Controlling the Pace Doesn’t Help NBA Teams Win Games

By: Mohin Banker (@mohinthelawn)

As the 2016 NBA Playoffs progress, you are going to see analysts and commentators compare and predict matchups by looking at a team’s pace and contrasting it with their opponent’s pace. For example, going into the playoffs, the Golden State Warriors played at the fastest pace of 101.6 possessions per game, which is drastically different from the Toronto Raptors’ 92.9 possessions per game - among the lowest in the league. Pace can tell you a lot about the way a team plays, which determines a team’s identity. And so, a common mantra in basketball analysis is that the team that controls the pace of the game, will win the game. In theory, this makes sense; if you can tailor a game to your style of play, then you might gain an advantage to win the game. But, analytics suggest the opposite.

What is Pace?

Pace is defined as the number of team possessions in 48 minutes, or the length of a regulation game. The higher the pace statistic, the faster a team plays. But, what is a possession?

Screen Shot 2016-05-24 at 10.47.26 PM.png

In this formula, we use

  • FGA - Field goal attempts
  • FTA - Free throw attempts
  • ORB% - Fraction of team offensive rebounds to total possible offensive rebounds
  • TOV - Player turnovers

In simple terms, a possession ends when the other team gains possession. A possession is when a team scores, when a team misses a shot and fails to get the rebound, or when a team turns the ball over.

Figure 1

From a box score, we cannot find the exact number of possessions because free throws (which end a possession) can come in bunches of one, two, or three. So, we estimate possessions by discounting free throw attempts and offensive rebounding percentage to account for free throws that do not end a possession. The discount coefficients come from Basketball Reference’s provided formula, although other sources use different coefficients. For that reason, our calculated paces can differ slightly from these sources, but it should not affect our analysis very much.


Pace Over Time

Using this formula, we calculate the pace of individual games and compare the pace of a game to teams’ season average paces. We calculate the paces for games and teams from the 1983-84 to 2015-16 seasons because Basketball Reference does not provide offensive rebound numbers in games before the 1983-84 season.

Why would a faster-playing team correspond to a faster pace? Since pace is time-adjusted, a team would have a higher pace if:

  1. The team selects or allows shots early into the shot clock
  2. The team is poor at grabbing offensive and defensive rebounds
  3. The team forces and commits many turnovers

These attributes correspond to pushing the ball quickly up the floor, small lineups, and weak defense. From Figure 1, a moving histogram of pace in the league over time, you can see that teams in the late 80’s played at a much higher pace than the following decades. The 80’s-era NBA was largely characterized by a run-and-gun style of play - a strategy that deemphasized defense and set plays in lieu of fast break points. Doug Moe was renowned for applying this strategy to the Denver Nuggets at the time, and coached the team to three of the top five fastest seasons from 1983 - 2016.

Fastest Teams since 1983-84

Team Season Pace (Poss. per game)
Denver Nuggets 1990-1991 113.38
Denver Nuggets 1983-1984 110.25
Denver Nuggets 1984-1985 107.29
Golden State Warriors 1988-1989 107.25
Denver Nuggets 1988-1989 107.19

Slowest Teams since 1983-84

Team Season Pace (Poss. per game)
Cleveland Cavaliers 1995-1996 81.58
Cleveland Cavaliers 1996-1997 82.10
Detroit Pistons 1996-1997 83.91
Cleveland Cavaliers 1994-1995 84.08
Miami Heat 1998-1999 84.51

Pace slowed down across the league in the mid to late 90’s, as half court offenses developed with a heavy emphasis on long isolation possessions, and many teams adopted a large, hulking frontcourt. But, in the early 2000’s, the NBA removed its restrictions on zone defense, and enacted rules preventing hand-checking. As a result, teams moved away from isolation plays and more towards small-ball to capitalize on guard penetration and increased fast-break opportunities, led in large part by Mike D’Antoni’s Suns in the mid 2000’s.

Measuring Pace Control

When two teams play a game against each other, we assume that both teams have an equal control over the pace. Therefore, we would expect the game pace to be the midpoint of their season average paces. The difference between the actual game pace and the expected pace would be our measure of pace control, with its sign indicating

Screen Shot 2016-05-24 at 11.04.23 PM
Figure 2

whether the home or away team is in control. A positive pace control means the home team controls the pace, while a higher magnitude means that team had more control over the game’s pace.


As an example, we use the Golden State Warriors’ loss to the San Antonio Spurs on March 19th, as shown in Figure 2. At home, the Spurs defeated a strong Warriors team by a convincing 8 points, while holding Golden State to their slowest pace of the season. The Spurs’ season pace (using our formula) was 93.22 possessions per game, while the Warriors averaged 98.85 possessions per game. So, we expected the game pace to be the average of their paces, or 96.04 possessions. If the actual game pace were greater than 96.04 possessions, we would say that the Warriors controlled the pace; otherwise we would say the Spurs controlled the pace. The actual game pace was a snail-like 87.12 possessions, so we can calculate the pace control using the difference between the expected pace and the actual pace value: 96.04 - 87.12 = 8.92 in San Antonio’s favor.

One problem with measuring control of pace in a game is that controlling the pace may be more valuable in certain matchups. If similarly paced teams play each other, pace control becomes less significant because both teams play using similar styles. The game’s expected pace is very close to each team’s pace.

Does Controlling the Pace Help You Win?

Our dataset includes every regular season NBA game since 1983, a sample of 37,123 games. We want to see if our metric, pace control, can explain a team’s margin of victory. We examine the distribution of our two variables to see if there are games with extreme values for margin of victory or pace control that could influence a linear regression.

