More evidence ± is the worst stat of all time

Since the Avs are done and there's not much to talk about, expect more short quickie posts like this one for the next couple weeks or so.

Shane over at Avs talk linked to TSN's off-season gameplan for the Avalanche yesterday and I, unlike Shane, took a minute to read it. It automatically gets a fail for not mentioning the best things the Avs could do... Fire Granato and hire a quality coach.

Other than that, the overview was pretty decent, certainly a lot more in-depth and smarter than most national, or even local, Analysis of the Avs. So I'm a bit hesitant to criticize it. But two paragraphs caught my eye:

Veteran sniper Milan Hejduk tied Smyth for the team scoring lead, but also posted the first minus rating (minus-19) of his career. The 33-year-old is entering the final year of his contract and would have some appeal if he was willing to be moved, but he's also one of the club's few proven finishers so if the Avalanche do unload Hejduk it could make the rebuilding process all the more challenging.


As the most adept puck-handler in the group, John-Michael Liles is supposed to be a cornerstone, but last year's career-low minus-19 didn't shine a bright light on his defensive prowess. He's not big, nor is he aggressive, so he's not likely to be a defensive stalwart, yet Liles' skating and puck skills should produce better results if he has a stronger supporting cast (in terms of goaltending and forwards).


*sigh*. This is the exact reason (±) should just be tossed out the window like a 6-month old carton of eggs. Anyone watching the Avs knows that Liles probably had the best defensive season of his career, and Hejduk was one of the Avs best defensive forwards all season. In fact I remarked all year how I was commending Liles improved defensive zone play. His, and Hejduks, (±) was clearly a result of inept coaching combined with poor teammate play and is not even close to an accurate representation of their season defensively.

(±) is a useless statistic for individuals, and should never, ever, ever be cited in an analysis for someone's defensive abilities.

15 comments:

  1. The article also pointed out that Stewie had a rookie-worst -18. This stat certinly does not paint an accurate picture of him. Also Mike Mottau from NJ is at a +24 (18th best in the league) scoring only one goal in the regular season. This stat seems as useful as a book to a Red Wings fan.

    ReplyDelete
  2. Plus/Minus is a useful stat for both scoring and non-scoring players. The unfair criticism usually comes at times when a player's rating is unusually lower than it normally would be. Mike Mottau is probably a +24 despite scoring only one goal because he prevented goals from being scored against his line, whereas Hejduk and Liles did not. Milan has never exactly been known for his dedication to defensive zone coverage, and most offensive d-men struggle in the +/- category because they take more chances.

    ReplyDelete
  3. check out the hawks and ducks plus minus for the season and see if you still feel the same way. kane was minus 2 for the season no where near anyone else. ducks moved 3 people with the worst plus minus and the deadline and look what they're doing now. it can definitely be useful if looked at in the right context.

    ReplyDelete
  4. Mike Mottau is probably a +24 despite scoring only one goal because he prevented goals from being scored against his line, whereas Hejduk and Liles did not.• Or because he was playing against the other teams worst players, while Hejduk and Liles routinely played against the best.
    • Or he had good line-mates throughout the season who scored a lot of goals while he happened to be on the ice
    • Or his line mates were solid defensively
    • Or the defense on New Jersey is superb, the goaltending is outstanding and no matter who they put out there is going to get a a solid ±

    And this is the problem with ±, there are so many variables that go into the stat that it loses all meaning. All Mike Mottau being +24 tells me is that he was on a good team. It tells me nothing about his defensive ability. It tells me nothing about his offensive ability. It tells me nothing about what kind of player he is and, most importantly, it tells me absolutely nothing about his effectiveness on the ice, because there's too many other factors that are not under his control at play.


    It can definitely be useful if looked at in the right context.
    And that's the problem right there... There's so much context that needs to be weeded out, that by the time you give it proper context you know everything you need to know about the player, making his ± useless.

    Also i'll point out that St. Louis went on a terror by not adding or subtracting hardly anyone, so the Ducks losing their three worst ± players doesn't necessarily mean that's why they are playing better. I'll bet dollars to doughnuts that they are playing better because: Francois Beauchimin returned from injury and Jonas Hiller has been playing spectacular.

    ReplyDelete
  5. The biggest problem with +/- is that, as a counting stat (as opposed to a ratio stat), it fails to accurately describe what it ambitiously purports to depict - poor defensive play.

