NHL Rankings by way of Puck Prospectus/VUKOTA

[zeke]

As for the projections themselves, VUKOTA produces conservative predictions. Everyone, especially goaltenders, are projected to regress towards the mean. That's why the projected standings are so tight. We don't really think the last-place team in the NHL will get 79 points. The spreads will be greater that projected. But this is the first go-around. Improvements will come.

Just wondering - what kind of weight do you put on:

a) Ice Time for defensemen (the comment that Beauch and Komi were similar in value struck me as off, for example - as was the downplaying of what Bo meant to FLA)

b) Save Percentage for goalies (as compared to other goalie stats)


and any chance you guys can redo the Toronto numbers taking the Kessel addition into account?


P.S. Great Work! this kind of analysis is badly, badly needed in this sport. hell, a good number of GMs would be wise to take a look.

 

[Jarick]

These rankings don't take into account change in coaching, systems, or linemates do they? I can't imagine Koivu, who spent most of last season with Brunette and Miettinen, would be worse this year having Havlat and Sykora on his wings. Or that Burns, who was out half the year with a concussion and played wing the other half, would improve. Or that Johnsson, under a new system that involves offense, would improve his meager point totals.

And somehow it's odd that all that adds up to the Wild shooting up to 12th in the league, when Backstrom will likely face a lot more shots and scoring chances.

Oh well. I guess we'll see how well it does in practice.

 

[HankTheTank]

huge lol @ those spewing vitriol that this system got "their team" all wrong and these guys don't know anything.

Criticize their correlations by all means (and to be sure they're still pretty crude imo) but the heart tugging and cliche bs is useless when you're arguing against a spreadsheet.

For all the doubters just remember than in the span of less than a decade these guys went from being total unknowns to the point that every sportsbook in the world waited for them to print Baseball Prospectus 2009 prior to them setting team win over/unders and then COPYING the projections for every team!

Obviously that's not the case with Puck Prospectus now as there is a huge variation btwn their projections and the lines out there, but I'm confident their models will improve over time.

And for all those that are so certain of outcomes, feel free to back up the truck on your favourite teams over/under points total.

 

[Moobles]

This is a lot better than TSN's intuitive assignment of teams or some random pundit-homer's interpretation of who deserves the Stanley Cup over who could possibly contend.

Like people have said, VUKOTA is relatively new and hockey is an extremely difficult sport to model (given it's so chaotic and there's so many confounding variables, etc.) However, an empirical approach to season rankings is really refreshing and it's great to see some method to prediction.

Thanks OP

 

[Viqsi]

These rankings don't take into account change in coaching, systems, or linemates do they? I can't imagine Koivu, who spent most of last season with Brunette and Miettinen, would be worse this year having Havlat and Sykora on his wings. Or that Burns, who was out half the year with a concussion and played wing the other half, would improve. Or that Johnsson, under a new system that involves offense, would improve his meager point totals.

And somehow it's odd that all that adds up to the Wild shooting up to 12th in the league, when Backstrom will likely face a lot more shots and scoring chances.

Oh well. I guess we'll see how well it does in practice.

It might have those, but with different weights than what you or I would give them. Not much way to know, really.

I'm sticking by my "public beta test" analogy in the meantime.

 

[Benttheknee]

I find their point totals to be quite conservative, as others have mentioned. Look for the top scorers, and they are almost always ranked too low. Crosby, Malkin, and Getzlaf being the exceptions.

Here are a few guys under 80 points:
Thornton
Heatley
Spezza
Lecavalier
St. Louis

Now it is true that a few had rough years last year, but those were more of the abberations than trends. As I recall Lecavalier had a bumb wrist, so he only got 67 or so points.

There is something a little off in their projections. Perhaps their alogrithm is too heavily weighted to last year.

 

[Hockey Outsider]

For now, take these predictions for what they're worth, which is hardly nothing, even if they're not accurate. If you asked a hundred equally intelligent hockey experts to make predictions for this season, and gave half of them these predictions, I guarantee the ones who got the predictions would do better than the ones who didn't. It's much easier to look at objective information like this and correct for "errors" where you can tell that the model can't handle an unusual player or situation properly than it is to just make stuff up based on your own knowledge.

