College Basketball Rankings and Predictions
Let’s take a look at what the experts and mathematical are predicting for NCAA basketball games.
We’re not going to re-invent the wheel and create our own ranking model. There are too many smart people already doing this.
We’re going to look at what all the college basketball and betting experts are saying. We’re going to look at what the computer models and algorithims are saying. And then we’re going to come to our own conclusions based on past performance.
It should be a fun ride deciphering what our experts are saying that’s different from NCAA and Vegas rankings. And who knows, maybe you can make a little money along the way.
Basketball rankings vs predictions
Let’s start by defining some terms, but it’s going to be important going forward.
Sometimes people use ‘rankings’ and ‘predictions’ to mean the same thing.
Here, rankings are used to make predictions:
NCAA basketball rankings: These models look at what teams have done in the pasts (usually games in the current season, but especially early in the season, may include last season’s games) to come up with a ranking. Usually, the two main factors looked at are wins, margin of victory, and strength of opponents.
NCAA basketball predictions: This is more what Vegas does, and what we’ll do here. It’s predict who’s going to win games. The ‘NCAA basketball rankings’ are a big part of this, but then you have to look at other factors: home court is the biggest, but some models include things like time since the last game, injuries, momentum, and what’s on the line.
I’m not the expert
It’s worth mentioning again that I’m not the expert.
There are many mathematicians, college insiders, and gambling experts that I respect. I think it’s nearly impossible for anyone to get an edge on them, and if someone does, it’s not going to be me.
What we’re doing here is looking at years of predictions, and pulling out the people and computer models that are consistently the best. Then, we’re combining all these rankings and picks to come up with our own conclusions.
You could say we’re analyzing the analysts.
Then, we’ll report what our research is showing. We’ll focus on how it’s different than popular national rankings or gambling lines.
A sample of the ranking and prediction models we are looking at
There are somewhere between 5 and 50 NCAA basketball predictions we look at each week. The amount we look at depends on how much data we need, but it’s usually closer than 10. If there are wild numbers (numbers don’t agree), sometimes we look at more.
We’ll start with one you’ve probably heard of… KenPom Ratings. This is a rare ranking that only focuses on college basketball. KenPom uses statistical analysis. Offensive and defensive efficiency (points scored per 100 possessions) and strength of schedule are the main inputs to determine rankings of teams. Then, they mix in home court advantage to come up with game-by-game predictions.
KenPom crunches a bunch of stats. Things like luck, the tempo a team likes to play, and more random facts like how teams do after fouling late in the game. The key stat you’ll look at to determine the quality of a team is ‘AdjEm.’ It tells you how much the team would win (positive number) or lose (negative number) to an average opponent on a neutral court.
One important thing to point out is that KenPom predictions don’t take into account less mathematical items (or at least less data to use in calculations) player injuries, motivations, delayed team flights, etc.
Maybe one you’ve heard less about is The Power Rank, but it’s proven it’s worth looking at weekly. It’s a model created by Dr. Feng, a Stanford graduate PhD in chemical engineering. Okay, chemical engineering isn’t exactly sports. But it’s the same mathematical skill set, and his passion is sports. Either way, I find him to be better than most at predicting college basketball games.
For college basketball, the model looks at offensive and defensive points per possession, strength of schedule, and uses statistical modeling to spit out the predicted margin of victory. The NCAA basketball rankings shows each team’s predicted points they would win (or lose if a negative number) against an average opponent.
Another well-respected NCAA basketball ranking is Sagarin. Jef Sagarin is 1970 MIT graduate who has been creating computer sports prediction models for many years now, and his accuracy is amazing. His calculated basketballs spreads are nearly a perfect 50/50 on each side, which means it’s nearly spot on. Unfortunately, Sagarin is less revealing on how he calculates his rankings than other popular (it’s thought to be influenced by the chess rating system called Elo), but I guess that doesn’t matter if it’s consistently an accurate model. To predict games
An important note on Sagarin is there will be multiple different rankings. ‘Gold mean’ and ‘predictor’ are two slightly different models he uses. He also includes a ‘recency’ rating, which like it’s name, but it rates more recently games–this is different than a lot of models. All of these numbers go into his ‘Rating,’ which would be like an average of these three methods.
The Rating can be used to predict the score of a game. Team A with a 90 rating vs Team B with an 80 rating–Team A is expected to win by 10… on a neutral site. He lists the ‘home advantage’ (usually around 3) and that number is added or subtracted from the margin of victory depending on whether the team is home or away. In our example, if Team A was home they would win by 13, and if they were away, they would win by 7.
We take this and also see what our favorite college basketball experts and insiders are saying. Then, we filter out the noise and laser in on the trends.
Final sports predictions
The mathetmatical models above are surprisingly good at predicting college basketball games. But they have two big flaws:
- Most only take into account past games
- They don’t do well when there’s not a lot of data (earlier in the season or weird situations like new players coming back from injury)
This is where I think expert do a good job filling in the gaps. These people have connections inside college basketball teams from trainers to influential alumni. And they just know they sport.
So what the experts say can matter, but you have to read between the lines. Many times they’re like Jim Cramer on Mad Money and shout winning teams. This is mostly for show. But sometimes you’ll get insightful information like John Smith has looked like he’s avoiding jumping off his left foot, or little other insights that could impact a game.
This is where listening matters, and can sometimes give an edge to use on top of the models.
So let’s see where this all takes us. We’ll look at all the prediciton models, experts, and insiders, and come up with our own thoughts on where we think Vegas gets it’s wrong.