With the clock ticking down on the Golden State Warriors' legitimate playoff hopes in Game 4 of their Western Conference semifinal matchup against San Antonio, the explosive Warriors' offense needed just one final basket to break a tie in the game and knot up the series.
Naturally they called a play in which Jarrett Jack pounded the ball for several seconds before tossing up an awful fadeaway jumper with a hand in his face that never had a chance.
The same scenario played out in the other Western Conference semifinal Game 4. With the Thunder and Grizzlies deadlocked in the waning moments of the game, Zach Randolph received the ball on the perimeter and took a horrid fadeaway shot that wasn't even close to send this contest to overtime as well.
Both the Warriors and Grizzlies came back to win in OT, but I've never understood why teams would be willing to live with such bad shots that would not pass muster in any other part of the game during the most crucial moments of the contest, and making matters worse these attempts often come out of a timeout.
If Jack or Randolph attempted those shots with seven minutes left in the second quarter, I would not be surprised to see them hooked from the game. Yet these are the plays their coaches draw up to win crucial games.
The one mitigating factor is that by ensuring you take an isolation shot at the buzzer, you greatly reduce the chance of a turnover and make sure you don't give the opponent one final shot at the horn. Those are important things, but perhaps they can be accomplished in a manner other than having your (usually) best player take a horrible shot at the buzzer.
It's clear from watching way more basketball than I should that teams shoot lower percentages in the clutch, and to find out how much worse I analyzed five seasons worth of regular season data from the NBA's media stats site (2008-09 through 2012-13). I decided against comparing it to the same situations during the playoffs due to the skew toward good teams and the small sample size of the postseason, particularly when it comes to clutch situations.
I broke up the data into three distinct groups: when the shooting team trails by three or less or is tied with 10 seconds or less remaining, when a team trails by three or less or is tied with 30 seconds or less remaining and non-clutch situations (determined by subtracting the 30 seconds data from total shooting percentage data). There are more clutch situations than just these two examples, but I figured these scenarios are the most critical for a team when it absolutely must score to win a close game.
I found it important to include both the under 30-second situations and the under 10-second situations because the latter case so often ends up in an isolation pull-up jumper whereas the former allows for a better chance to run actual offense. However, they do overlap in this dataset because I wanted to compare each scenario individually to the non-clutch situation. The under 10 scenario made up just 0.23 percent of all NBA shots over this period and the under 30 example 0.36 percent.
I discovered that teams shot 27.33 percent in the under 10 seconds scenario, 32.57 percent in the under 30 seconds scenario and 45.87 percent in the non-clutch situations. There were no major outlier years as each season played out pretty close to the averages.
Next I pulled out some notes from Dr. Lackritz's statistics class, and found a two-group proportion interval, which ended up being very tight due to the incredibly large sample size in each of the respective groups.
The difference between proportions between the non-clutch group and the under 10 seconds group was 18.54 percentage points with a confidence interval between 18.51 and 18.56. The difference between proportions between non-clutch and under 30 seconds was 13.30 percentage points with a confidence interval between 13.28 and 13.32. Finally, the difference between the 30- and 10-second groups was 5.24 percentage points with a confidence interval between 5.21 and 5.27. Because there are no zeroes in these intervals, these are significant findings at a 99 percent confidence interval.
Furthermore, this means teams shoot on average 40.4 percent worse in the final 10 seconds than they do in the previously defined non-clutch situations, and 29.0 percent worse in the final 30 seconds than non-clutch situations. They even shoot 16.1 percent worse in the final 30 seconds than they do in the final 10 seconds, which is particularly interesting because of the overlap between these two groups.
I sent these results to ESPN NBA analyst David Thorpe, the executive director of the IMG Academy Pro Training Center, and asked him what NBA coaches can do to be more creative with their crunch-time play calls.
"The key for coaches is this: create an action with your best player or scorer, then ask for the ball to find the most open guy," Thorpe said. "That leads to another factor in poor field goal percentages -- hero shots rather than moving the ball to the most open guy."
Thorpe also noted that the difference between 11-30 seconds and 10 and under is "a huge variable" because there is "more time to set screens, run decoys, read and react, etc."
