Prediction Analysis ... how did Elo go?

I know you've been waiting on the edge of your seat for this analysis, asking yourself, how well did ELO do so far? Oh my gosh, I can't wait any longer to see Dave's predictions for the conference championships. Well, get excited, cause here it is! Disclaimer: the delusions of grandeur that I have about my blog following are real.

 Image: nhl.com

Image: nhl.com

New York and Tampa Bay! My initial Elo predictions had the Rangers duking it out with the Canadiens for supremacy of the Eastern conference. I had Tampa Bay as third in the East, but the deciding factor in the Tampa/Habs series turned out to be the superhuman play of Goalie Ben Bishop, who in many ways out-played Vezina trophy winner, Carey Price of the Canadiens.

Interestingly, the goalies in the Rangers/Lighning matchup (Lundqvist and Bishop) are currently ranked 1 and 2 for goaltender Elo. That means they are contributing +0.84 and +0.78 goal difference per 60 minutes, respectively, after accounting for team, and opposition quality. That makes sense when you consider that teams take an average of around 30 shots per game, and Henrick and Ben have been saving 0.944 and 0.931 all season long, compared with league average of around 0.91. 

The cool part is that the Elo rating can tell that goalies are good without even looking at their save percentages! The process of constantly modifying player ratings according to their plus/minus outcomes works for goalies just as well as skaters.

We can compare individual players in a new way, with the metric I call Corsi-Elo:

Corsi-Elo is just like Elo, except it is a rating for adjusted shot +/-, rather than goal +/-.

A rating of 1.5 is set as the league average for both Elo and Corsi-Elo. As many hockey bloggers love to point out - shots are a good way to measure possession. I'm definitely not a Corsi preacher, though I do believe that comparing Corsi-Elo to Elo can provide valuable information about team's and player's styles. Let's look at all the players on the Rangers (NYR) and Lightning (T.B) in terms of Elo and Corsi-Elo:  

 NOTE: Goalies are not included in shot-based metrics, so they are given league average Corsi-Elo of 1.5 in the plot above.

NOTE: Goalies are not included in shot-based metrics, so they are given league average Corsi-Elo of 1.5 in the plot above.

In general, there is a reasonable relationship between Elo and Corsi-Elo for T.Band NYR players, though many players maintain consistently different Elo and Corsi-Elo ratings. For example, Martin St Louis is a player who tends to create quality chances over quantity, so he has a high Elo, but low Corsi-Elo. Whereas, Kieth Yandle is a player who dominates the possession stats, but does not have the goal difference to justify it. Yandle is well known for his aggressive puck movement and fore-checking, which occasionally leads to odd-man rushes and quality scoring chances against his team.

 Image: pucksofafeather.com

Image: pucksofafeather.com

Chicago and Anaheim! The last teams standing in the strong Western conference. Anaheim, like New York and Tampa Bay, have benefited from solid goaltending. Frederik Andersen is currently the seventh ranked goaltender in Elo and is contributing around +0.45 goals per 60 minutes. Conversely, Chicago has managed to make the Western conference final in spite of relatively mediocre goaltending. Crawford is contributing -0.05 goals per 60, while Darling is at +0.1. Individual player Elo and Corsi-Elo ratings for this matchup are shown below.

 Elo and Corsi-Elo ratings for players on Chicago and Anaheim.

Elo and Corsi-Elo ratings for players on Chicago and Anaheim.

We can see in the plot above, that Andersen's Elo is much higher than Crawford and Darling's. In spite of this, with Kane and Timonen returning from injuries, Chicago is stacked with skaters who are able to outscore and outshoot most opponents. With all that quality on the ice, this should be an amazing series to watch!

OK, so what are my predictions for the conference finals and Stanley Cup? Once again - I used team-Elo ratings, weighting players ratings by their average time on ice. I then 100 thousand Stanley cups, as I did in the previous post, to determine the likelihood of each team advancing. The table below shows the results.

 Comparison of Elo before playoffs - and after round 2, and comparison of my original probabilities and the probabilities as they stand going into round 3.

Comparison of Elo before playoffs - and after round 2, and comparison of my original probabilities and the probabilities as they stand going into round 3.

So there you have it. Not including any results from the conference final round, I have New York beating out Anaheim for the Stanley Cup! The major story here is the huge increase in Anaheim's team Elo since the playoffs began. The answer to this is actually quite simple. The Ducks were predominately using goaltender John Gibson before the playoffs began, and then switched to Frederick Andersen. Since goalies play 60 minutes each game and Andersen's rating is around 1.6, compared to Gibson's 1.5 - this had a large influence on the team's overall Elo. This goalie-fueled boost in Elo tips the scales in favor of the Ducks for this matchup!

So far, team Elo was able to pick 8 of 12 games correctly. Who will advance out of the East and will it be a goaltending extravaganza as we expect? In the West, will Chigaco's dominant skaters be able to make up for their lack of elite goaltending? Elo says no, but as we know - it's the cup and anything can happen!