspective. I think its kind of apples and oranges trying to compare baseball and hockey.

[quote]
it appears that the criteria that you put forth give more weight to players and teams that take a lot of games to overtime, don't get out on top of their opponents, and play more in even-strength and short-handed situations. None of these are good things, and yet it appears that you're rewarding them.
[/quote]

I think I agree that this is the biggest weakness, but I'm not convinced. Ottawa does all these things on a regular basis and Ottawa has choked in the playoffs with this strategy every single year. The truth is these are great in the regular season but in the playoffs more is on the line, defenses tighten up, teams don't take stupid penalties and you have to learn to gut out wins. When you depend on power plays for scoring and don't get them in the playoffs, a major part of your game plan is gone. Teams that don't practice this in the regular season are soft regular season teams, and end up getting spanked in the playoffs. Ottawa vs Aniheim, QED.

[quote]
Campbell doesn't reward players and their teams nearly as much for being in poor situations, as his calculations focus more on scoring the first goal, go-ahead goals, and game winning goals.
[/quote]

All these things are actually factored into the stat via the goal differential at the time.

I think what this stat is trying to do is blend the team/individual perspective. Both count because its how an individual plays within a team concept that makes him great or garbage. Thoughts/comments?

Anonymous said...

Nathan, i think you're having a hard time grasping the team concept. A guy like AO would (arguably) do better (ie. score more points) on the top line playing with either Alphy, Spezza, or Heatley. Nothing to question therer.

Sure the author of this statistic may have flaws, but in no way could you compare, say the +/- of Allfredsson, to others like Jokinen, AO, or others would dont play on more all-around sound team. He is just trying to level the playing field for those who are not forturnate enough to have stronger supporting players.

Anonymous said...

on another note, I'd like to see a play-off version of this stat.....

Anonymous said...

Cool, I heard about this on the Fan960 in Calgary and thought it was interesting!

One question, where do you get the stats? I assume you've got a database or spreadsheet with the info, just wondering if you have a good source for online stats helping you out, or if you did it more manually. I've had a couple of ideas that I wanted to try with a database of season stats, but getting that info without paying big $$ or writing a screen scraping program has been impossible.

JavaGeek said...

Just to make the connection to the baseball 'clutch-ness' I figured I'd note that in baseball they would look at a statistics like ABA clutch vs. non-clutch where ABA = hits/at bats.

In hockey to do the same you need some metric of hits/ at bats, for example goals/second (or hour) and compare how a player does in clutch vs. non-clutch.

Of course in baseball ABA is easy to calculate, in hockey the clutch vs. non-clutch calculation would take quite a bit of effort (you need ice time data for every player to correlate with clutch parts of the game).

By your CF metric there should be correlations from season to season. Why, because the best players get the most clutch ice time (look at Tampa if you don't believe me). Iginla is out is the last 30 seconds of a 3-2 game, not Godard. Godard's opportunity for a clutch goal is much lower than Iginla's. In fact if a team was up 10-1 Godard might actually get a bit more ice time (so the opposition doesn't hurt their best players).

That all said, I'll make the same bet Phil made for baseball for hockey, given you get the statistics in the form out something per hour....

Anonymous said...

What about clutch defense, that seems more important, so in that case, Nick L for Heart and Norris

Anonymous said...

Thanks for the comments.

First of all, I completely agree that it is a difficult task to come up with a "Clutch Factor". The elements which should be included and excluded are, of course, debatable.

For example, we chose to exclude such factors as the first goal of the game. Yes, the first goal of the game is an important stat, and is somewhat predictive of the game's outcome. But is that goal any more (or less) valuable than the go-ahead goal scored with one minute left in the third?

My contention is that the goal differential is the best way of measuring this.

Another factor which we decided not to use was whether or not a goal was scored in a win or a loss. Again, you could make a case for and against including such a stat. I would be interested to know what everybody thinks.

Nathan, I think your comments are reasonable, and well-thought out, and I appreciate the effort. You are obviously a knowledgeable sports guy. Regarding methodology, really, I have nothing to hide. Essentially the factors which I listed in the original post were given a weighting, this weighting was applied to each goal scored in the league, and the player who scored the goal got credit for the adjusted goal value. The cumulative stats were generated, and a clutch factor was distilled. The details are somewhat boring and laborious to recount, but if you have specific questions, let me know.

The data was just input manually from reputable sites, so unfortunately, I don't have a pipeline to Bettman's office. Sorry Photon.