Screen Shot 2016-05-24 at 11.07.18 PM
Figure 3

In the red histogram of pace control, we can see that the distribution is unimodal and symmetric, but skinny, which implies that most of the games fall close to the mean. However, while we can’t see the actual bars, there are a few games with a pace control lower than -20 or higher than 20. Games like this are highly unusual, so we will ignore them in our analysis. In the blue histogram, we see that margin of victory is also unimodal, but is slightly asymmetric and skewed slightly to the right. We can attribute this to the inability of games to end in ties, as well as the influence of home teams’ home court advantage.

I chose to exclude games with extreme pace controls by removing games with pace controls outside of (1Q - 1.5*IQR, 3Q + 1.5I*QR) where 1Q and 3Q are the first and third quartiles of our pace control variable. IQR is the interquartile range. These boundaries corresponded to games with pace control values between -12.40 and 12.30 possessions. Using this method, we can identify 398 games which would be considered outliers, and exclude them from the dataset. Many of these games come from the 80’s and 90’s, when the pace of a game was more volatile. Another characteristic of these outliers are very high scores, often having combined scores well over 200 (i.e. a game between the Boston Celtics and the Indiana Pacers during the 1990-91 season, with the a final score of 152-132, Celtics).  Since these games have such extreme pace control values, they have high leverage, meaning they can be more influential when we construct our regression. As it turns out, including these points don’t change the results of our findings, but it is safe practice to exclude games that we would expect to have an unusually large impact on our analysis.

Here, I’ve conducted a linear regression on the remaining set of games:

Screen Shot 2016-05-24 at 11.08.45 PM

Screen Shot 2016-05-24 at 11.09.41 PM

We conduct a hypothesis test on the coefficient for pace control predicting margin of victory with a null hypothesis that 1= 0. The test yields a p-value of 0.2261. Under the assumption that pace control does not predict margin of victory at all, there would be a 


Screen Shot 2016-05-25 at 8.27.50 PM
Figure 4

22.61% chance of finding a coefficient of our magnitude or more extreme. We cannot conclude that the true coefficient is not 0 for a reasonable level of confidence. But, even if the true coefficient was a non-zero value, it assuredly has a negligible effect on a team’s margin of victory.  So, in general, controlling the pace does not mean a team will be any more likely to win. But, maybe we are looking at it from the wrong perspective; many of the matchups are between teams with similar paces, a situation where we would not expect pace control to have much of an effect. So, we could exclude games where the average paces of the two teams are very similar.

I chose to remove games in which the difference in teams’ paces over the course of the season was fewer than 4 possessions per game. The number is relatively arbitrary, but it is about the difference between the Spurs’ and Warriors’ paces this season. Therefore, we can hopefully exclude most games between “similar” teams. However, creating a similar regression and conducting a hypothesis yields a p-value of 0.111. We still have no reason to conclude that pace control can be a predictor of margin of victory. Even so, its predictive power would be small enough to ignore.


Pace is a valuable statistic for identifying styles of play, and understanding how teams played differently in different eras of the NBA. But, basketball analysts perpetuate the myth that “controlling the pace” - forcing an opponent’s speed of play to align with the team’s ideal playing style - is key to winning games. We conclude, from a statistical perspective, that controlling the pace doesn’t have any relationship with how well a team fares against its opponent.

A possible explanation for the lack of an effect is that playing closer to one’s style does not translate to outscoring the opponent. In Spike Lee’s documentary Kobe Doin’ Work, Bryant comments on how both teams know exactly what the other team is going to do. Both teams know each other’s plays, game plans, and defensive schemes. It all comes down to execution.


If you look back at the previously mentioned Warriors-Spurs matchup this season, you will find that the Spurs did manage to slow the game down. Their disciplined transition defense prevented the Warriors (the league leaders in fast break points per game) from pushing the ball down the court for quick, easy points. As a result, the game was played primarily in the halfcourt. Curry’s low field goal percentage that night, coupled with the Spurs’ hot shooting, gave the latter team the victory. But, just a month later, when the Warriors faced the Spurs again in San Antonio, the teams played at a similar pace. But, this time, the Warriors won convincingly. The real difference was that Curry made his shots even though the game was played in the halfcourt. It was not about who dictated the speed; it was about the team that executed their plays and hit its shots at a higher clip.
Ultimately, the fact that pace control has so little influence over a team’s margin of victory is surprising. When a team controls the pace, that team is playing their preferred style of basketball, and we would expect that to be beneficial. However, we find that in reality, there is no discernable advantage - when it comes to winning games, what matters is simply which team is more talented, not how fast the game is.


3 thoughts on “Controlling the Pace Doesn’t Help NBA Teams Win Games

  1. Hi, I love the article but have a major qualm. The projected pace of a game is not the average of (team A pace,team B pace). I think a more realistic projection of a games pace is... team A pace + team B pace - League Average Pace. So if Team A plays at 100 poss/game and team B plays at 95 poss/game and league average poss/game is 95. I would project that game to be played at 100 poss. Not 97.5. The reason is because Team A was able to attain 100 poss/game while playing a variety of teams and play styles. They average 100 poss/game against the average opponent. If they then play the average opponent I would expect them to play at a pace of 100 poss.

    I would be very interested in seeing this redone when faster than average paced teams play slower than average paced teams.
    Re-think exactly how you want to predict the expected pace of a game and the pace control formula. Then see if controlling the pace had any effect.

  2. Hey,
    Great article! I would be interested in a followup to see if pace control had an effect within the gameflow itself, ie if margin of victory/time in game was correlated with pace of play control/time in game. The drawback here, is of course a massive amount of noise, but this could help vindicate the idea that if one controls the pace of play during stretches of time during the game, they are more likely to regain ground or increase their lead.

Leave a Reply