Predictions and Texas Basketball

Each fall it happens -- I try to predict what will happen in the coming basketball season. It is an unavoidable annual occurrence -- no matter how hard I try to avoid it -- for anyone who writes about the sport. What is going to happen?

I am here to tell you the truth. I don't have a clue. Shaka Smart and the Texas Longhorns may be good, may be bad, or may be something in between. My history at predicting how things will go is terrible.

I really don’t know how good Jarrett Allen or Andrew Jones will be as freshmen. I can tell you some things about their games, but I don’t know exactly how things will play out on the floor against D-I competition. I can’t tell you how much offseason improvement sophomores Kerwin Roach, Eric Davis, and Tevin Mack have made, even if I can potentially propose some things that I would like to see improved. I can’t predict exactly how incoming grad transfer Mareik Isom’s game will translate to a different conference and new style of play. All I can say is that he can really shoot.

In the coming weeks I will be making some predictions. I haven't given up on that. What I have given up on is caring all that much on how close my predictions come to the truth that will play out between mid-November and early April. I just don't care about being right about this stuff anymore, or at least in the same way that I used to be, because now I see it as pretty much a hopeless exercise.

It is curiously simple to make predictions that work fairly well in aggregate (the general trends of college basketball are typically easy to anticipate) and yet impossibly difficult to nail outcomes for specific teams. We know in general that teams that were good last year and return most of their important players are likely to be good again. We know that teams with more experience and more talent will do better than teams without these things. These basic principles work well for making general predictions coming into the season of the sort you will find at places like kenpom.com.

But we also know there are many individual exceptions to these trends and that the odds of a particular team being an exception -- either positively or negatively -- are large enough to frequently render these sorts of predictions useless at the individual team level.

Why is this so hard to do? At first glance, predicting the quality of an individual basketball team seems straightforward, at least at the conceptual level. We have very good tools for quantifying at the team level precisely what made a particular team good or bad. To turn this information into a prediction it should then become a "simple matter" of looking at these data and projecting them into the future, given the inevitable changes each team goes through each season.

But it is not a simple matter. I might figure, for example, that the Texas Longhorns are likely to improve in perimeter shooting and in forcing opponent turnovers in this coming season compared with last season. I also might project that Shaka Smart's team will turn the ball over more often and will be weaker on the defensive interior than what we saw last year. If I can quantify the magnitude of these changes, I can put all of it into some simple mathematical machinery and come up with a projection of the quality of the team's offense and defense in the coming year.

That all seems great, but there are really two problems with the approach. First, quantifying the magnitude. Will Texas turn the ball over in 16 percent of its possessions next season? 17 percent? 18 percent? What may seem like a small difference in turnover percentage turns out to have a meaningful difference in total offensive efficiency. The story is the same for three point shooting percentage, two point shooting percentage, and rebounding rate. Being off in the projection of each of these by seemingly small amounts can lead to quite different predictions for both an offense and a defense. So even if I know enough to correctly predict the general trends in these measures, it is hard to convert that into accurate projections of total offensive and defensive quality.

And then there is a second problem with this approach. Even if I can accurately estimate the quality of a team, this does not map directly onto how a team will actually perform in terms of wins and losses. Teams of more or less equal ability over the course of a season playing against similar levels of competition can end up with very different records at the end of the year simply because enough games will end up being decided by a small number of possessions where minor random events (a missed call, a bad bounce of the ball, or a halfcourt heave at the buzzer) decide outcomes. Some very small number of teams will be good (or bad) enough that these sorts of events are less likely to matter, but for most teams this just isn't the case.

There is always the less mathematical approach. I can simply look at the roster, close my eyes, and give you my best guess as to what will happen. This approach is easier and more intuitive, and when applied to a single team probably not meaningfully better or worse that the one that involves a spreadsheet. This intuitive approach is perhaps better because it is less work.

In the coming month there will be predictions made — both by people focused locally as well as those with national scope — about Texas basketball. Some will use math while others will use the gut. I have made some predictions (they make a modest appearance in Smart Texas Basketball 2016, which will soon be released as an ebook); others will make them, too. We will probably all be wrong — although hopefully we can be wrong in a manner that is at least interesting and informative.

I don’t know anything about anything. What are your predictions for the season?