College Basketball Skill Curves

    Welcome back! Today I decided to look at skill curves for various college basketball players. What exactly is a skill curve you might ask? Unfortunately, I can't take credit for their creation as Dean Oliver discusses them in his 2004 book Basketball on Paper. The basic concept is to compare a player's offensive performance relative to how much is demanded of him when he's on the floor. This is achieved by comparing a player's offensive rating (a complicated formula that estimates how many points a player contributes for every 100 possessions) to his usage rate (the percentage of plays that a player is involved in). In his book, Oliver creates these graphs for NBA players only, so I decided to try to extend the idea to the college game.

     For example, here's what this type of graph looks like for Marquette's Markus Howard on a per game basis. This graph as with all following graphs is fitted with a trend line with varying levels of accuracy. In Howard's graph we see the data points are quite sporadic, but the trend line is still decent. We'll see later examples where this isn't the case. Also, in Oliver's book, he determines a year-by-year average O Rating for the NBA. In the absence of a predetermined average for college basketball this past season and the current limitations with how I'm getting my data, I think a rough estimate is probably somewhere around 105.

    As you can see, as Howard's usage rate increases, his offensive rating stays relatively the same. This is pretty unexpected as you would assume that as the player is tasked with doing more for the team, his offensive performance would start to decline due to fatigue, shooting inefficiency increasing with the volume, etc. While Howard's graph doesn't show the expected decrease, it doesn't mean that we can just assume that increasing Howard's usage rate up to say 70-80% will yield the same production. However, it does confirm what many already believe to be true: he can be a very effective high-volume offensive player. 

    As mentioned earlier, this is a case where the trend line comes out looking pretty strange, and suggests that Toppin's offensive rating steadily declines until his usage rate hits somewhere around 28%, and then suddenly his offensive rating gets briefly better. This isn't necessarily true, but the trend line does show the expected decline unlike Howard's above. This suggests that coach Anthony Grant was wise not to let Toppin's usage rate creep too much above 35% to preserve Toppin's incredibly productive and efficient season. This is emphasized by the fact that Toppin only had four games with an O Rating under the estimated average of 105 and two of those games were barely under this mark.

    After looking at two star players for their respective teams, I thought that it would be good to look at a more role-player type starter. Davide Moretti came to mind because I wanted to see the impact of his decision to leave Texas Tech to go pro. He certainly was valuable to the team as he had an O Rating above 105 in over half of his games. However, we see that his optimal usage rate is less than 25%, and so it shouldn't be too difficult for Tech to bring in a replacement with similar production at this less-demanding usage rate. It's certainly easier than a team like Marquette trying to replace Markus Howard whose optimal usage rate was much higher. Another reason for Tech fans to not be too upset is the fact that we see a downward trend line. It doesn't appear that Moretti was in for a breakout senior season offensively because his offensive performance declined as he was relied upon more.

    Unlike Moretti who doesn't look like he would've seen an offensive boost this coming season, we move into bench players and see that Alex O'Connell seemingly could be ready to make such a jump. Ignoring the two huge outliers on the right, O'Connell's O Rating generally improves as his usage rate goes up and peaks around 25% which is backwards. Like Toppin and Moretti (Howard is a weird exception), we should see O'Connell's O Rating steadily decline as his usage rate goes up, but that's just not the case. My guess is that it's because whenever O'Connell's usage for a game was over 20%, he likely played a lot of garbage-time minutes where he was actually able to do things on offense and help his O Rating. Regardless of reason, I'm curious to see how he would fare consistently having a usage rate of around 30%. Unfortunately, I don't see this happening next season even in spite of his transfer to Creighton given how strong the Blue Jays project to be. The only way we might get the chance to see if this skill curve is a fluke or not is if O'Connell can't get a waiver, and after sitting out a year sees a larger role awaiting him.

    Overall, these skill curves aren't the most informative because we're limited to the roughly 30 games that each player competed in this season. Players who missed games are even more difficult to assess with these graphs. Compare this to the NBA's 82 games and skill curves clearly paint a better picture for NBA data. Another downside is that Basketball on Paper is from 2004, leaving O Rating to be a bit dated. However, I think that skill curves still have some use for college basketball in that they told us what is generally assumed such as Markus Howard being a strong offensive asset despite being relied upon so much, and also some surprising information like Alex O'Connell potentially ready to see a vastly expanded role. I think that these skill curves could be particularly valuable in finding players ready for a breakout, as well as those who should see less usage. I know that the players that I selected for this article didn't necessarily show these trends (except possibly O'Connell), so if readers have suggestions for who to analyze next, my Twitter is @cbb_statistics. 

    Thanks so much for reading if you made it this far! If you enjoyed this article, be sure to check out some of my other ones such as my one on left versus right side layup percentages!