Playoffs and Fancy stats thus far

[toewsintangibles]

Unsustainable 2, Corsi 0 thus far

With both the Rangers and the Ducks beating their opposition in 5 games or under the trend of puck possession being the key to victory seems to be bucking..

Also other series are trending toward the worse team:

The Capitals (13th) are leading 3-2 against the best team in the league, the Islanders.
The Hawks (7th) are leading 3-2 against the fifth best team in the league, the Perds.
The Flames (26th) are leading 3-2 against the 18th team, the Nucks.
The Wild (11th) are leading 3-2 against the Blues (9th)

To be honest there are a few series still leading the charge for the cWo (corsi World order)

Montreal (17th) is leading the Sens (22nd) and Tampa (4th) is tied with Red Wings (10th)

Why do you think this is happening? Is this unsustainable?

There is a realistic chance to have the 6 best team, in terms of corsi, out of the playoffs in the 1st round!

My source for 5v5 fenwick for when score is close:
http://stats.hockeyanalysis.com/teamstats.php?disp=1&db=201415&sit=5v5close&sort=FFPCT&sortdir=DESC

Click to read more...

 

[Chris Hansen]

Why are you using the terms "Corsi" and "Fenwick" interchangeably?

 

[Quarter]

Two results against the popular train of thought is "bucking" the system?

Rather hasty, wouldn't you say?

 

[duckhead32686]

Ducks since trade deadline were tops or near the top In the league in both Corsi and Fenwick. They also carried possession in the series for the most part against Winnipeg.

Also, this "unsustainable" argument is really getting old. Even if the ducks weren't trending upwards on the possession stats they were still a #1 seed going against the "#8" seed.

 

[EssendonBombers]

Yes, with the inherent small sample size that goes along with playoff match-ups and the whole host of other variables that arise when tuning your game plan towards one specific team, cumulative stats like Corsi have less ability to accurately predict outcomes. Well done.

 

[eyetest is useless]

OP fail

First off, in a seven game series this year's sabres could beat the gretzky oilers. It wouldn't happen often, but it'd happen. Obviously in reality the skill edges are much closer, making the playoffs a crap shoot. That's hard to predict with any method.

Also, it's not like winnipeg and the rags were dogs from an analytics standpoint. People just said it's closer than most realise. In the pens case, their injured d-core was even used as a qualifier.

Take the ranger's top defense pairing away and see what happens.

 

[fedfed]

My source for 5v5 fenwick for when score is close:
http://stats.hockeyanalysis.com/teamstats.php?disp=1&db=201415&sit=5v5close&sort=FFPCT&sortdir=DESC

Haven't "close" numbers been debunked in a big way?

And yeah, I've heard an opinion that the possession stats as a predictive tool is best when a sample is about 25 last games.

 

[CorgisPer60]

Yeah, the probable allowance of corsi and fenwick has diminishing returns after 25 games, hence the more popular statisticians use a 25 game rolling sample size, as things such as personnel, injuries, and puck luck all factor into the equation at some point.

You know as well as anybody else that the Nashville Predators and the New York Islanders are relying on very early season success, and have been faltering quite significantly over the last stretch of games. Also, advanced stats deal in probability of success, not prediction of success.

 

[GKJ]

Yeah, the probable allowance of corsi and fenwick has diminishing returns after 25 games, hence the more popular statisticians use a 25 game rolling sample size, as things such as personnel, injuries, and puck luck all factor into the equation at some point.

You know as well as anybody else that the Nashville Predators and the New York Islanders are relying on very early season success, and have been faltering quite significantly over the last stretch of games. Also, advanced stats deal in probability of success, not prediction of success.

Nashville was losing games, but they weren't 'faltering' possession-wise. From the trade deadline to the end of the season they were still at 52.9% SAF which was 5th among teams who made the playoffs (tied with Winnipeg) and 9th in the league. Compare that with previous to the trade deadline, they were actually 4th with 'only' (in comparison to where it would be now) 53.4% SAF, not a big difference, but there were a few teams who missed the playoffs who passed them.

The bigger reason for Nashville falling off in wins and losses was the fact that Pekka Rinne taking sudden drop-offs. Compare this (this is a 10-game rolling 5v5 sv% graph) to Jonas Hiller who was something of a roller coaster, has been on the up for a couple months, and on a Burke-alytics team.

One reason Calgary was and is able to still sustain their success is the disparity in power plays to penalties. If you're getting that many more chances on power plays than penalty kills, it's more obviously chances to score, which means it's less chances the other team has to score. Also: high PDO when trailing.

 

[IdealisticSniper]

Yeah, the probable allowance of corsi and fenwick has diminishing returns after 25 games, hence the more popular statisticians use a 25 game rolling sample size, as things such as personnel, injuries, and puck luck all factor into the equation at some point.

You know as well as anybody else that the Nashville Predators and the New York Islanders are relying on very early season success, and have been faltering quite significantly over the last stretch of games. Also, advanced stats deal in probability of success, not prediction of success.

Spectacular post. Well said.