There is no doubt that the "clutch factor" is just one measurement. It does not replace the official goals, assists and points, which are a valuable measurement of "quantity". But if you use this in conjunction, you can look at the "quality" of the goals as well. Nobody is saying that a player who scores one overtime goal in an important game, and ends up with a high clutch factor, is more important than Alfredsson, who scores tons of goals. But if you want to compare Alfie with his high-scoring peers, that's where maybe this might be useful.

And of course, when you're looking at the Norris, as well as other stats, defensive skill should not be overlooked. There is nothing built into this stat which would account for that. This is purely an offensive statistic. However, recent memory suggests that the Norris tends to go to an offensive defenseman.

Anonymous said...

Ah thanks, I had a glimmer of home that someone had found an RSS feed or XML feed somewhere that I haven't been able to dig out with my google-fu.

We were talking about it in the office today and a buddy thinks the only flaw is that it doesn't take the ultimate outcome of the game into account. I can see good reasons why not to do it, as then it would favour winning teams (for one).

Plus it could indirectly be taken into account depending on how the math is done; overtime goals for sure, and goal differential as well if a go-ahead goal is weighted more than a tie it up goal.

There's a bit of discussion at our site here: http://forum.calgarypuck.com/showthread.php?t=54062

The Pest said...

Not sure how to quote... so I'll just italicize what you said. :)

I'm having no issues grasping the team concept. I disagree with the assumption seemingly made by these calculations that players are made by good teams, rather than good teams being made by players.

I'm not sure the stat reflect that. The adjustments are still based on things the individuals do. If you score goals when the team is down, or score a go ahead goal, you get a bonus. If its against a particularly good defensive team, you get a bonus. Those are all individual things that exist within team sports. I think you may understand that this stat is trying to say someone like Alfredson sucks. I don't think its saying that. Its saying that he's not in many high pressure 'clutch' situations. And if you're not in clutch situations, you can be a great hockey player but you can't be clutch.

If Ovechkin had a centerman like Pavel Datsyuk, yes, I would expect him to have better +/- numbers because Datsyuk is a Selke-calibre player.

Again, I think +/- is a totally ineffective stat simply because it depends on things like team strategy and has way more to do with what line you're on then your individual contribution. Its a team/line stat, not an individual one like this is.

And yes, I would expect him to score 70 goals this season instead of 60. The point is this, though. If A.O. had Datsyuk on his team, the Caps would be better, and as such, A.O. would not have to score as many goals while trailing, tied, or in overtime. But nobody would suddenly dispute how amazing he is!

But they would dispute how important he is to his team, as is done in an MVP debate. If you're team is way better when someone else comes compared to when you're not there, its pretty likely your contribution to the team are not as significant, therefore not as valuable.

Wally, I agree with your point about comparing baseball to hockey. That was essentially my point. This type of metric is better suited to baseball because it's strictly turn-based, and puts more responsibility on individual performances than it does total team-play.

I think that indicates the need for more abstract stats in hockey though. Its easy to look at baseball stats and see exactly what someone hits, therefore theres no need for a stat like this. In Hockey, as you've pointed out, its more difficult. What the clutch factor attempts to do is extract individual metrics from the team concept, and tangibilize the intangible.

The problem is that such a drop-off falls on the shoulders of the entire team, as well as the preparation of the coaching staff...

Organizational depth is important in the playoffs, and says nothing of the "clutchness" of star players...
that's not and indication that the top players aren't "clutch" -- it's an indication that the opposition is trying to stop the top players at all costs, and forcing the role players to win games.


I totally disagree. Last year, Ottawa's only successful line was the third line who played with passion and heart. Ottawa's stars were not clutch. Sure, they faced a great team, but they shrank when challenged. They skated out of hits and were eventually beaten mentally. Their top players, their leaders, did not lead, they got shut down because they were soft.

the truly great players make the adjustment. Dany Heatly doesn't forget how to shoot the puck when the playoffs start, and Spezza doesn't forget how to pass.

Hmm... then why can't he score? I'd suggest its because when the heat is on, as it is in the playoffs, he does not forget, but just doesn't know because he's out of practice... this is what the stat suggests.

In the playoffs, teams are willing to take their chances with the second line if they're shutting down the first one.

Yea, but its not like teams don't have a choice because the defense is keying in on the top line. It happens, but you're a star hockey player: you're supposed to find ways to make it work. Sports is about overcoming those challenges. Truly great players are clutch players, and know how to adjust like you said. Ottawa did lose because their stars shrank under the pressure, everyone watched it... they played scared. I think this stat is somewhat indicative of this.

The Pest said...

Hey Photon, email me at wallyo@hotmail.com... sorry, but don't know how else to reach you.

JavaGeek said...

Please e-mail a complete breakdown of the points awarded for each factor listed in the post. hockey@javaplus.cjb.net