I have been interested in the hockey stat plus-minus (PM) for some time now. It suffers from many well documented problems. A lot of those problems, though, are people upset when a player has a PM stat over a short period of time (one game, one week) that doesn’t reflect what seems appropriate for how good everyone thinks that player is.
In any case, scaled plus-minus (SPM) corrects some problems with PM, and makes no attempts to fix others. It is still based on PM, so the problem of where shifts start, and what lines a player is on are still prevalent. But how it is considered in the broader context of judging the strengths or weaknesses of a player suddenly become useful. For example, SPM approaches a constant value – the player’s true value – over time, instead of steadily increasing or decreasing. Also, if a player has a PM of +3, it is hard to tell if the player is above average or not. It is much more clear with SPM: SPM within the range of -1 to +1 is considered average.
Read more in part 1, a discussion of the changes (and a tiny bit of math – the relevant wikipedia page on random walks is mostly about drunk people and has pretty pictures/gifs), here.
For those looking to jump to how SPM affects the way we look at sports, click here.