Thursday, December 14, 2006

Statistics are for losers....

How many times have you heard that? I'll wager a lot. And, I'll also wager by people who think numbers are merely characters sewn onto the backs of jerseys.

Statistics are the revealed performance of individuals and teams. There isn't anything inherently "winning" or "losing" about statistics, they are merely "data." The trick is in knowing which of them serve to illuminate, and more important, which of them can be used to distort.

Today's lesson . . . plus/minus. Some folks look at that number and think that the player with a high "plus/minus" is a fine defensive player. Others might look at the same number and think that he plays in a good "team" scheme. Personally, I think both people are missing a bigger question . . . "how was that number arrived at?"

Let's take two players. Player "A" has a plus-minus rating of +14. You might think this player is a top scorer (he is). You might think he is an excellent defensive player (he's not...not bad, mind you; it's just not the strength of his game). You might think his club plays a successful system efficiently (well, no, the club is in the middle of the standings pack). It pays to ask oneself, "how was that number arrived at?" Well, this player is a +14, but he's accumulated +12 of that against one club. He is +2 against the rest of the league in 22 games played. Some guys are "killers" against a particular team, and this guy (in this instance) certainly is. Here we have one form of statistical bias, a tendency toward uncommon (by his own standards) success against a specific team.

Let's look at Player "B." He has a plus-minus rating of -2. You might think of him as an indifferent defender (he's no Selke candidate, but he isn't incompetent, either). You might not think of him as a top-flight scorer (you'd be wrong). You might think his club is an average-at-best defensive team (actually, they don't rise to the level of "average" -- they are in the lowest third of the league in goals against). So, what's his deal? Why the minus number? Well, he is a -2 for the year, but he had two especially awful games, both a -4. Against the rest of the league, he is +6 in 28 games played. Another form of bias -- outlying data.

Two guys . . . very different in their official numbers, but not so when one looks more closely at what makes up those different results. And, these guys play for teams that are tied in standings points this year.

Meet Player "A" . . . Sidney Crosby
Meet Player "B" . . . Alexander Ovechkin

When you look at numbers, ask yourself how the player might have arrived at them. And when you hear someone claim that "statistics are for losers," remember that all of those "statistics" are the bricks and mortar from which wins and losses are built.


JP said...

Averages are particularly susceptible to being misleading as the result of "outlying data," especially relatively early in the season. It'll take a goalie a month of good games to lower his goals against average noticeably, but only takes one shelling to raise it a half a point or more.

Look at Kolzig, for example. Take away just one of his 6 goals against performances and his GAA is around .15 lower.

The Peerless said...

The problem is variation -- the aggregate numbers (and, by expension, averages like GAA) often mask considerable variation from event to event. That's why I find Ovechkin's performance in his ten game splits so compelling. There isn't much variation. The kid is like a metronome. Wind him up, and he'll get somewhere around 13 points every ten games.

Insanity said...

I think you need to adjust one of the comments regarding "defense". It isn't team defense, it is line defense. Or more importantly, time in offensive zone.

Neither Malkin nor Crosby showed excellent defensive zone coverage (couple of turn overs there), but they maintain great puck control when they are on the ice and don't give up many opportunities.