The story of a clutch player with ice water in his or her veins pulling out a victory through sheer force of will is the type of high drama that draws fans in. Examples of such feats live on in sports lore: Tiger Woods in his prime making incredible shots under the pressure of a Sunday at a major tournament, Joe Montana putting together a 92 yard drive at the end of the game to win Super Bowl XXIII, and Michael Jordan scoring 38 points while playing with the flu in game five of the 1997 NBA finals.
In hockey the “clutchiness” of Patrick Roy, who earned three Conn Smythe Trophies backstopping the Montreal Canadiens and Colorado Avalanche to two Stanley Cups each, is legendary. Roy was listed as number one on Bleacher Report’s top 50 most clutch hockey players of all time, and it is hard to argue with this selection. More recently Jonathan Toews was given the nickname “Captain Clutch” for seemingly raising his game when it mattered most to will the Chicago Blackhawks to three Stanley Cup championships in a six year span.
As analytics entered into professional sports some began to question whether clutch performances actually exist. For example, although Patrick Roy’s career .918 save percentage during the playoffs really that much better than his .912 regular season save percentage? This type of difference is save percentage amounts to allowing roughly one fewer goal every ten or so games. That definitely could make some difference in long playoff runs but it does not seem to justify being rated by some as the top clutch performer in NHL history. Along the same lines Jonathan Toews’ .87 career points per game in the regular season is almost identical to his .86 points per game during the playoffs. The lack of ability to tease big difference in performance over large sample sizes leads some analysts to conclude that clutch does not exist beyond very good players simply continuing to be very good players. But does this really mean clutch does not exist, or is grossly overrated if it does? Or is this simply a methodological issue where the tools used on large sample sizes and datasets are not really designed to tease out the nuances of social psychological phenomena?
With the goal of getting to the heart of what clutch is and how it works this article will revolve around three key questions: What is clutch? How can we measure clutch? And, where does clutch come from? Although this article is about clutch play it will also deal with choking, which is essentially the flip side of the same coin. As always the underlined words are hyperlinks that will take to you the articles I refer to in this piece.
What is clutch?
Although there are some minor variations in details all definitions of clutch refer to superior performance, also known as performance increment, that occurs in pressure situations. Conversely, choking is defined as an inferior performance (performance decrement) under pressure. Pressure is simply situational factors that lead to stress and may impede performance, such as playing in front of a larger than usual audience, making it further into the post season, or being on the ice when the game or the season is on the line. When it comes to clutch performances context is everything and the athlete must know that it a pressure situation at the moment it is happening for it to count as clutch. For example, if a hockey team is up by five goals the sixth goal would not be considered a clutch play since it was not scored in a pressure situation. If the opposing team came back and scored five goals of their own, meaning that sixth tally was the game winning goal, the big comeback does not retrospectively make the sixth goal a clutch play.
The main issue with contextual definitions is that contexts are not always straightforward as multiple factors come into play at the same time. A good example of the difficulties of defining clutch plays in hockey is this goal by Erik Karlsson against the New York Rangers in game six of their 2017 second round playoff series:
— Sportsnet (@Sportsnet) May 10, 2017
When the goal was scored the Senators were already up by one in the second period of the game. An argument could be made that this was not necessarily a clutch play because there was still a lot of hockey left to be played and at the time it seemed like an insurance goal. However, it could just as easily be argued at any point of a playoff series is high pressure and thus all superior performances and plays can be considered clutch. Stephen Burtch used this latter approach in these tweets and looked at overall playoff performance compared to regular season performance to draw his conclusion about whether a given player was clutch:
@JDylanBurke Williams prod was only sig outsized in one playoff year. Reg Season 0.25 g/gp 0.63 pts/gp. Playoffs 0.26 g/gp 0.68 pts/gp.
— Stephen Burtch (@SteveBurtch) August 17, 2015
While defining clutch is relatively straightforward figuring out which contextual factors matter most can be a subjective process that is open to varying interpretations. This should not be taken to mean that clutch does not exist or that it is not important. It is simply a reminder that good and consistent definitions are an essential part of producing meaningful empirical results.
How can we measure clutch
Academic researchers looking at clutch draw on mix of quantitative and qualitative approaches including experiments, surveys, and interviews. The key feature of the methods used is they are sensitive to the situational factors that create pressure situations that separate clutch players from the pack.
Most social scientists looking at clutch performance construct experiments designed to tease out athlete performance under pressure while holding as many factors as possible constant. To give one brief example of how such experiments may work, a study of clutch in basketball may involve getting a group of players to take free throws. The first round of free throws would act as a baseline, a second round could be used to help take a warmup effect (i.e., players shoot better after they have a chance to warm up and get used to the surroundings) into account. A stressor would be introduced and then subsequent rounds of free throws “under pressure” would be taken. This stressor sometimes uses imagery (“think about a game being on the line”) but more often it involves some type of reward that could be money or some other type of meaningful prize for those who perform well. Such experiments are very effective in identifying groups of clutch players.
A second component of empirical clutch research is typically some type of survey to measure assorted factors that may be associated with clutch performers. The survey could be paper and pencil (or a version that can be done on a mobile device) but my personal preference for situations like this is for researchers to use coded structured interviews where someone asks the survey questions face to face. My experience is that people are more likely to be thoughtful and conscientious when answering questions asked by someone who is in the room with them. It also provides an opportunity to ask follow up questions that can lead to important insights. Paper and pencil surveys are best left to situations when the sample sizes get bigger and one to one interactions become too time consuming to manage.
