The Most Important Metric that Determines a Game’s Outcome

There are many metrics that fans use to describe a basketball game: points, field goal percentage (FGP), 3-pointer field goal percentage, free throw percentage, assists, steals, blocks, turnovers, rebounds, and many, many more.

My question is this: “What’s the most important metric that determines a basketball game’s outcome?”

Luckily for us, the answer is rather quite obvious. Every basketball game boils down to which team has scored more points. Therefore, the most telling metric is points scored. By comparing points scored by a team and its opponent, we can say which team has won with 100% certainty.

However, the question might not be so obvious if I asked about the second most important metric related to a game’s outcome. What are some other metrics that are associated with wins? Would it be the number of assists? Rebounds? Blocks? Steals? Maybe the field goal percentage?

I’ve set out to answer this question by analyzing regular season NBA games data since the 1995-1996 season.

For this study, I decided to look at the following metrics:

  • points scored
  • points scored from 2-pointers
  • points scored from 3-pointers
  • points scored from free throws
  • points scored from fast breaks
  • points scored “in paint”
  • steals
  • assists
  • blocks
  • defensive rebounds
  • offensive rebounds
  • rebounds
  • turnovers
  • fouls
  • estimated possessions
  • estimated offensive efficiency
  • field goals attempted (FGA)
  • field goals made (FGM)
  • field goal percentage (FGP)
  • 2-pointer field goals attempted (2-pointer FGA)
  • 2-pointer field goals made (2-pointer FGM)
  • 2-pointer field goal percentage (2-pointer FGP)
  • 3-pointer field goals attempted (3-pointer FGA)
  • 3-pointer field goals made (3-pointer FGM)
  • 3-pointer field goal percentage (3-pointer FGP)
  • free throws attempted
  • free throws made
  • free throw percentage
  • percentage of points scored from 2-pointers
  • percentage of points scored from 3-pointers
  • percentage of points scored from free throws
  • percentage of points scored from fast breaks
  • percentage of points scored “in paint”

For each metric, I obtained the metric’s values for two opposing teams. I then “retroactively predicted” each game’s outcome based on a simple comparison of the two values. Something like this: “If Team A recorded a higher Metric X than Team B, then predict that Team A has won the game,” where Metric X could be any of the metrics listed above.

I could then calculate the accuracy of these “retroactive predictions” using the following formula:

\begin{aligned} Accuracy = \frac{\text{Number of Correct Predictions}}{\text{Number of Total Predictions}} \end{aligned}

This “prediction accuracy” of a metric is a proxy for the metric’s association with winning, where a high accuracy suggests a strong association.

Not surprisingly, if we use points scored to make retroactive predictions on game outcomes, we’d predict teams who recorded higher points against their opponents would have won the games. And this yields 100% prediction accuracy, or 100% association rate. In other words, having a higher number of scored points is 100% associated winning games.

Using this method, I compiled the results of various metrics into the following table:

Metric Accuracy Number of Predictions
Points Scored 1.000 51972
Offensive Efficiency 0.964 51972
Field Goal Percentage (FGP) 0.797 51740
Field Goals Made (FGM) 0.782 48532
Defensive Rebounds 0.740 34196
Assists 0.727 48920
2-Pointer FGP 0.721 51676
3-Pointer FGP 0.670 50990
2-Pointer Points Scored 0.662 48886
2-Pointer FGM 0.662 48884
Rebounds 0.650 49790
3-Pointer Points Scored 0.624 46944
3-Pointer FGM 0.624 46944
Free Throws Made 0.624 49396
Blocks 0.615 45956
Points Scored “In Paint” 0.614 33868
Fouls Forced 0.608 48096
Free Throws Attempted 0.601 49832
Steals 0.599 46806
Fast-Break Points Scored 0.595 34396
Turnover Forced 0.589 47722
Free Throw Percentage 0.562 51328
% of Points from 3-Pointers 0.544 51930
% of Points from Fast Breaks 0.525 36224
% of Points from Free Throws 0.514 51938
3-Pointer FGA 0.501 49438
Offensive Rebounds 0.472 48174
2-Pointer FGA 0.464 50054
% of Points “In Paint” 0.460 36208
Field Goals Attempted (FGA) 0.459 49670
% of Points from 2-Pointers 0.441 51938
Estimated Possessions 0.430 51608

According to this table, the metric (other than points) that is most highly associated with winning is offensive efficiency, which measures how well a team converts a possession into points.

\begin{aligned} \text{Offensive Efficiency} = \frac{\text{Points Scored}}{\text{Number of Possessions}} \times 100 = \text{Points Per Possession} \times 100 \end{aligned}

This was interesting because having a higher offensive efficiency was associated with wins 96.4% of the time, far more than having a higher field goal percentage, which was associated with wins 79.7% of the times.

In addition, while having more blocks, steals, and forced turnovers was positively associated with winning (61.5%, 59.9%, and 58.9%, respectively), having more defensive rebounds and assists showed profoundly stronger associations with wins (74% and 72.7%, respectively).

Moreover, as expected from my earlier blog post, having more offensive rebounds was more often associated with losses (52.8%) than with wins (47.2%).

Lastly, having a higher percentage of points scored via 2-point shots was more often associated with losses (55.9%) than with wins (44.1%). However, having a higher percentage of points scored via 3-point shots was more often associated with wins (54.4%) than with losses (45.6%). This is somewhat consistent with our finding from a previous blog post that teams that shoot more 3-pointers stand a better chance of scoring than those that don’t.

So how does this help us? This information is useful because it provides a method to focus on metrics that matter. When we have hundreds of metrics to look at, being able to focus on metrics that truly matter gives us an advantage.

You May Also Like

About the Author: Howard Song

I’m a data practitioner by day, a web developer by night, a semi-competent swimmer, an active basketball player, a collector of cool ideas, an aspiring entrepreneur, a college dropout but a lifelong learner, and a self-professed nice guy. I love all things basketball, data, programming, and entrepreneurship.

Leave a Reply

Your email address will not be published. Required fields are marked *