Fenwick Tied Rankings
- Thread starter Jason MacIsaac
- Start date
Jason MacIsaac
Registered User
You're preaching to the wrong crew here. These stats have been proven to positively correlate with winning.
Last edited:
This guys right. A stat is useless if it doesn't correlate positively winning % at a consistent level. All it does is tell us how well the team is producing at whatever that stat is measuring.
Just saw this, mind linking me to where it's been tested/reviewed?
Well, for one, "chances" are so incredibly biased, I won't go into what constitutes one or not. I never made a claim we outchance the opposition. I say the style we employ makes the chances we do get "high quality". It's why we play well against teams that employ the cycle, like Vancouver. They can throw all the point shots they like, but when their point man fumbles the puck, we get a 2 on 1. "Outchanced"? Sure, according to the stats. We'll still win the game.
The difference for the Rangers is that when the effort isn't there, we don't force those turnovers and don't create those scoring chances. And we, obviously, lose. However, unlike other teams, when we have a stinker we register sub-20 shots. Let the other team take shots for a minute, go dump the puck and go for a change. When the effort isn't there, that's what a crappy Rangers game looks like. When we aggressively forecheck and pressure the shooters, they cough up the puck and we get an odd man rush every minute and we put on our "normal" amount of shots.
This you're just pulling out of your ass
The truth is neither you nor I will know where they will end up in the end. These stats are interconnected. Letting a few more shots get by boosts Lundqvist's save % stat, sure. We will likely let a few less get by later in the year and it'll decline, but his GAA is right where it was last year.
Shot percentage and how many shots you take is also connected. Gaborik this year has an atrocious shooting percentage - however he's also throwing more shots on net than he has ever done in his career. Even the ones that aren't high quality, he just puts it on net. This boosts his shot/game stat, while lowering his shot %.
The last two years we've about the same amount of shots on net than we allowed to get to us. Last year was +.5. The year before that was -.3 (the year we didn't make the playoffs and our team was all Gaborik).
This year we're at -6. If you think we'll stay there, that's on you. We're icing essentially the same team, just with the emergence of two defensemen who can carry the puck, the blossoming of Stepan (who has looked nothing short of incredible) and the adding of Richards, who will not likely change much even strength, but has already added so much to our PP. Despite it being worse statistically than last year, the change is night and day. Just a lot of unlucky posts in a small sample size can do that.
Its because he will never produce "points/big numbers". Happens all the time, and most people dont have the time/dont make an effort to watch games from all teams consistently. And when they watch the game, its not to focus on a specific number of individual players, and when it is, its always going to be on the "stars" (offense stars...)
For example, you being a St. Louis fan, would likely agree that Pietrangelo has been phenomenal this year and no doubt their best defender. Shattenkirk is the one who is getting praise from leaguewide fans though. Shatty has been good too but he is much more sporadic in his own zone and overall less consistent and solid. The impact of both, net positive and negative, is in favor of Pietrangelo.
Oh no, the fact that I forgot the Blues played last night invalidates everything I've ever said.
An 82-game sample size isn't enough to guarantee that random variance won't propel undeserving teams into the playoffs while deserving teams are denied qualification. The bounces just don't even out over that small a period. Fans and some GMs need to be looking longer term than that. The facts are that the Blues were a dominant possession team last year but received extremely poor goaltending thanks to Halak having a below-average season and being injured for an extended period without a competent backup. I fail to see how that was Payne's fault.
Jason MacIsaac
Registered User
I'm not searching that up again, I looked probably an hour ago and the corrlation is .61. On the Cohen scale that is considered strong.
bluesfan94
Registered User
A .602 correlation isn't exactly the strongest correlation, nor does it prove causation.
So I present you with a numerical, reality-based look at the fact that the Rangers do not get more "high quality" chances than the other team and you refute it...based on zero evidence. Yes, they're winning the games. Because their shooting percentage is currently at a ridiculous rate that no team has sustained over the course of a season the last four years and Lundqvist is stopping pucks at a rate that no non-Thomas goalie has come close to in that span.
This you're just pulling out of your ass
The truth is neither you nor I will know where they will end up in the end. These stats are interconnected. Letting a few more shots get by boosts Lundqvist's save % stat, sure. We will likely let a few less get by later in the year and it'll decline, but his GAA is right where it was last year.
Shot percentage and how many shots you take is also connected. Gaborik this year has an atrocious shooting percentage - however he's also throwing more shots on net than he has ever done in his career. Even the ones that aren't high quality, he just puts it on net. This boosts his shot/game stat, while lowering his shot %.
I don't know exactly where they'll end up but I sure as hell know they won't be able to sustain their 1039 PDO given that the highest PDO team in the NHL at the end of last season was Boston at 1022, buoyed by Thomas' incredible even-strength SV% (Nashville was 2nd with a 1015 PDO).
Like I said, we can expect the Rangers to somewhat improve territorially based on past seasons' performance but whatever gains they make in that area are going to be dwarfed by the negative regression that occurs in their even-strength SV% and SH% which are literally the only reasons the Rangers are where they are in the standings right now.