At the end of the season, it would be interesting to look back on the VUKOTA projected standings, and projections from the mainstream media. It wouldn't be difficult to calculate which ranking system was more accurate.

 

[Neely2005]

They added Boston yesterday:

http://www.puckprospectus.com/article.php?articleid=288

3rd Overall.

Only 2 Teams left to go.

 

[dcsfanatic]

So Chicago and Washington are left?


Ohh....the anticipation

 

[Teemu]

So Chicago and Washington are left?


Ohh....the anticipation

Caps are #2, so Hawks are #1

 

[Steve Larmer]

Caps are #2, so Hawks are #1

http://puckprospectus.com/article.php?articleid=290

 

[LTIR Trickery]

All this does is prove VUKOTA is still a very raw tool. I love the idea, I love PECOTA/sabermetrics for baseball, but the tools for hockey are very much in infancy.

 

[MayDay]

It's fine to use pure stats to make a ranking. But what I take issue with is that the written analysis is just poor.

For example, for the Sabres, the analysis completely misses the signing of Mike Grier this summer (he's not even listed on the projected roster, nor mentioned once). The analysis seems to regard Gerbe as the most likely prospect to make the roster (when really, it's Myers).

The reasoning of the analysis seems to be that, well, the Sabres just missed the playoffs last year, they didn't make any major changes, so therefore they'll just miss the playoffs again. This reasoning ignores the fact that the two most important players on the team (goaltender Miller and leading goal-scorer Vanek) missed significant time to injury last spring, and that the Sabres were sitting in the 6th seed before those injuries occurred.

 

[MathMan]

For example, for the Sabres, the analysis completely misses the signing of Mike Grier this summer (he's not even listed on the projected roster, nor mentioned once).

Dunno why he was missed, but IIRC Mike Grier is more of an "intangibles" sort of guy anyway. Grinders like him don't weigh for much in statistical projections, typically.

This reasoning ignores the fact that the two most important players on the team (goaltender Miller and leading goal-scorer Vanek) missed significant time to injury last spring, and that the Sabres were sitting in the 6th seed before those injuries occurred.

Two points with that --

1-individual health is a skill, and a player that gets injured does tend to get injured again, so the projection system includes games missed in its projections. Whether that's fair or not for individual players is a different matter, but a statistical system can't differentiate between freak injuries that will never happen again and players who just tend to get hurt a lot -- especially since that's a subjective appreciation and a lot of what gets dismissed as the former turns out to be the latter.

2- Perhaps more importantly, the Sabres also lost their #1 D. That's a rather significant deletion from your roster, especially since "tough-minute puck-moving #1 defenseman" might just be the most impactful player on any team, up there with "starting goaltender" if not even ahead of him (depends on your backup, I guess).

 

[HankTheTank]

Pretty surprised they have the Caps rated so highly. I thought they "ran good" in terms of the number of points they got in regulation vs their goal differential based on doing a regression similar to the Pythagorean in other sports. They also ran way above .500 in the shootout.

Even if they improved over the summer (how do you quantify that it was a lot of player improvement?) .... I dunno.

 

[Dr Quincy]

It's fine to use pure stats to make a ranking. But what I take issue with is that the written analysis is just poor.

For example, for the Sabres, the analysis completely misses the signing of Mike Grier this summer (he's not even listed on the projected roster, nor mentioned once). The analysis seems to regard Gerbe as the most likely prospect to make the roster (when really, it's Myers).

The reasoning of the analysis seems to be that, well, the Sabres just missed the playoffs last year, they didn't make any major changes, so therefore they'll just miss the playoffs again. This reasoning ignores the fact that the two most important players on the team (goaltender Miller and leading goal-scorer Vanek) missed significant time to injury last spring, and that the Sabres were sitting in the 6th seed before those injuries occurred.

1) They didn't factor in Mike Grier? Do you really think that would change the projection very much?

