clock menu more-arrow no yes mobile

Filed under:

Cracking the NET: Inside the NCAA’s new men’s basketball metric

What’s all this then?

NCAA Basketball: NCAA Tournament-First Round-North Carolina State vs Seton Hall Peter G. Aiken-USA TODAY Sports

The NCAA Evaluation Tool (NET) is replacing the RPI as the main ranking system for the NCAA Tournament. It’s supposed to be a secret formula developed by Google and approved by the NCAA. It’s been puzzling people since it first appeared in late November and I’ve been trying to figure the puzzle out. Well, I have the basic picture figured out, using only five variables. Here’s how the ACC teams are doing. The best teams are at the bottom-right of each of these five plots.

Q & A:

Q: What are these and how are they calculated?

A:

Efficiency - Net Efficiency. It’s (your points divided by your possessions) minus (your opponents’ points divided by your opponents’ possessions). “Possessions” are estimated by field goal attempts minus offensive rebounds plus turnovers plus 0.475 times free throw attempts. Unlike KenPom rankings, this is not adjusted by quality of opponent. Every point against Mount Saint Mary’s counts!

Margin: Average margin of victory, limited to 10 points a game and 1 point for overtime. For example, if you win a game by 44, lose another by 5, and win another in overtime, your margin is (10-5+1)/3 = 2.

AdjWinPercent - Adjusted Winning Percentage. This gives more weight to road wins and home losses, and less weight to home wins and road losses. For example, the calculation for a home win, a home loss, and a road loss would be (0.6+0+0)/(0.6+1.4+0.6)=23%. This is 25% of the RPI formula.

OWP - Opponents’ Winning Percentage. The average winning percentage (not adjusted) of your opponents, excluding when they played you. This is 50% of the RPI formula.

OOWP - Opponents’ Opponents’ Winning Percentage. The average of opponents’ OWP. This means you can’t benefit from playing weak teams who play even weaker teams. This is 25% of the RPI formula. Read more here.

(Note that only Division 1 games count. You could lose to Rust College and it wouldn’t affect your NET.)

Q: Why did you pick these metrics? How good are they?

I picked them because the NCAA told us that’s what they’re using.

The only difficult part is figuring out the “Team Value Index”. I included the factors from the RPI because I think the NCAA is afraid to come up with anything too different from what they already had. It’s not perfect but it’s most of the puzzle pieces.

Using nothing but a “line of best fit” with these five factors, I can estimate NET rankings within an average of 14 spots. That may or may not sound like a lot, but an example is picking Colgate at 151 instead of 137.

It’s difficult to say what factors are the most important because they’re all so darn correlated. Whenever you tend to score more points than your opponent, you probably have a good efficiency, a good margin, and a good win percentage. Sure, it helps to play a tough schedule too, but it’s not as important as it used to be.

Q: How is this working out for the Wolfpack?

A: So far, extremely well! We’ve spent most of the season beating up on bad teams. That means we have one of the worst “Opponents’ Winning Percentage” in the country. With the RPI, that would rank us around 160. But in the NET, a few 30-50 point wins really help you to the top! Look at how efficient we are! We’re right up there with Duke and Virginia! This might not impress KenPom or the human voters, but it sure impresses the NET!

Believe it or not, our “Opponents’ Opponents’ Winning Percentage” is actually pretty good. Basically, even though nobody’s impressed that we beat Maryland-Eastern Shore, we get some credit that the Hawks also played good teams like St. John’s, Georgetown, and North Texas (only one loss, to Oklahoma). Meanwhile, Duke’s OOWP takes a little hit for playing Kentucky because Kentucky played several bad teams like Monmouth, Tennessee State, and North Dakota.

As fun as it is to have the top 20 ranking, we’ll get what we’ll deserve in March. Our winning percentage will get worse, our OWP will get better through conference play, and our efficiency and margins will go down, but we will still be buoyed by the big wins at the beginning. Hopefully we can continue to at least beat the bad teams. Besides Carolina, the best team we place twice is Boston College.

Q: Can we use this knowledge to our advantage?

A: If I really thought I knew some big secret that would benefit the team, I would be bugging Kevin Keatts and Debbie Yow on Twitter instead of talking about it here. Regardless, here is some obvious advice:

  • Run up the score whenever possible. Yes, the Margin is capped at 10 points per game, but the Efficiency is unlimited. I’m not saying that it’s worth being unsportsmanlike or risking injury, but the math says that every point counts, on offense and defense, every second of every game.
  • It’s better to schedule bad teams that play good teams than bad teams that play bad teams.
  • Avoid scheduling tough games at home. Losing at home hurts your Adjusted Win Percentage much more than winning helps it. And if the opponent doesn’t finish in the top 30, it’s not even a Quadrant 1 game.
  • Consider scheduling away games against teams in the 76-135 rank area. It’s fairly winnable, it can boost the Adjusted Win Percentage, and it counts as a respectable Quadrant 2 game.