 

[Flamer12]

Maybe I am just too old for this s***, but who came up with this Fenwick and Corsi stuff. Seriously, who is the person that brought it to the forefront? Just curious if it is some GM or exective from a good team or a bad team.

 

[Levitate]

Also, advanced stats deal in probability of success, not prediction of success.

Yeah, big thing there. Sure you're statistically more likely to be successful over the long run with good possession numbers but that doesn't necessarily mean you'll beat X team in one game or even a series because your numbers are better

Maybe I am just too old for this s***, but who came up with this Fenwick and Corsi stuff. Seriously, who is the person that brought it to the forefront? Just curious if it is some GM or exective from a good team or a bad team.

Jim Corsi, goalie coach for the Sabres back in the day, came up with a version of Corsi as a way to track the workload on his goalies I think. Then some other people popularized it and it started taking off

 

[Devil Dancer]

For judging a team over the course of a season, I'm much more interested in goal differential.

 

[Jeti]

Corsi makes no attempt to capture goaltending. In each series you mention, the better goalie won.

Also, Corsi is just a simple stat. It's descriptive, not predictive (well... very weakly predictive). If people are naive enough to look at it as the sole predictor of a playoff series, they deserve to be disappointed. If you wanted to build a predictive model for playoff series, you'd need to consider many more features than just even-strength shot differential.

 

[groov2]

Because in hockey, he who scores more goals, wins; he who has the puck more, only puts himself in a better position to succeed. Is there an advanced stat on capitalizing on opportunities?

 

[Doctyl]

Because in hockey, he who scores more goals, wins; he who has the puck more, only puts himself in a better position to succeed. Is there an advanced stat on capitalizing on opportunities?

Shooting percentage?

 

[CorgisPer60]

PD0 is the closest one, as it factors in Fenwick, SH%, and a goalie's SV% when determining an effect.

 

[Vesa Awesaka]

worth noting that after firing Maclean the sens became a much better corsi team then montreal and much better advanced stat team overall. So you could argue that series is trending totwards the unsustainable

 

[groov2]

It was not until I joined HFBoards, that I started hearing so much about advanced stats. I have seen so many threads where people hang onto this stuff like religion.

On the Canucks board, people seem to absolutely hate Derek Dorsett because he has bad possession stats; any hockey guy I know thinks he is a very valuable piece to the organization. Sbisa, on the other hand, is an apple turnover machine and will never understand that signing :-P

We can write off my time playing Junior hockey because my last game was 10 years ago and we didn't use advanced stats at the time. We did use them; however, as recent as a few years ago when I was scouting. We used advanced stats as a means of gaining information for what our players and what other players were bringing to their clubs. It was simply a piece of the pie when scouting players and teams; it gives us added information on what they may be good at.

Advanced stats are very useful because information is useful, but there are so many more variables that on base percentage in hockey.

Sorry for the rant!

 

[WilderPegasus*]

Maybe I am just too old for this s***, but who came up with this Fenwick and Corsi stuff. Seriously, who is the person that brought it to the forefront? Just curious if it is some GM or exective from a good team or a bad team.

It took off because there are too many dumb keyboard warriors out there who watched Moneyball and thought it would make them look smart to spew that garbage.

 

[Doctyl]

The stats aren't even that advanced. It's literally the +/- of shot attempts. Its the best way we have to record possession. The hate it gets is ridiculous.

 

[Hammer Time]

Obviously injuries are a big factor, and so is goaltending.

In the NYR/Pittsburgh series you have 3 out of the top 4 Pens D-men out, and Lundqvist in net on the other end. That's going to make life difficult for your team even if you're a top Fenwick team.

Chicago was actually sitting tied for 2nd in Fenwick with Tampa when Patrick Kane got injured. They were 11th during the two months when Kane was out. Now that the Hawks are healthy again, they should be favoured over Nashville even from an analytics perspective (and with Weber out for the Preds, the edge got bigger).

And then of course you have Anaheim trending up in a big way (they were in the top-5 in Fenwick after the trade deadline), while the Islanders had a fairly average Fenwick in the second half of the season after being very good in the first half.

Even if you look at the full-season Fenwick standings, there's about a 2% difference from #3 to #14. That's great for parity, bad if you're trying to make predictions. When all the teams are that close, winners are really up in the air. And I'd say that's a good thing.

 

[Jaynki]

Corsi and all youre ****ing super advanced stats sucks and means nothing

 

[Doctyl]

Corsi and all youre ****ing super advanced stats sucks and means nothing

1) they aren't really advanced
2) yes they do mean something

 

[Sojourn]

Yes, with the inherent small sample size that goes along with playoff match-ups and the whole host of other variables that arise when tuning your game plan towards one specific team, cumulative stats like Corsi have less ability to accurately predict outcomes. Well done.

Interestingly enough, Anaheim was used as an example of a team regressing in the playoffs the last two seasons despite the small sample size. The 82 game season was deemed less "true", while the playoffs were seen as a more accurate representation. There seem to be a lot of contradicting arguments.

That being said, I don't think the OP is making the argument he thinks he is.