Which Olympic Hockey Team Do The Analytics Favor?

[Bear of Bad News]

http://www.fiveforhowling.com/2014/1/7/5286390/which-olympic-hockey-team-do-the-analytics-favor

Spoiling the results, this would predict Canada/Sweden/USA as the medalists (in that order).

Completely GVT based analysis, which is the standard used at Hockey Prospectus. Answer seems reasonable to me.

Click to read more...

 

[TheDevilMadeMe]

http://www.fiveforhowling.com/2014/1/7/5286390/which-olympic-hockey-team-do-the-analytics-favor

Spoiling the results, this would predict Canada/Sweden/USA as the medalists (in that order).

Completely GVT based analysis, which is the standard used at Hockey Prospectus. Answer seems reasonable to me.

Obviously based off of NHL play on the smaller ice surface because there isn't a big enough sample of games of NHLers on the big ice.

Because of the difference in ice surface, I really don't think the analysis is particularly useful though. USA generally stinks on larger ice, and while Canada still does okay, they are less dominant themselves.

Still, an interesting read on how the teams would likely stack up if playing under NHL rules. If you think of it like that, even the decision to consider KHL players "replacement level" seems fair.

 

[hatterson]

Thinking of a GVT analysis brings up a couple of thoughts.

1.) Who is the biggest "snub" from a GVT standpoint and who is the "worst" inclusion? I suppose this question really only applies to the nations filled with NHL caliber players.

2.) Where would Canada's "B" team slot in from a GVT standpoint? There's clearly some incredible talent that had to be left off (First line could be Giroux, Thornton, MSL), but how does it actually stack up?


Also, in addition to the generic limitations of GVT mentioned in the article there's the extra consideration of ice size differences. Although I haven't seen any work that puts a specific quantifiable change on performance based on ice size, it seems to reason that certain skill sets would map better to a larger ice surface, specifically speed.

 

[wgknestrick]

About all that GVT is good for is comparing forwards as the D and goalie values are incredibly off relative to the forwards. Goaltenders get all the GVT credit of their defense in front of them, while the best defender in the league is equivalent to an average 2nd line forward. Don't D men play significantly more minutes than forwards too?

Every season there are about 6 goaltenders "deemed" more valuable than Crosby/Malkin/Stamkos by GVT. IF this was even remotely true, we'd see goaltenders coveted with equivalent, elite salaries.

I love the premise of GVT, but think some work needs done on the "assumptions" it makes in the formulas. The top 4 teams are much closer than GVT would have you believe.


Biggest snubs would be James Neal and worst decision is Orpik over anyone....Yandle esp.

 

[TheDevilMadeMe]

GVT definitely overrates goaltenders and underrates defensemen over single season samples, that is for sure

 

[soireeculturelle]

Time zones.

No North-American team has medaled at the Olympics since 1998, and you might as well pick the finalist teams based on who's playing closest to home.

98: CZE - RUS (Nagano)
02: CAN - US (Salt Lake City)
06: SWE - FIN (Turin) - note that neighbouring Switzerland beat CAN
10: CAN - US (Vancouver)

The top 4 this year would probably be a mix of RUS, CZE, FIN and SWE, given this

 

[wgknestrick]

Time zones.

No North-American team has medaled at the Olympics since 1998, and you might as well pick the finalist teams based on who's playing closest to home.

98: CZE - RUS (Nagano)
02: CAN - US (Salt Lake City)
06: SWE - FIN (Turin) - note that neighbouring Switzerland beat CAN
10: CAN - US (Vancouver)

The top 4 this year would probably be a mix of RUS, CZE, FIN and SWE, given this

I think you are on to legitimate factor to break "talent ties".

 

[Czech Your Math]

Seems like the North American countries would generally be rated higher by this method than non-NA countries, since the former utilize solely NHL players, while the latter often use many players from their home leagues (which don't count toward GVT, correct?).

 

[hatterson]

Seems like the North American countries would generally be rated higher by this method than non-NA countries, since the former utilize solely NHL players, while the latter often use many players from their home leagues (which don't count toward GVT, correct?).

Players who didn't have NHL data were considered replacement level NHL players, so yes it would under-rate teams who have a large number of non-NHL players. Although there's really not too many that are much above replacement NHL level anyway so I can't imagine it making a huge difference.

 

[Feed Me A Stray Cat]

Time zones.

No North-American team has medaled at the Olympics since 1998, and you might as well pick the finalist teams based on who's playing closest to home.

98: CZE - RUS (Nagano)
02: CAN - US (Salt Lake City)
06: SWE - FIN (Turin) - note that neighbouring Switzerland beat CAN
10: CAN - US (Vancouver)

The top 4 this year would probably be a mix of RUS, CZE, FIN and SWE, given this

While it makes a degree of sense, that's a pretty limited sample.

 

[hatterson]

While it makes a degree of sense, that's a pretty limited sample.

Not to mention that CZE and RUS weren't exactly in home timezones in Nagano anyway. The US and Canada are just as close to Japan as Europe is.

 

[wgknestrick]

Players who didn't have NHL data were considered replacement level NHL players, so yes it would under-rate teams who have a large number of non-NHL players. Although there's really not too many that are much above replacement NHL level anyway so I can't imagine it making a huge difference.

If this is true, then the analysis doesn't hold much water. RUS has significant players that are anything but replacement level, GVT-wise. Stars in other pro leagues are certainly above replacement NHL level IMO.