2) The problem with your criticsm about Miller and Vanek is this: It's not enough for a team to improve. They have to improve RELATIVE TO THE LEAGUE. Every team had important players miss time. Every team (including the Sabres) will have important players miss time this year. Vanek played 73 games last year. Do you honestly think that was a hardship greater than what other teams faced? There's a good chance Miller plays more than 59 games this year, but every team has an injured played or 2 who will play more games this year than last.

They have the Sabres being in contention for a playoff spot, but just missing out. Given that:

Pitt
Was
Phi
Bos
Car
NJ

are all pretty much givens and that NYR and MTL are probably (right now) a slight cut about the rest, I don't see why it's outrageous to predict (if you have to make a prediction) that Buffalo just misses.

CAN they get in? Sure. Both NYR and MTL have flaws, but there are a few other teams right on that bubble with Buffalo.

 

[Ninja Hertl]

The Sharks could get better and not win the Obama trophy due to the possibility that other teams are also better.

 

[Alistar]

ok I have a couple points to make (I'm in a hockey pool with a few of the guys from PP and have been following their website closely from day 1).

#1: Those point projections are based on the player in question playing with league average linemates. I'm 99.9% sure of this. The whole point of this website is trying to isolate individual contributions to the hockey in a way that's better then just goals / assists / plus minus / whatever else, so it makes sense that they'd stabalize for linemates and competition. That's why Dany Heatley's point projections didn't change one bit going from Ottawa to San Jose. So yeah, if a player is going to be on a top line then they'll probably do as well or better then these predictions. These predictions are totally objective and having the author guess at line combinations would kind of ruin that.

#2: They never said Philly got worse on defense - they actually pointed out that they got much better (the Flyers were one of the bottom 5 defensive teams in the league last year going by GVT, their rating stat). The goaltending on the other hand.. Martin Biron was a borderline top 10 starter last season. Ray Emery played in Russia, so he has no useable stats, so their basing their prediction of the 2007-08 season in which he sucked ass. If he plays decent to above average to good the Flyers will probably beat their prediction, but I'm not exactly counting on it.

#3. On the topic of goaltenders in general, they noted in their Boston preview (and have mentioned a few times before in previous articles) that goaltenders are much more "volatile" then other players and tend to see their performance drop and rise at a much greater rate from season to season. Goaltending performance is (surprise) the biggest individual determinant of team success. So if your goaltender does surprisingly well, so will your team.

 

[Cashew]

Look, I don't think the Leafs will be a powerhouse. I hope they make the playoffs but it is FAR from a sure thing. But honestly, HONESTLY does anyone think they will finish as low as 29th in the league?!? They finished 7th worst last year, and they have improved on that team...

I think they'll make the play-offs.

 

[Iain Fyffe]

Sorry for not replying sooner. Hope no one minds me bringing back old threads.

Just wondering - what kind of weight do you put on:

a) Ice Time for defensemen (the comment that Beauch and Komi were similar in value struck me as off, for example - as was the downplaying of what Bo meant to FLA)

b) Save Percentage for goalies (as compared to other goalie stats)

and any chance you guys can redo the Toronto numbers taking the Kessel addition into account?

I would suggest contacting Tom Awad from PP for answers to these. He's the sytem's developer. The Kessel addition certainly makes Toronto a better team. Their 0-5 record so far (and league-worst -12 goal differential) shows just how badly they need him.

It's fine to use pure stats to make a ranking. But what I take issue with is that the written analysis is just poor.

For example, for the Sabres, the analysis completely misses the signing of Mike Grier this summer (he's not even listed on the projected roster, nor mentioned once). The analysis seems to regard Gerbe as the most likely prospect to make the roster (when really, it's Myers).

Grier was an oversight, but as someone says upthread, he wouldn't make any real difference to the projection. Read the article again; the most you could infer is that both Gerbe and Myers would be competing for roster spots. I implied that Myers would get a longer look because of his size, but on the other hand that the Sabres don't seem to be very biased against small players. So I'm not sure where you're getting that from.

