Introducing WNBA RAPM

Oct 14, 2015; Minneapolis, MN, USA; Minnesota Lynx forward Maya Moore (23) dribbles in the third quarter against the Indiana Fever at Target Center. The Minnesota Lynx beat the Indiana Fever 69-52. Mandatory Credit: Brad Rempel-USA TODAY Sports
Oct 14, 2015; Minneapolis, MN, USA; Minnesota Lynx forward Maya Moore (23) dribbles in the third quarter against the Indiana Fever at Target Center. The Minnesota Lynx beat the Indiana Fever 69-52. Mandatory Credit: Brad Rempel-USA TODAY Sports /
facebooktwitterreddit

While the NBA has seen a surge of advanced stats and a whole universe of new metrics and stats over the past few years, the WNBA has remained mostly unexplored. Most WNBA stats compiled are basic ones, and the advanced sections have old metrics borrowed from the men’s game, like Win Shares and PER. This is unfortunate because the WNBA has been given too little attention for basically its entire existence already, and the public sports analytics community has long had the power to help rectify that.

Thus, I’m enthused to share one of the basic building blocks of advanced stats in men’s basketball: RAPM, or ridge regression adjusted plus/minus, for several seasons of the WNBA. Plus/minus is a way to measure a player’s value independent of traditional stats, giving people a gauge of impact without relying on basic box score stats which can often be misleading or tell an incomplete story.

More from Nylon Calculus

For those unfamiliar with the WNBA, there are some important differences between the league and the NBA. There are fewer games and fewer minutes per game — 34 and 40, respectively. Consequently, the data set is slightly smaller in general. But there’s one difference I want people to understand before the unveil the RAPM numbers — there’s far more variation in the league at the player level. For instance, the top PER scores in the NBA rarely clear 30, and the top one is a shade under 32; however, there are six such seasons in the WNBA over 32 with the top one at 35, even though the league has significantly fewer seasons.

There’s generally a greater spread in player ratings, but the scale isn’t too extreme. Also, for better results, I calculated the RAPM in three-year chunks weighed by recency. You can view the results below where 2015 RAPM, for example, means seasons 2013, 2014, and 2015 were used with 1-2-3 weighing. The seasons end at 2006 because that’s the last year where a full three-year span was available. I only had usable play-by-play lineups going back to 2004.

For anyone unfamiliar with RAPM, it’s an estimate of a player’s net impact per 100 possessions. It relies on play-by-play data to measure how a team performs when a player is on and off the court, and then uses statistical techniques to try and account for the quality of opponents and a player’s own teammates. It is set so that an RAPM of 0.0 represents an average level of production.

Version: 1.0

The titans of WNBA RAPM so far are Maya Moore, Tamika Catchings, and Lauren Jackson. All three players have won an MVP award — they should be no surprise. Moore’s dominance in her rookie season is a little surprising, but her team went from being one of the worst in the league to one destroying others on the way to a title with a margin of victory over ten points in the regular season. This is the point of plus/minus: measuring player impact without box-score stats. It’s all about how a player’s affects her team’s ability to outscore opponents.

Unfortunately, the data is scarce and lacking. There were a lot of errors to handle and issues to overcome, so if you see any mistakes — a player might be listed twice, for example, because the name isn’t consistent — alert me on Twitter and I’ll try to fix it as quickly as I can. Right now I don’t have access to the 2016 season or anything before 2004, but that could change. Expect some improvements in the future.

Next: Highly plausible 2016-17 win projections

When plus/minus was first introduced, the concept helped revolutionize the way people thought about the NBA game. The WNBA has been neglected, but with several seasons of RAPM publicly released we could start a revolution there too. Advanced analysis is sorely needed.