GVT still has a hard time dealing with "opportunity" in these types of evaluations. Almost all of these guys are getting PRIME minutes to build their GVT on their current teams. IE they are the stars and "goto" guys to score. These PRIME minutes will only be passed onto a select few of the players on each team, so team depth isn't quite of a magnified advantage as it appears to be by looking at GVT. This tends to even the teams slightly as stacked teams (CAN) don't get as much of an advantage as GVT would leave you to believe.

Every forward isn't going to get 18mins 5v5 with 2min of PP time a game. You need to calculate GVT/min and then assume some sort of tiering for ice time to the players.

For example, when Toews is only getting 13min 5v5 and 2min of SH time, he isn't contributing a huge GVT (which is about 75% scoring based for forwards) to his team like he normally does.

 

[hatterson]

If this is true, then the analysis doesn't hold much water. RUS has significant players that are anything but replacement level, GVT-wise. Stars in other pro leagues are certainly above replacement NHL level IMO.

Kovalchuk has GVT data so he's not counted, and outside of Radulov there doesn't appear to be anyone who is significantly above replacement level. Granted I don't follow the KHL all that closely (read, at all) but it seemed that projections had the guys from the KHL slotted in as 4th liners and 3rd d pairing guys slotting behind guys like Anisimov, Kulemin, Tyutin, and Emelin who, although being solid players and valuable contributors aren't really huge GVT pushers.

GVT still has a hard time dealing with "opportunity" in these types of evaluations. Almost all of these guys are getting PRIME minutes to build their GVT on their current teams. IE they are the stars and "goto" guys to score. These PRIME minutes will only be passed onto a select few of the players on each team, so team depth isn't quite of a magnified advantage as it appears to be by looking at GVT. This tends to even the teams slightly as stacked teams (CAN) don't get as much of an advantage as GVT would leave you to believe.

Every forward isn't going to get 18mins 5v5 with 2min of PP time a game. You need to calculate GVT/min and then assume some sort of tiering for ice time to the players.

For example, when Toews is only getting 13min 5v5 and 2min of SH time, he isn't contributing a huge GVT (which is about 75% scoring based for forwards) to his team like he normally does.

They did do ice time adjustments for GVT for the major countries knowing that an example exactly like you gave with Toews will happen. The GVT contributions of guys who are expected to be 3rd and 4th liners on team Canada, US and Russia were reduced to reflect lower expected ice time and opportunity.

And yes, the usual caveats with GVT is clearly still in effect here, they mention as much in their article.

 

[soireeculturelle]

While it makes a degree of sense, that's a pretty limited sample.

very true. maybe a corsi-based analysis would shed more light.

started thinking about this when watching canada lose in the semis vs finland at the world juniors. it just looked like something was physically wrong with the players (matt dumba had the flu, which seemed to affect his play a lot, and other players seemed to have trouble making simple plays which could be a result of jet lag).

at the highest level, not sleeping or eating well can make a big difference in performance, especially with so many games in so few days, and the fact that most players on the team have never been in that part of the world.

 

[Bear of Bad News]

Edmonton Journal blog article looks at why Hall and Eberle didn't make Team Canada, using qualcomp and zone starts:

http://blogs.edmontonjournal.com/20...ylor-hall-jordan-eberle-didnt-make-the-grade/

 

[Bear of Bad News]

Similar article, but from Bleacher Report and focusing upon Team U.S.A.:

http://bleacherreport.com/articles/1921465-usa-olympic-hockeys-riskiest-players-in-sochi

 

[Bear of Bad News]

Ottawa Citizen article (Phil Birnbaum) on why the best Olympic hockey team might not win:

http://www.ottawacitizen.com/sports/best+Olympic+hockey+team+might/9405325/story.html

Focuses upon the impact of luck in a short tournament.

 

[hatterson]

Ottawa Citizen article (Phil Birnbaum) on why the best Olympic hockey team might not win:

http://www.ottawacitizen.com/sports/best+Olympic+hockey+team+might/9405325/story.html

Focuses upon the impact of luck in a short tournament.

Luck certainly does play a huge factor when every game beyond your 3rd is a win or go home game.

Even if we assume the "best" team has an 85% chance to win against any opponent (that seems like an absurdly high percentage) and earns a bye into the quarter finals via winning their bracket or being the best 2nd place team that still means their chance of winning 3 straight games is only .85^3 = 61.4125%, not even two out of three. Dropping that to "only" a 79% chance to win means you're worse than a coinflip to get through three single elimination games.

Even if you tier the rounds and say the "best" team has only a 5% chance to lose in the quarters, a 15% chance in the semi's and a 25% chance in the finals you still end up with roughly a 39.5% chance of them not winning gold.

 

[TheDevilMadeMe]

Similar article, but from Bleacher Report and focusing upon Team U.S.A.:

http://bleacherreport.com/articles/1921465-usa-olympic-hockeys-riskiest-players-in-sochi

Nice to see some statistical backing for what a lot of people have been saying:

One final point that jumps out is how the actual selection of shutdown defensemen might have been improved by taking Erik Johnson and/or Andy Greene instead of Brooks Orpik and Paul Martin.

I assume there is a typo here and "shot-based advantage" is supposed to be "shot-based disadvantage"?

They all bring similar skills and experience to the table, but Johnson and Greene don't put their teams at as much of a shot-based advantage as Pittsburgh's pair. But, again, you could argue that such an argument is also a cosmetic one that is safely left to the team's management. Let's move on to the forwards.