Background
College basketball holds a special place in my heart. My sister “snuck” me into Cameron Indoor Stadium posing in 10th grade when I posed as one of her graduate student friends. I was there for Jimmermania in 2011. My first published research paper was about college basketball TV audiences in the Final Four. My first major project at ESPN was to redo the college Basketball Power Index (BPI) and create Strength of Record (SOR). My algorithm is still what feeds those rankings on the NCAA selection committees team sheets.
It is no secret that college basketball has among the largest home advantages in all of sports. There’s many possible reasons why. From travel, rest, unfamiliarity of surroundings to more tangible ones like using a different basketball team to team to the uniqueness of college student sections. In the last decade student sections have gotten much more creative in their attempts to affect the outcome of opposing team shooters making their free throws.
Previous Research
A Duke medical student tabulated the miss percentage for games in the 2020-2021 and 2021-2022 seasons for various different distractions the graduate student section made behind the basket in the second half.
Arizona State is known for its “curtain of distraction” which rose to fame in 2015 and at that time had a surprise distraction of shirtless olympian Michael Phelps.
The Harvard Sports Analytics Collective did a study in 2015 about the Arizona State’s “Curtain of Distraction” and reported that visiting teams shot 60.6% from the free-throw line in the second half compared to 68.6% in the first half. Their estimate was that this distraction was equal to approximately 1.41 points per game. They reported that there was not a statistically significant difference between first half and second half free throw percentages due to the small sample size of their data.
Luke Benz, a current Harvard PhD student and author of the ncaahoopR package, revisited the Arizona State distraction in 2019 and found that there was not enough evidence to conclude that teams shoot worse against Arizona State than they do at other road games.
Surely with another 5 years of data one should be able to discern once and for all if Arizona State’s students have made a noticeable effect and if other teams have had similar impacts as well.
Data
I used data from the 2015-2016 college basketball season until the current season’s data (2023-2024). I excluded the 2020-21 seasons due to many arenas not having fans at their games due to the reaction to the COVID-19 pandemic. I grouped any home team that has less than 15 games in the dataset into one collective team.
I made sure to only look at true road games, not games where the visiting team is playing at a neutral site. Whether or not something is listed as neutral site is sometimes subjective, I used the neutral site tag that can be found via the ESPN.com API.
Methodology
Each of the previous methods take a slightly different approach to estimating a crowd’s affect on free throw shooting. Each has its advantages but with more seasons I will attempt to revisit the 2nd half analysis as well as estimate which home venues affect free throws throughout the game as opposed to only the second half.
2nd Half Student Section Effect
I filtered the data to teams that have at least 200 free throws attempted at their home court in both the first half and second half over the course of this data set. The top 10 in terms of largest drop-off in free-throw percentage from the first half to the second half can be seen in the table below.
I also plotted the histogram of the differences in free throw percentage from the first half to the second half by each home team. What is suprising in looking at first glance is that, on average, visiting teams shoot 1 percent better in the second half than compared to the first.
This is likely due to the effect of the end of game scenarios in basketball. Typically, if a team is winning the losing team starts to foul towards the end of the game. The winning team knows this will happen and does its best to get the ball into the hands of their best free throw shooters before these fouls occur. You can see this trend in the following plot, where the difference in 2nd half free throw percentage goes up when a team wins by about 5-10, which is typically the range of outcomes where the other team was fouling.
This doesn’t explain the entire effect as even teams that lose tend to shoot better percentage wise for free throws in the second half. There’s several possible explanations for this but my intuition was that as players got tired their free-throw percentages would decrease in the second half, which doesn’t appear to be the case.
To somewhat account for this second half effect I take that one percent in the second half off of a team’s free throw percentage when assessing the difference between the first half and the second half via the z-score and p-value.
