What is Clutch (Baby Don’t Lose This, Don’t Lose This No More)

I was looking for another project to do this weekend – something time and attention consuming but not actually too hard because I was fairly distracted. Someone suggested looking at what makes a player “clutch” and I thought “hmm, that’s interesting.” So here we are.

What does make a player clutch? If I was going to use fancy stats to rank players, I needed to figure out how to define the term first, as it would set the guidelines for the investigation. From dictionary.com, they define it as:

Clutch, adj.

done or accomplished in a critical situation

example: a clutch shot that won the basketball game.

In hockey, we have already proven that teams play very differently depending on the various “game states” (eg, tied, winning by 1, losing by 1, etc). Micah McCurdy has an excellent presentation illustrating this exact phenomenon, and the term “score effects” is used frequently amongst analytics types. In fact, most data you’ll see presented (here or elsewhere) is usually what we call “score adjusted”, meaning it has modifiers to counteract these score effects.

But if we’re trying to determine who is the most “clutch”, that is, who is the most effective in critical situations when the game needs to be tied or won, then these score effects are exactly what we want to examine.

First, some parameters. Because the burden of scoring most often falls to forwards, that is who is in this data set. I also only looked at 5v5 data because it’s easier to score on the power play (which is kind of the point), and then I set the minimum 5v5 minutes played at 750 for the season, to cut out call ups and  the guys the coach doesn’t trust even in winning scenarios (sorry not sorry Tanner Glass). All data is from war-on-ice, and is regular season only (actually am sorry, Justin Williams).

Next, I tried to determine which factors to use in my ranking. While Time on Ice was important for my special teams rankings, per McCurdy’s work, there seems to be little change in “roster strength” across game states (especially when Leading by 1, Tied, or Losing by 1), which leads me to believe it’s not important in this scenario. Instead, I chose to look at individual effort in the form of individual High Danger Scoring Chances per 60*, effect on team in the form of Scoring Chances per 60, and efficacy in the form of Points per 60.

Now to determine who was actually being “clutch”, I had to look at the difference between their play in a state of relative comfort (Leading by 1) vs states that could be considered crucial scenarios (Tied, Trailing by 1, Trailing by 2). I didn’t look at Trailing by more than 2, because it’s less likely a team will come back and win.

Once I had these differences, I ranked them. Large differences were good, as it meant a player was outperforming their play in the “Leading by 1” state. A player received a rank for iHSC/60, SCF/60, and P60 in each of the three game states. Then, I gave each of those rankings equal weighting, to create one overall rank per game state. Last, I weighted each game state – 40% for Tied, 40% for Trailing by 1, 20% for Trailing by 2**, to achieve a final “clutch” ranking.

Top 20 Clutch Forwards, 2014-15 Season

Top 20 Clutch - F - 1415

Top 20 Clutch Forwards 2013-14 Season

Top 20 Clutch - F - 1314

While it may not look it, there were actually several forwards who proved clutch each year – Bonino was #1 and #18 in 14/15 & 13/14 respectively. Bergeron (#2 & #25), Horcoff (#5 & #28), Getzlaf (#8 & #36), Seguin (#23 & #17), B. Boyle (#29 & #13), Girgensons (#33 & #1), and Jokinen (#34 & #14) all also put in good showings. In 13/14 the list contained 258 forwards; in 14/15 it had 267.

There were 61 names who didn’t make the 14/15 list who were on the 13/14 list, mostly because of injury (all of CBJ’s roster), or for other obvious reasons like retirement (Saku Koivu) or just not being very good at hockey in general (FHBF Brandon Bollig). One interesting bit of trivia – players who were traded during these two seasons typically dropped an average of 22 spots in the ranking. Players who remained on the same team? Average drop of 1 spot.

