Fun with Stats Episode 1: Why the Statgeeks Were, and Still Are, Wrong about the Minnesota Wild

As we all know by now, the Wild is in free fall mode.  They are two losses away from the “Fail for Nail” campaign.  Panic has nearly set in along the streets of St Paul, meaning passive aggressiveness as risen a full 2% across the state.

Several weeks ago, as the Wild were winning game after game, we were treated to article after article about how the Wild’s winning ways were “unsustainable.”  Corsis and Fenwicks and PDO’s and Offensive Zone Starts were trotted out to explain just how bad the Wild were, even as they continued to climb up the leaderboards.  They were wrong.

Instead of seeking to explain why their precious stats were failing them, the Statgeeks were penning articles predicting the implosion of the team.  Of course, everybody and their grandma could tell you a team that was near the bottom of the league in scoring and severely lacked talent and depth up front was going to fall back to Earth.  That’s a no brainer.  Still, Statgeeks stuck their necks out with several key stats to tell us Wild fans just how bad our team was going to be.

And even though the Wild have lots seven straight, the Statgeeks are STILL wrong.

The Mikko is not impressed with your PDO

The Wild’s season thus far can be split into three neat segments:

1. Stumbling out of the gate, going 3-3-3 to start the season
2. Winning 17 of 21 games to shoot up the NHL standings
3. Dropping seven straight games, going 0-5-2

So let’s look at some stats over each of those time frames.  Goals per game, goals against per game, shots per game, and shots against per game are self explanatory.  Power play and penalty kill stats as well.  Save percentage is obviously one minus goals against over shots against.  Corsi is the net of total shots, missed shots, and blocked shots for versus against.  Fenwick is the same thing without blocked shots.  I didn’t do PDO because I didn’t want to go through game by game (again) and grab all the special teams numbers.  Someday I will.

(all stats courtesy of NHL.com)

During the first 9 games (0.500 win%):
G/G: 2.22
GA/G: 2.56
S/G: 25.3
SA/G: 28.4
PP%: 13.9%
PK%: 74.2%
Save%: 91.0%
Corsi: -8.22
Fenwick: -4.56

To start the season, the Wild’s numbers were pretty mediocre.  Struggled to score, not great defensively, poor special teams, and negative Corsi and Fenwick indicated they were getting “outchanced” by the other teams.

During the winning streak (0.810 win%):
G/G: 2.81
GA/G: 1.95
S/G: 26.0
SA/G: 33.7
PP%: 18.4%
PK%: 89.9%
Save%: 94.2%
Corsi: -18.52
Fenwick: -12.81

During the streak, the Wild picked up the offense somewhat, tightened up significantly defensively, and improved special teams quite a bit.  Yet the Corsi and Fenwick were EVEN WORSE than before.  These stats are already calculated on a per game basis, so it’s not a cumulative error.  You can also clearly see that the shot differential was much wider.

During the losing stretch (0.143 win%):
G/G: 1.29
GA/G: 3.14
S/G: 28.9
SA/G: 31.4
PP%: 9.1%
PK%: 79.2%
Save%: 90.0%
Corsi: -6.57
Fenwick: -5.00

As the team lost, clearly they fell apart both offensively and defensively (although a 0.900 save percentage isn’t atrocious) and special teams have fallen as well.  But the Corsi and Fenwick are better than ever and that shot differential has tightened up quite a bit.  Hurrah!

Stats are boring.  Here's a monkey playing tennis!

We have two options here:

1. We can believe the Wild came back to Earth, are defying all kinds of luck and logic, that the goaltending has been below average the last couple weeks, that the pucks “just aren’t going in” for the team, and that they’re showing a lot of improvement and will break out of the slump at any point in time.

2. We can throw shots, Corsi, and Fenwick out the window.

In future episodes, I’m going to tell you why #2 is the correct answer.  I’ll have a lot of evidence backing up my claim.  Over 8,000 games worth of evidence.  And if I can find access to a database with special teams statistics, I’ll analyze those as well.

But for now, I’ll give you the short and sweet version.  Why do the Wild “defy” statistics?  Because shots, and any derivative of shots (Corsi and Fenwick) are terrible predictors of success.  The Wild and other teams who play defense are perfectly content to let the other team take a shot through a half dozen bodies from the point or boards.  That’s a low scoring chance.  More importantly, the Wild DON’T want you taking a shot on a 3-on-1.  That’s a high scoring chance.

And no, playing defense doesn’t rob you of offense.  Lacking offensive talent robs you of offense.  Boston plays a pretty tight system but scores plenty of goals.  Stamkos has done quite well for himself under Guy Boucher.  Crosby isn’t exactly playing in a run and gun system.  

The bottom line is this: in hockey, you win by outscoring your opponent.  Look at the goal differential during each of the three sections of the season.  The Wild were getting slightly outscored while playing 0.500 hockey (which is really not 0.500 hockey because of the charity point), they were outscoring their opponents significantly when winning, and are getting crushed by opponents during the losing streak.

Yes, it’s obvious.  But the Statgeeks haven’t figured that out yet.  Let’s wait a little longer to tell them.

4 comments:

  1. It certainly doesn't help that the two top lines were destroyed by injuries.

    ReplyDelete
  2. That would be a rational explanation, not allowed in the Statgeek world!

    ReplyDelete
  3. "In future episodes, I’m going to tell you why #2 is the correct answer. I’ll have a lot of evidence backing up my claim. Over 8,000 games worth of evidence. "

    You are correct. That is why most "statgeeks" use Corsi or Fenwick Tied / Close to account for changing styles of play, as shown in the first link. If a team gets out to a quick 3 goal lead, they aren't going to be playing too aggressively the rest of the game. For evidence of this, please click the second link for an article by the talented Gabe Desjardins.

    Also, lest you believe that teams don't value these numbers, you should know that Gabe is being paid by 3 teams to offer his insights and analysis and give them a different perspective (3rd link). Here's a good quote from that article: “I've seen people use Corsi to make trades,” he said. “I'll put it that way.”

    Also, in case you believe that people who think stats have value simply don't "watch the games", there's one final link for you to check out at the bottom.

    Cheers,

    http://objectivenhl.blogspot.com/2011/03/loose-ends-part-i-predictive-validity.html

    http://www.arcticicehockey.com/2010/4/13/1416623/corsi-and-score-effects

    http://www.theglobeandmail.com/sports/hockey/the-moneypuck-revolution/article2178766/

    http://www.arcticicehockey.com/2011/4/6/2081929/the-luck-involved-in-scoring

    ReplyDelete
  4. I would love to see the reaction of statgeeks when they realized that "advanced stats" were made up by casinos in order to help people gamble away money by assuming these predictors filled in the blanks.

    ReplyDelete