A New Stat: Shot Quality

Over the past decade or so, it seems like innovation has been a common theme in every facet of life. New technologies and ways of thinking are being created by the day, and with that our level of knowledge about a given topic grows. This not only includes, but is near perfectly reflected in the sports world. While it started with baseball, seemingly every sports team and league is looking for the next edge. The next statistical breakthrough. The next way to zig while others zag. And with March Madness on the horizon, I felt the time was right to explain that college basketball is no different.

Kenpom stats such as offensive and defensive efficiencies, as well as effective field goal percentage, have been around for a good chunk of time now. But thanks to the great minds at Shot Quality, I've discovered a new and interesting metric. The metric attempts to look at every single field goal attempt based on defensive proximity, and distance from the basket to determine the percent chance of a shot going in. Over 90 other variables go into the shot quality metric's calculation, which, along with other more detailed explanations of their statistics, can be found on the shot quality website. It also adjusts so that three pointers get valued slightly higher than they would, similar to how effective field goal percentage and true shooting percentage do. Because the stat is so new, it only has data since 2020. Additionally, I will only be focusing on 2022 data, as it is the only season in this dataset that was full-length from start to finish. All of this is to say: take the following information with a grain of salt.

All that being said, I wanted to first look at how well this correlated to offensive production as a whole. I decided to use adjusted offensive efficiency and correlating it to offensive shot quality. Here's what I came up with.

An r value of 0.89, as well as an r squared value of 0.79 is pretty great. Not perfect, but still shows high levels of correlation. Additionally, the website measures defensive shot quality by looking at the probability of an opponent making shots in the exact same way that offensive shot quality does. Here's a look at how the defensive shot quality correlated to adjusted defensive efficiency.

Here we have a 0.82 r value, and a 0.67 r squared. Not as good of a correlation as with offense, but a good correlational value nonetheless.

Now that we've seen the two main metrics shot quality produces, it's important to see how well these stats correlate to wins as a whole. Both offensive and defensive shot quality are combined into one over-arching metric simply called adjusted shot quality.

Definitely another solid correlation here, with a 0.83 r value and a 0.69 r squared. It seems as though the shot quality metrics are ones to take seriously, as they certainly correlate to wins and offensive/defensive production. However, there is definitely a little bit of error, specifically with the win percentage correlation, that must be discussed.

Shot quality is really a way to measure luck in college basketball. If a team takes a lot of low percentage shots and happens to make them, and they come away with a close win, that team was lucky to come away with a victory. Conversely, if a team got a lot of open looks that rimmed out, that team was super unlucky. Good offenses don't make shots, they create shots. And good defenses can't directly stop the ball from going into the hoop, but they can make it tougher on the offense by giving them less open looks. An example of this is on February 1st of this year, when Wake Forest visited Duke. Duke won the game 75-73, but while watching I couldn't help but think Duke snuck away with the victory. And sure enough, according to shot quality metrics alone, Wake Forest had an 89% chance of winning the game. The shot quality score was 82-67, with Wake Forest winning--a 17 point swing from shot quality metrics alone. These situations happen all of the time, and shot quality measures this impact on wins and losses directly. Alongside the actual win percentage, shot quality measures the shot quality win percentage, based on expected wins and losses from their own data. While Wake Forest may have lost at Duke, they won according to shot quality, which explains the difference between the two winning percentages.

It is key to note here that the luck is measured by subtracting the win percentage from the shot quality win percentage. For instance, if you have a win percentage of 80%, but a shot quality win percentage of 60%, you have a record luck of -20. Negative values indicate luck, positive values indicate lack of luck. Interestingly, when looking at last year's data, it seems as though teams with more luck are making the tournament, as displayed by the above density plot. Intuitively, though, this makes some sense. The luckier you are, the higher your winning percentage is, so the more likely you are to make the tournament. Obviously there could be other explanations, but that's just my theory. However, does shot quality luck catch up to you after you've made the big dance? Rephrased--do really lucky teams in terms of shot quality have their luck run out in the tournament?

What this plot tells me is there might be a slight edge to the luckier teams when it comes to making the sweet sixteen. However, the emphasis is on slight, as these density plots are very similar. This doesn't come to much of a surprise for me, however, as very few stats on their own, if any, are able to even somewhat accurately predict what goes on in march madness. It got its name for a reason, after all. It is complete madness. Absolutely random. Therefore I never expected shot quality to be any different.

Overall, I would say shot quality is a very interesting metric. While it has only existed for a short time, and thus is hard to draw solid conclusions from, it seems to be off to a good start in terms of correlating to production and wins of a program. And when filling out your bracket next week, I recommend you take shot quality into account. In 2022 and 2021, every final four team was in the top 26 in adjusted shot quality, and both champions were in the top 2 in that metric. It's a very interesting concept that has yet to become mainstream, but I believe it has a bright future in the world of basketball analytics.