Comparing the two years gave rise to some excellent questions about the mental part of the game. For instance – in 13/14, the entire Chicago 4th line (Bollig-Kruger-B.Smith) made the top 20. In 14/15, Bollig & Smith had been traded away, and Kruger, formerly #4, dropped to #111 – 107 spots difference. Now, we know that Bollig isn’t a particularly great hockey player, but it certainly looks like his absence on Kruger’s wing was felt. On the other hand, Brandon Saad was a 3rd/2nd line guy for most of 13/14, but last season spent much of his time with Hossa & Toews. He went from #212 to #68. Hossa also climbed 94 spots, and Toews jumped 35. Coincidence or chemistry?

In 13/14 Zemgus Girgensons held down the #1 spot. In 14/15, he was out of the top 30 – and doing better than most of his Buffalonian peers. On average, the untraded Sabres forwards dropped 38 spots year over year, higher than the rate of traded players. Buffalo was considered the model for “tank nation” in 14/15 – is that just a coincidence? Vanek, who was traded from Buffalo to about a million teams in 13/14, ended up at #69 in 14/15, climbing 131 spots. Coincidence?

Anyway, I think the best data both answers and raises questions, and this certainly did that for me. I’ve stuck the rankings in a google doc for everyone to look at themselves. I think the next logical step for this kind of ranking is to look at playoff performance – see if guys like Justin Williams and Jonathan Toews have really earned their “clutch” reputation.

Please check out part 2, where I address some of the weaknesses with this methodology and include defenders.

* I chose High Danger Scoring Chances over Scoring Chances for the individual portion because they are both more difficult to create (separating good players from bad) and more difficult to defend (greater effectiveness)

** Trailing by 2 received a lesser weight because it’s more difficult to win a game when down by 2, and therefore “less clutch” to score in that state.


4 thoughts on “What is Clutch (Baby Don’t Lose This, Don’t Lose This No More)

  1. This is super interesting. I wonder about the repeatability of it, though – you showed that several players made the list in both years, but there seemed to be a pretty big variance in position. The other thing that struck me is that it seems to specifically pick out players who are not good all the time; that’s the only way to get a really big difference in level of play like you’re showing. It’s cool if you can really up your game at tactical points, but I feel like a coach, given the choice between someone who is adequate most of the time and really good in high-pressure situations, and someone who’s really good all the time, would probably prefer to go with the player who’s consistently good. You picked out Jonathan Toews as someone you’d like to look at in the playoffs, and I’d be interested to see that too, because he’s “supposed” to be an utterly reliable player – at least that’s the argument made by people who say he’s one of the best in the NHL, despite lesser scoring than many superstars. Anyway, really enjoyed the article, it made me think!


    • I’ve been playing around with the Playoff numbers today – and while several of the same people did show up at the top (Bonino, Bergeron), and some of the “clutch” playoff guys (Williams, Ward) also came up high, Toews (and most of the Hawks) didn’t rank that well. I think part of it has to do with methodology – ranking them against themselves (guys who play at a high level at all states of play would end up in the middle ground, incidentally that’s where Toews is), so I’ve been looking at the various averages based on TOI to see if that would change anything. Anyway, it’s basically been proven (as linked) that coaches don’t change who’s on ice in those states, so this is more of an exercise in determining the psychology of it all.


  2. Hey Carolyn, interesting article. I did have to call Merrin up and ask her to clarify/confirm some things though.

    I would probably question your interpretation of “clutch” somewhat – as I understand it, you are comparing a player’s typical 5v5 performance against their 5v5 performance in “clutch’ situations and then ranking the difference, correct? But wouldn’t that penalize already high performing players? Even though they could be scoring more than the lower performing player, even in clutch situations? I was thinking in an academic way – if I have a student who consistently gets 90% on projects, and then scores a 95% on an exam vs the student who consistently gets 60% on projects but then comes up with an 85% on the exam. Even though the second student’s gain was greater, the first student still outperformed them.

    IDK, just wondering I guess.


    • Yes, that is definitely one of the weaknesses with this particular ranking, however, you’d also think that guys who are considered “clutch” would perform better vs their own high level of play, and poor players will still rank poorly. Mostly that is what you see on the big list, but I’ve been messing around with the playoff numbers today, and trying to see if there’s any real difference in TOI averages. Anyway, I think there’s a lot more investigating to be done, but it’s a fun starting point.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s