The point of stats are to proxy or estimate how good a player is in a specific area of the game. Does a player average 30 points or 5 points a game, 12 rebounds a game or 2? Conclusions can be drawn based on how players perform relative to each other. One of the biggest problems with stats is small sample size. We have all seen it before, a player comes out of nowhere in the second half of the season on a bad team and puts up big box score numbers. Then in the offseason we all wonder if those performances were trustworthy or not. If the player only appeared in 25 out of 82 games, it’s hard to trust those results. The larger the sample the more comfortable we are believing in something. Five thousand 5 star reviews on amazon, I guess I will pick this foam roller over the one with two reviews. Stable Stats are an attempt to remedy this problem by making more predictive stats.

Here at Basketball Index we have raw versions of stats (3PT%) and stable versions (Stable 3PT%). Stable Stats work by adding a small amount of league average results to a player’s stats, thus stabilizing them. The idea is that a small amount of league average padding pulls everyone towards average, giving less extreme results in small samples. Because of this, Stable Stats are more predictive than raw stats when it comes to how a player will perform in the future.

A good example of how Stable Stats work would be Max Christie and Buddy Hield in 2023. Former Laker Christie shot 42% from three as a rookie, but he only took 62 in total that season. What does that tell us? If you are a Laker fan you hope it means you’ve added a deadeye to the roster. After all, he shot the same percentage as professional movement shooter Buddy Hield. The difference is Hield took 677 threes that season. If this data was presented to most basketball fans and they were asked which player is likely to shoot a higher percentage next season, almost everyone would say Buddy. They would make some sort of calculation in their head and weight Buddy’s volume as more predictive. The next season (2024) Max Christe shot 36% from three on 118 attempts and Buddy shot 39% from three on 586 attempts. Hield’s larger sample turned out to be more predictive. There’s multiple reasons for this. Over a larger sample shooting variance plays less of a role, meaning Buddy had both hot and cold steaks over the season. Averaging them together gives you a better estimate of his shooting skill. When Christie only took 62 threes in 2023 that could have been the hottest shooting of his life. Another reason is the larger sample gives time for the players shot quality to stabilize. It’s possible Christie was only playing in garbage time, or maybe his minutes matched perfectly with LeBron’s and he was getting nothing but amazing passes. If we look at both players Stable 3PT% in 2023 we would have had some built in context to start our analysis from. Buddy’s Stable 3PT% was 41% (42% raw), Christie’s was 37% (42% raw). The stabilization pulled both of them back towards average, but Hield’s large sample was barely affected, whereas Christie’s small sample moved a lot more. 

Another case for stable stats is not overreacting to hot or slow starts to the season. For the sake of example let’s use some made up numbers. Let’s say the padding in Stable 3PT% is 100 attempts at 36%. In the year 2049 LeBron James III starts the year hitting 7 of his first 10 threes, meaning he’s shooting 70% from three. No one thinks he’s going to shoot 70% from three for his career. His Stable 3PT% would be (43 for 110) 39%. This is a much more realistic number, and likely more accurate than the math we are all doing in our heads looking at 7 for 10.

Any stat can be stabilized and each one has its own unique padding value. Some stats like blocks stabilize much faster than 3PT%, (Meaning they need a smaller sample before they are trustworthy). Another reason stable stats are helpful is when you sort by raw 3PT% random players with small samples will litter the leaderboard. Even with a minute and attempt filter you can see the difference in the leaderboards below.

2025 Raw 3PT % Leaderboard – 250 minute min, two stable 3PTA/75

1. Dru Smith 53% 
2. Seth Curry 45.6%
3. Larry Nance Jr 44.7%
4. Zach LaVine 44.5%
5. Taurean Prince 43.8%
6. Ty Jerome 43.8%
7. Matisse Thybulle 43.7%
8. Craig Porter Jr. 43.7%
9. Vit Krejci 43.6%
10. Aaron Gordon 43.6%

2025 Stable 3PT% Leaderboard

1. Zach LaVine 41.9%
2. Taurean Prince 41.6%
3. Harrison Barnes 40.4%
4. Malik Beasley 40.3% 
5. Kevin Durant 40.2%
6. Keon Ellis 40.2%
7. Seth Curry 40.1%
8. Ty Jerome 40.0% 
9. AJ Green 40.0%
10. Grayson Allen 39.8%

The raw leaderboard has some ob vious flaws even with the filters. The point of looking at 3PT% is to figure out who’s good at shooting threes, and Stable 3PT% does a better job at that.

A common push back to Stable Stats is that they aren’t what literally happened. This is an understandable hangup. My objective in this article isn’t to stop you from using raw stats (I still use them), but instead offer a valuable resource when analyzing players.

Follow me on Twitter for more basketball analytics @taylormetrics. Thanks to Head of R&D at Basketball Index Krishna Narsu @Knarsu3 for making all of our Stable Stats and explaining how they work to me over and over.