The reasoning of the analysis seems to be that, well, the Sabres just missed the playoffs last year, they didn't make any major changes, so therefore they'll just miss the playoffs again. This reasoning ignores the fact that the two most important players on the team (goaltender Miller and leading goal-scorer Vanek) missed significant time to injury last spring, and that the Sabres were sitting in the 6th seed before those injuries occurred.

Really? Vanek missed only 9 games, and Miller played 59. So they didn't miss much timey. And your argument misses the fact that other teams also missed their star players for a few games (so you can't just apply this reasoning to the Sabres), and that the Sabres will likely have other injuries this year.

 

[MayDay]

Grier was an oversight, but as someone says upthread, he wouldn't make any real difference to the projection.

Really?

Grier has already helped the Sabres win two games that they likely would not have won last season.

He's a great lock-down defensive forward, and he was on the ice helping the Sabres hold their one-goal lead over Phoenix for the last two minutes of the game last week. The Sabres had major problems last season holding onto one goal leads late in games.

And of course he scored the GWG (and only goal of the game) late in the 3rd against Nashville Saturday night.

So I respectfully disagree with your assertion that the addition of Grier won't really have any impact on the team's W-L record. I think he will have an impact, and I think he's already had an impact.

 

[Teemu]

Iain, maybe you can help me with this. I can't help but notice when I look at the projected VUKOTA, that offensive defensive have absurdly increased DGVT. A prime example is with the Canucks, where you list 2008-09 and projected 2009-10 values. Christian Ehrhoff had a DGVT of 0.9 last year, but is projected to increase to a whopping 4.9 (same with Bieksa, 0.5 to 3.7). I'm sure the case is the same with the Hawks' Brian Campbell, who is projected to get a DGVT of 5.3, unrealistically on par with Brent Seabrook. What's causing those heavy biases to the defensive stats of offensive defensemen?

 

[Matt714*]

BTW, Emery's stats from the preseason. 0.943 SV% and 1.76 GAA. And he looks very very good.

Douglas Murray had a hat-trick during the preseason.

He has 1 goal in 216 NHL games.

 

[Iain Fyffe]

Grier has already helped the Sabres win two games that they likely would not have won last season.

...and those may be the only two all year. Do you really expect me to judge a player based on a sample of three games?

So I respectfully disagree with your assertion that the addition of Grier won't really have any impact on the team's W-L record. I think he will have an impact, and I think he's already had an impact.

No significant impact, over the course of an entire season. He may well do something to lose the Sabres the next game, thus negating some of the good work so far. Three games is not enough to judge.

What's causing those heavy biases to the defensive stats of offensive defensemen?

That's one of the flaws of the GVT system. It incorporates plus-minus into the defensive values. Now plus-minus is often called a "defensive" statistic, but it isn't because it also includes offence. So when a defenceman prevents a goal from being scored, it helps his defensive GVT (through his plus-minus). But when a defenceman creates a goal on offence, it helps both his offensive GVT and his defensive GVT, since the defensive GVT includes plus-minus.

So given the same defensive ability, a more offensively-talented defenceman (or forward for that matter) will have a greater defensive GVT.

It's the biggest flaw in GVT as I see it.

 

[MayDay]

...and those may be the only two all year. Do you really expect me to judge a player based on a sample of three games?

No, I'm saying that he does things to help teams win games that don't show up in stat sheets or box scores. I was using the games last week as an example of the kind of things Grier does (and I'm well-familiar with what Mike Grier brings from his previous stint with the Sabres).

One of the things that he does is help lock-down one-goal leads late in games. He's a good defensive forward. The Sabres had major problems holding onto one-goal leads late in games last season. Therefore, the addition of Grier's skills in this area should make a direct contribution to win totals, even though most of the time his contribution won't show up on the boxscore.

People talk about "intangibles" in hockey because they're actually important and they matter. So perhaps a purely statisics-based projection is not the best way to make predictions in hockey. I just don't think such a method works as well in hockey as it does in baseball. I like baseball too, but hockey is a completely different sport, a less stat-based sport. There are too many other factors that go into Ws and Ls that don't show up in stats.