With the 0.928 even-strength SV% and 8% even-strength SH% the Rangers managed last season, they would currently be being outscored 2.24-1.90 per sixty minutes of even-strength hockey instead of outscoring teams 2.5-2.0 as they are right now.
Last edited by a moderator:
Most people who dont come from a strong gambling or statistic background are not going to appreciate just how dominant variance is and how/when to account for it. In reality, everything is governed by variance. Nothing plays to the average, yet everything converges to it. But it only does so over the long run, and people are not going to believe how long the long run can take.
For example in hockey, outshooting your opponent by 3 on average for 82 games (still relatively short term, but getting there, I consider 200 games to be a meaningful sample size) can be the difference between a 5th seed and a 12th seed, while at the same time you can win any single game while getting reamed in possession, lets say getting outshot 17-41.
If you take it to the extreme, getting outshot 17-41 every game is not going to translate into success, yet if you win 5 of those in a row (statistically unlikely, but not even remotely impossible) then there will be, for sure, people (in the media or just casual fans) who spin it somehow into "X goalie does better when facing more shots" or "they bend but dont break" or something like that. It's very easy to see the impact when you take things to the extreme -> when its brought back to reality, the same impact is there just magnitudes smaller, which, for any with a gambling/statistic background can tell you, increases variance, which makes short term results even more unpredictable and even less meaningful.
For example in hockey, outshooting your opponent by 3 on average for 82 games (still relatively short term, but getting there, I consider 200 games to be a meaningful sample size) can be the difference between a 5th seed and a 12th seed, while at the same time you can win any single game while getting reamed in possession, lets say getting outshot 17-41.
If you take it to the extreme, getting outshot 17-41 every game is not going to translate into success, yet if you win 5 of those in a row (statistically unlikely, but not even remotely impossible) then there will be, for sure, people (in the media or just casual fans) who spin it somehow into "X goalie does better when facing more shots" or "they bend but dont break" or something like that. It's very easy to see the impact when you take things to the extreme -> when its brought back to reality, the same impact is there just magnitudes smaller, which, for any with a gambling/statistic background can tell you, increases variance, which makes short term results even more unpredictable and even less meaningful.
Jason MacIsaac
Registered User
bluesfan94
Registered User
Yeah, but variance can't be tapered out over a long sample size for each hockey team, as teams shift year to year as players develop, decline, or switch teams.
TOML
Registered User
EAST
Pittsburgh - 0.538 - good system and sturdy players - but injury-prone stars
Washington - 0.534 - system in place, but shaky player foundation
Boston - 0.525 - solid lineup throughout
Montreal - 0.525 - system decent, players could be better
Florida - 0.523 - solid rebirth here - surprised
New Jersey - 0.520 - system making team appear to be good
Ottawa - 0.504 - new coach, new energy
Philadelphia - 0.503 - should be better. Coaching is done by Pronger
Toronto - 0.501 - good efforts by players who are dropping like flies
Winnipeg - 0.497 - better than expected after meh start
Buffalo - 0.486 - should be a lot better under Ruff. Forwards beyond top line not doing as well as hoped
New York I - 0.478 - lack of defensive depth is taxing them
Carolina - 0.474 - new signings not that good. Not much going right lately
New York R - 0.453 - collapsing effectively around Lundy
Tampa Bay - 0.448 - re-tooling needed after Stamkos' big signing
WEST
Detroit - 0.571 - same ol' Detroit. This will plummet with no Lidstrom. Or will it?
St Louis - 0.567 - good depth all around. Kinda injury-prone though
Chicago - 0.551 - decent rebound season. Defense needs work. Crawford might not handle too large a workload
Vancouver - 0.537 - expected dropoff. Need players to be healthier in order to go anywhere
Colorado - 0.521 - and i thought Sacco was the problem... They need a few more vets to help the kids. Esp. on D
Phoenix - 0.509 - Same ol' Tippett
San Jose - 0.508 - pick control of top line continuing it's steady decline. But depth is elite.
Calgary - 0.505 - good veteran depth. But hardly elite.
Columbus - 0.501 - Eh? The record says goaltending. Brutal goaltending.
Edmonton - 0.499 - Young. Still learning. Excellent at home. Not so much on road.
Los Angeles - 0.491 - Should be a lot better, but there are some missing pieces
Dallas - 0.454 - very hot goalie starting to cool off
Nashville - 0.433 - no surprise. The trifecta needs help up front
Anaheim - 0.417 - also no surprise. No depth up front. Imagine if Teemu retired...
