Introducing Wins Above Bubble

The method behind one objective measure of résumé strength

Thanks for reading the Her Hoop Stats Newsletter. If you like our work, be sure to check out our stats site, our podcast, and our social media accounts on Twitter, Facebook, and Instagram. You can also buy Her Hoop Stats gear, such as laptop stickers, mugs, and shirts!

Haven’t subscribed to the Her Hoop Stats Newsletter yet?

---

One of the most fascinating aspects of college sports is the fact that it’s often really hard to rank teams. The WNBA has 12 teams, and all 12 play a relatively similar schedule, so win-loss records are generally a fair way to order the teams for the playoffs. In contrast, there are over 350 NCAA Division I women’s basketball teams, and the difference between the toughest schedule and the easiest is gigantic.

Difficulty ranking teams doesn’t stop us from trying, of course. There’s the AP Poll, the RPI, the NET, the Her Hoop Stats Ratings, the Ken Massey Ratings — each of these systems attempts to rank teams in its own unique way. The AP Poll, for example, is a collection of media members’ opinions. Our Her Hoop Stats ratings use machine learning to try to predict which team would win if two teams played each other today.

There’s another system that has gained traction on the men’s side in recent years: Wins Above Bubble (WAB). Made popular by Seth Burn, WAB has received endorsements from analytics pioneer Ken Pomeroy, ESPN writer John Gasaway, and NBA GM Monte McNair. It’s even made its way onto men’s analytics websites. It’s time to add it to the growing list of tools in the women’s game as well.

We’ll debate the advantages and disadvantages of WAB in a coming article, but for now let’s take a closer look at what exactly this metric is. The basic explanation of WAB is that it is precisely what the name sounds like: the number of wins a team has above a typical bubble team’s expectation. In other words, it answers the question, “How many more wins does Team A have than the number of wins a bubble team would be expected to have against the same schedule?”

For example, if Team A is 14-2, and a bubble team would be expected to go 11-5 against Team A’s schedule, then Team A will have a WAB of 3. A negative WAB indicates a team that has won fewer games than what would be expected of a bubble team against its schedule; if Team A were instead 10-6 against that same schedule, its WAB would be -1. A notable feature of WAB is that it doesn’t take into account margin of victory. It isn’t attempting to predict future results, rather it’s analyzing the strength of a team’s résumé based purely on wins and losses.

So how is it calculated? The key is the “expected” part. There are a few steps to figuring out how a bubble team would be expected to do against a given team’s schedule. First, we need to know how many points each team scores or allows per 100 possessions, adjusted for strength of schedule. We can use our Her Hoop Stats Offensive and Defensive Ratings for this.

Next, we use those ratings to calculate each team’s probability of beating an average team. Bill James’ Pythagorean expectation formula can be used here. This formula began as a tool in baseball but has since been adapted for use in many other sports. For more of the nitty-gritty math behind it, check out the details here. We’ve found that an exponent of 10 works best for college basketball. We can use this formula three times — once each for home, road, and neutral-site games — by adjusting the offensive and defensive ratings accordingly. Home offensive ratings are typically about 4.4 points per 100 possessions better than road offensive ratings, so we’ll adjust the neutral-site numbers by 2.2 each way for the home and road numbers.

Take Indiana, for instance. If the Hoosiers played an average team on a neutral court, they would be expected to win about 95% of the time. That figure increases to 97% in Indiana’s own building and falls to 93% for Indiana road games. (Keep in mind that there are approximately 350 teams in Division I, so an “average” team is ranked somewhere in the 170s.)

Now that we have an expected winning percentage for each team, we’ll use log5 (another Bill James method) to generate win probabilities for each game of the season. For each game for each team, what we want is the probability that a typical bubble team would win that game at that location. For the four seasons for which Her Hoop Stats has data and an NCAA Tournament was played, the eight teams that were either the highest-rated team left out or the lowest-rated at-large team had an average rank of 40.88 in the Her Hoop Stats Rating. To simplify things, we’ll use the 40th-best team as our typical bubble team. This means that for each game, we will use log5 to calculate the probability that the 40th-best team by Her Hoop Stats Rating would have won that game.

The last step is to award or revoke credit to a team based on whether they won or lost the game. The bubble team’s win probability in a given game is the same as the expected value of wins in that game. If a bubble team has a 75% chance to win a game, that means it is expected to, on average, come away with 0.75 wins. A team that actually wins the game obviously comes away with one win, so it will have 0.25 more wins than what a bubble team would have been expected to earn. Likewise, a loss would be 0.75 fewer wins than a bubble team’s expectation, which would result in a -0.75 WAB for that game. After WAB is assigned in this manner for each game, all that’s left is to sum the WAB values for every game a team has played. That’s the team’s WAB for the season.

The full WAB rankings can be found here and will be updated regularly throughout the remainder of the season. Due to the uniqueness of this season and the high variance between teams in the number of games played, each team’s winning percentage above bubble is also included in column K. This is just the difference between a team’s actual winning percentage and the winning percentage a bubble team would be expected to have against the same schedule. It’s the same concept as WAB, but doesn’t penalize teams who haven’t played as many games and thus haven’t had a chance to accumulate wins. Here are this year’s current top ten:

There are several tools for evaluating and comparing the performances of different teams, and just about all of them have value in the right context. As we near Selection Monday and start having more and more discussions around which teams are the most deserving, WAB can be a useful piece of the conversation. Let’s give it a seat at the table.

---

Thanks for reading the Her Hoop Stats Newsletter. If you like our work, be sure to check out our stats site, our podcast, and our social media accounts on Twitter, Facebook, and Instagram. You can also buy Her Hoop Stats gear, such as laptop stickers, mugs, and shirts!

Haven’t subscribed to the Her Hoop Stats Newsletter yet?