In sports sciences research in particular there is also a body of research that relies on mixed methods and draws upon full open ended interviews with athletes who talk about what they are experiencing. Qualitative methods such as this are not generalizable (i.e., we cannot say that themes from one group of interviews are representative of all athletes in a given sport), but they can provide insight into what to look for when we are setting out to measure how clutch works. For example, in their study of the 2004 Chinese Taipei women’s Olympic archery team a group of researchers were able to collect comments about the training protocol that both fed into their experimental design and created a feedback loop with the coaches who used the information to fine tune and individualize their training.
Where does clutch come from?
Empirical research into clutch performers, and by extension chokers, typically does two things. First, it must establish that clutch performances exist. While members of a given analytics community typically do not find much supporting evidence for clutch performance social scientists looking at sports such as archery, basketball, and golf almost always do. To be completely blunt, the question of whether clutch exists is not particularly interesting and if that was the extent of the research question it would not really worth using up time and/or grant money to answer. The second and far more interesting question is where does clutch come from, and going down this road ultimately leads researchers into an exercise of mapping out how clutch works.
The problem with clutch performances in sports is that they should not exist at all. Athletes are paid a great deal of money to keep themselves in shape, eat right, workout, manage their sleep schedules, etc., in order to operate at peak efficiency. We know that pressure can have a negative impact on performance so choking can be understood relatively easily. But when athletes turn pressure situations into a greater than usual performance where does that extra oomph come? To put it another way, clutch performance is a lot like when Pac Man eats a power pellet except information about what that power pellet is and where to find it is not given to us when we start playing the game.
Social scientists and sports scientists who research clutch in sports typically try to isolate psychological traits that are associated with clutch performance. Traits such as confidence, self-consciousness, implicit knowledge, and anxiety are typically looked as researchers begin to map out how clutch works. For example Mark Otten, who is probably the most well known researcher looking at clutch, produced the following Structural Equation Model based on data from an experiment he ran on free throw shooting:
There is a lot going on in this model, but for purposes of this article we can limit our focus to the links between expertise, sport confidence, self confidence, implicit knowledge, perceived control, and performance under pressure. The way this works is being very good at a sport is associated with being confident in that sport which in turn leads to a general sense of self confidence. As athletes train and grow confident they rely more on implicit knowledge (AKA “implicit motor learning”) which essentially means they start to play automatically without having to think about every action. This fosters a sense of control over situations within the sport which in turn leads to the ability to perform well under pressure. Perceived control is typically associated with clutch performers, so players who can run the gauntlet of developing expertise and confidence to the point where they feel in control during stressful situations will usually be the ones that can “raise their game.”
Researchers who use this type of approach typically find that when the pressure mounts the chain of factors listed above may break down for some athletes. In particular, those who have not gained the confidence to do things automatically typically find themselves focusing on minutia when they are in stressful situations in what is commonly described as a shift from implicit to explicit knowledge. For example, a rookie playing in his or her first NHL game may find himself overthinking things such as where to be on the ice, and some players describe the anxiety associated with this leading them to “squeeze the stick so hard it practically turns to sawdust.” Liao and Masters found that anxiety leads to self-focused attention which in turn led to poor performance under pressure. Following this line of thought researchers began to focus on helping athletes overcome tendencies to choke through interventions to lessen the impact of anxiety such as use of imagery , and techniques designed to increase a sense of control such as quiet eye training (i.e., pausing and fixating on a target or goal to enhance focus during aiming tasks).
While this line of research is very useful as far as it goes it provides a much fuller account of choking than clutch performance. More recently researchers have tried to fill in the gaps by looking at how different psychological traits lead chokers and clutch players interpret pressure situations very differently. Qualitative research that draws upon interviews with professional golfers and mixed methods research on netball players show that clutch players are narrowly focused on the task at hand and see pressure situations as opportunities to succeed while those who let pressure get to them focus on not failing which leads them to focus intensely on normally automatic movements in order to try to avoid making a mistake. Although there are very few studies to date that go very far in this direction, it certainly is an intriguing possibility that one source of clutch performance rests in the ability to focus on the opportunity such moments present, and push hard towards the goals as well as the rewards that go with achieving them, without any thought of failure in mind (e.g., “I will be a hero”).
I can’t really relate to extended debates over clutch players. I get that sometimes random point patterns sometimes lead to point streak during the playoffs that are not really associated with over “clutchiness.” However, I am skeptical that you can really objectively tease out whether someone is clutch using publicly available hockey data. When pressure situations occur, like playing late in a playoff game with one team up by a goal, coach on each side would put out players who he trusts to play well under players so there would be several players on the ice on each side who are supposedly clutch. There is also the matter of the goaltender who may make a clutch save to cancel out a clutch play to set up a goal. When Tiger Woods was making a 30 foot put on a Sunday at The Masters he did not have to deal with another clutch golfer trying to tackle him to the ground as he did so, and there was no one standing in front of the hole trying to kick the ball away.
But even if you could miraculously isolate one player’s clutchiness based on in-game event data I still think doing so largely misses the point in at least two important ways. First, the “elevating performance” standard is largely a crock. Considering the negative potential impact of pressure on performance I would be thrilled to have players on my team who can simply maintain their usual standard of play when pressure begins to mount. Second, the biggest bang for the buck for hockey teams is probably identifying who chokers are and figuring out interventions that can help those players. In fact, a lot of social science research, and almost all sports science research, goes down this road and focuses on identifying ways of helping players who typically choke to better cope with pressure. I have written about player selection models (which makes up most of hockey analytics) versus player improvement models before and I think there is typically a lot of potential to improve a team by helping players who struggle get a bit better.