    The biggest defensive liabilities on a team will almost never put up the worst +/- numbers, simply because they don't get a chance to. A fourth line player who sees less than half the TOI as a Top 6 forward, will inevitably be on the ice for fewer goals against.

    Should Tjarnqvist and Mackenzie be considered the Avs' best defensemen for being the only two with a plus rating?

    Additionally, the stats' dependance upon overall team performance, makes its use in comparing a player on a losing team to a player on a winning team, utterly meaningless.

    Was Paul Coffey the most defensively sound blueliner of the 80's, simply because he posted a +216 while with the Oilers?

    Obviously, most of the goals scored against Edmonton came with Coffey on the ice (he rarely sat down), but since he contributed to even more goals for, the number of goals scored against him remains hidden.

    If one is looking for a stat that describes defensive prowess, I'd suggest eliminating the plus side of the equation entirely, and simply reporting the minuses.

    ReplyDelete
  6. Of course, that would also ignore the quality of opponent factor, painting the guys who coaches employ to shut down opponents' best players, as some of the worst offenders. If your job is to shadow Ovechkin and Malkin game-in game-out, you'd be hard-pressed to keep those minuses down.

    Should Rod Brind'Amour be stripped of his back-to-back Selke Awards because he only went +8 and +7 during those two seasons? (or for his career -10 rating?)

    The only way to fully describe a player's defensive worth with a stat, would be to look at his minuses, AND his TOI, AND the quality of his average opponent on the ice.

    ReplyDelete
  7. Over the years, I have endeavored to develop a SABRmetric-style series of stats, including several that relate to +/-, or at least to the minus side.

    The most basic is simply to record every skater's goals against. Every goal against, powerplay included.

    Just as goalies' total GA is not widely reported, a more widespread stat would become a skater's GAA. This would make comparing fourth liners to top defencemen more useful than comparing Hendricks' +0 rating to Liles' -19.

    Using the same technique, every player would also be assigned a GF and a GFA, that would serve a useful purpose on its own, rather than lumping that measure into the convoluted +/- number.

    The SABRness doesn't come into play until you attempt to measure the quality of opponent on the ice. The details are too long to get into now, he says after making us read this friggin' phonebook of a comment.

    Perhaps I'll use this excuse to return to my desk and publish something - ANYTHING - for the first time this month.

    Look for my detailed description of the proposed "ultimate" defensive measure in the next short while.

    ReplyDelete
  8. You should look at behind the net, which has a GAA/20 and they have taken a stab at quality of teammates and quality of opponent. I'm not sure how to factor in their weighting there, but it's certainly light years better than ±

    ReplyDelete
  9. Yeah, I love his shit. You a friend of Hawerchuk?

    The ignoramus in me pictures the entire Bay area as one tiny village.

    Do you know John Smith? He also lives in Cali, so I figure you two have met.


    Perhaps I haven't been following BTN as fervently as I should have, and now I feel too embarassed to ask questions that have likely been answered already. Otherwise, I'd wonder aloud how to calculate the "quality of" numbers.

    My own measure simply relates an opponent's average GFA to the player's own GAA, but I have yet to introduce "quality of linemates" into the equation.

    If you are, in fact, a buddy of Hawerchuk's, do me a favor and assure him that I'm not as retarded as I may seem, should I "sack up" and ask for some definitions.

    ReplyDelete
  10. TT -- the quality numbers are based on +/-, i think.

    ReplyDelete
  11. 1)+/- like any stat should never be used alone to promote/denounce a particular player. It does give some measure of a player's effectivenss in in playing well on both sides of the rink, etc, but we agree that it can be misleading at times.

    ReplyDelete
  12. @ Faux:

    I don't think it gives any measure. Sure no stat alone should be used, but ± is so misleading that it's worthless, IMO. I guess, in other words, any usefulness that could come from ± could easily be shown with much better statistics.

    ReplyDelete
  13. What? There were Avs players with + stats? Well, slap me silly and let pigs fly out me arse.

    ReplyDelete
  14. At the same time, the year Foppa and Hedjuk had the insane +53 (or something) can't be written off as a product of which lines they play against–shut down OR top lines from opponents

    ReplyDelete
  15. @ wflan:

    That's because Foppa wakes up in the morning and pisses excellence.

    ReplyDelete