Thorpe feels defenses are likely better than average in these crucial situations since teams can play their best defensive lineups regardless of an individual player's offensive deficiencies, so there is a "potential for better matchups."
Although the small sample sizes perhaps mitigate the usefulness of the following data, it is noteworthy that the 2008-09 and 2009-10 Hornets ranked first and second in under 30 seconds shooting percentage, after shooting 12-of-18 (66.7 percent) and 12-of-21 (57.1 percent), respectively. Those New Orleans squads were led by an elite point guard in Chris Paul who could get to the rim at will, nail jumpers and penetrate and kick with the best of them, so it leads one to believe that the Hornets would have been more apt than an average team to actually run offense in these situations.
On the flip side, the 2009-10 Indiana Pacers missed all nine of their clutch shots, the 2009-10 Pistons shot 2-for-23 (8.7 percent), the 12-13 Pacers went 4-for-31 (12.9 percent) and the 11-12 Bucks went 3-for-23 (13.0 percent).
Those 08-09 Hornets knocked down an uncanny 8-of-12 shots under 10 seconds within a bucket, and the 2009-10 Grizzlies were 7-for-13 (53.9 percent). The Derrick Rose-led Chicago Bulls hit 7-of-15 (46.7 percent) in 2011-12.
Perhaps the most noteworthy poor performance belongs to the Big Three-led 2010-11 Miami Heat, who hit just 1-of-18 clutch shots in the final 10 seconds. You might remember that was the Dueling Banjos year in which LeBron James and Dwyane Wade traded off taking bad isolation shots at the end of a contest. Apparently this strategy does not even work for the best players in the world. Miami improved to 6-of-13 the following year when it won the championship.
As for overall team percentages over the five-year period, the Hornets' 41.2 shooting percentage led the league, followed by Dallas (39.6), Denver (38.3), Atlanta (38.1) and San Antonio (37.8). The bottom five were Philly (25.0 percent), Milwaukee (25.2), Sacramento (26.7), Boston (26.8) and Minnesota (27.8).
The Hornets were only 9.6 percent worse in under 30-second clutch situations than they were normally whereas the Sixers were a whopping 45 percent worse, the Celtics were 43.8 percent worse and the Bucks 43.2 percent worse.
New Orleans led the under 10-second category as well by shooting 41.9 percent in such situations, followed by Memphis more than a standard deviation below the Hornets at 35.9, Denver (34.8), Cleveland (34.3) and Dallas (32.8).
Many familiar faces crowded the bottom of this list paced by Philly (19.1 percent), Boston (20.3), Detroit (21.4), Indiana (21.5) and surprisingly the Durant/Westbrook-led Oklahoma City Thunder (21.9).
The Hornets decreased by just 8.2 percent whereas the second-best Grizzlies were 22 percent worse in a clutch scenario with under 10 seconds to play. The Sixers were 58.1 percent worse and the Celtics 57.5 percent.
Last season many NBA players struggled in such situations, including Paul Pierce (2-for-11), Damian Lillard (1-for-9), Mo Williams (1-for-7), Paul George (0-for-7), Jordan Crawford (1-for-7), Deron Williams (0-for-6) and Rajon Rondo (0-for-5). Yet players are often better remembered for the clutch shots they make than all the ones they miss.
A future study could try to discover why certain teams shoot so much worse in the clutch and whether good shooting teams or poor shooting teams decline by more in the biggest moments. It would also be interesting to analyze whether isolation shots dramatically increase in the final seconds of a game and what teams shoot on those attempts in comparison to other shot types.
There are many reasons why teams shoot worse in the clutch, including the lack of fast-break opportunities in such scenarios, having to go up against a set defense often coming out of a timeout and the pressure of the moment as well as the explanations Thorpe offered. In addition, some of these teams likely were playing for the last shot in a tied game to prevent the opposition from getting one final look, which led to a worse shot than they would get normally.
However, the NBA clearly has a crunch time problem. By watching the games it surely seems like teams are taking poor shots in clutch situations, and these statistics strongly support that conclusion.
In clutch situations it would behoove teams to run their most creative plays rather than their least to free up an open player to take the shot instead of a star against a defense full of pressure. After all, it is quite the paradox that teams often take their worst shots when they need a basket most.