The flaws of the plus/minus statisticNovember 19, 2012, by
If you’ve been frequenting this blog for a while, then you know that we aren’t exactly fans of the plus/minus statistic. It’s useful in some cases, especially when measuring some prospect’s defensive value when they are in a shutdown role (Sam Noreau and Peter Ceresnak come to mind here). But at the NHL level, when other metrics are readily available, the plus/minus stat has very little value.
One of the main problems seen with plus/minus is that it is a counting stat that relies heavily on ice time. When ice time is involved, it skews the numbers –positively or negatively– for those that simple play more. Those playing on high scoring teams –specifically when matching up against weaker lines– generally have higher plus/minus ratings. The best example of this being skewed is with the Washington Capitals. Not exactly noted for their defensive prowess, players like Alex Semin (+36 in 2009-2010) and Mike Green (+39 in 2009-2010) had outrageous ratings.
Another problem is that it doesn’t account for the quality of competition being faced. Ryan Callahan finished last season with a -8 rating, but no one ever suggests he’s a poor defensive player. Marc Staal finished with a -7 rating. Ruslan Fedotenko (-7) and Brandon Prust (-1) also didn’t have favorable ratings. But you don’t say they are bad defensive players.
Those five are all given ice time against top quality competition, and expected to shut them down at even strength. But if you’re looking solely at plus/minus, then Marian Gaborik (+15) was one of the best defensive forward for the Rangers.
So what stat do we use to try to counter the clear faults in the plus/minus stat?
Well, the guys at BTN use a metric called +/- per 60, which eliminates the ice time factor and gives every player an even playing field. It doesn’t address the quality of competition or quality of teammates playing, but it’s definitely a more fair metric than raw plus/minus.
To my knowledge, there is no metric that accounts for QoC and QoT in addition to eliminating the ice time factor from plus/minus. The closest metric that combines them all is GVT, but there are also additional stats used in the GVT formula. Personally, I prefer the combination of GVT, RCorsi (which we use here frequently), zone starts, and QoC to get a full picture of a player’s defensive value.