Minnesota - 0.413 - and yet they lead the west? This is not good for the league. But when their goaltending cools off, meh
Pittsburgh - 0.538 - good system and sturdy players - but injury-prone stars
Washington - 0.534 - system in place, but shaky player foundation
Boston - 0.525 - solid lineup throughout
Montreal - 0.525 - system decent, players could be better
Florida - 0.523 - solid rebirth here - surprised
New Jersey - 0.520 - system making team appear to be good
Ottawa - 0.504 - new coach, new energy
Philadelphia - 0.503 - should be better. Coaching is done by Pronger
Toronto - 0.501 - good efforts by players who are dropping like flies
Winnipeg - 0.497 - better than expected after meh start
Buffalo - 0.486 - should be a lot better under Ruff. Forwards beyond top line not doing as well as hoped
New York I - 0.478 - lack of defensive depth is taxing them
Carolina - 0.474 - new signings not that good. Not much going right lately
New York R - 0.453 - collapsing effectively around Lundy
Tampa Bay - 0.448 - re-tooling needed after Stamkos' big signing
WEST
Detroit - 0.571 - same ol' Detroit. This will plummet with no Lidstrom. Or will it?
St Louis - 0.567 - good depth all around. Kinda injury-prone though
Chicago - 0.551 - decent rebound season. Defense needs work. Crawford might not handle too large a workload
Vancouver - 0.537 - expected dropoff. Need players to be healthier in order to go anywhere
Colorado - 0.521 - and i thought Sacco was the problem... They need a few more vets to help the kids. Esp. on D
Phoenix - 0.509 - Same ol' Tippett
San Jose - 0.508 - pick control of top line continuing it's steady decline. But depth is elite.
Calgary - 0.505 - good veteran depth. But hardly elite.
Columbus - 0.501 - Eh? The record says goaltending. Brutal goaltending.
Edmonton - 0.499 - Young. Still learning. Excellent at home. Not so much on road.
Los Angeles - 0.491 - Should be a lot better, but there are some missing pieces
Dallas - 0.454 - very hot goalie starting to cool off
Nashville - 0.433 - no surprise. The trifecta needs help up front
Anaheim - 0.417 - also no surprise. No depth up front. Imagine if Teemu retired...
Minnesota - 0.413 - and yet they lead the west? This is not good for the league. But when their goaltending cools off, meh
Variance is never going to "even out" per se, but it will converge to the "true average". No one knows the "true average" for anything in the real world, but we generally have an idea of the ranges for any particular event. As players develop/decline and teams change makeup, the "true average" obviously has to change as well, but every game you play converges you closer to the true average (not the old one, but the new one, which in fact may move you further from the old one, but that's irrelevant now). So you can run "below" your old average for example and now "above" your new average but see similar results, which makes assessing talent/results/anything very difficult.
So yes, you cant wait for things to average out given the current set of variables, since they are always changing, but results are always still converging to the true average, we just have to evaluate and assign a new range for that.
So 82 games isn't a sufficient sample size...when the outcome runs contrary to your opinion. However, it's apparently a large enough sample size for your precious little stat that you base everything upon.
And that doesn't address the crux of what I said anyway.
Jason MacIsaac
Registered User
System making team appear to be good? How does achieving a 52% Fenwick only appear to be good. It is good.
This is actually backwards. Marleau, Thornton, Pavelski and Couture, who have all spent significant time on the top line, have individual on-ice score-tied Fenwick ratios of 0.496, 0.496, 0.514 and 0.589 respectively -- particularly impressive considering Marleau, Thornton and Pavelski have carried the mail this season (as Patty and JT did last year) in terms of going head-to-head against opposing teams' top lines. The depth is the issue. The third line is crap with Mitchell, Handzus and McGinn registering score-tied ratios of 0.368, 0.410 and 0.468 despite playing easy minutes against nobodies. Demers and Braun have struggled on the back end as well.
Jason MacIsaac
Registered User
What he is saying is that it isn't a large enough sample size to completely regress to the mean but it still is positively correlated with success. That means ****** teams like Anaheim can make the playoffs last year.
Compare the volume of shots and missed shots taken in a given season to the volume of goals scored and get back to me about sample sizes.
It's been definitively proven that, largely due to the parity in the league, shooting percentage differences among teams at the NHL level are the result of luck, not skill. Over the last four seasons, teams have had zero year-to-year correlation in shooting percentage. While goaltender talent does exist, single-season SV% is just as much the product of random variance as single-season SH% and that can be shown by the fact that teams strongly converge to a PDO (even-strength SH% + even-strength SV%) of 1000 by the end of the season. However, it isn't a perfect regression and there are inevitably going to be teams that are affected by the percentages going in either direction. There's more than enough evidence to conclude that evaluating teams by goals scored and allowed in a single season is a flawed metric.
TOML
Registered User
Compare the standard deviation of shots to the standard deviation of goals scored and get back to me about confidence intervals.
bluesfan94
Registered User
Yeah, I understand variance will never be gone, but if an 82 game sample size isn't enough for the convergence to the average, then the whole idea of correlation is a nonfactor in hockey, with teams changing on a yearly basis. True, variance will likely have less of an e effect over 82 games, but if teams like STL and NJD last year were out of the playoffs despite statistics suggesting they should be in, then correlation shouldn't be regarded as a predictor.
I haven't looked at the stats, but I would imagine that over a meaningful sample size, the two would both get exponentially lower and have a decent correlation, which is his point. Obviously, over an infinite sample size, I'm nearly positive you would find a strong correlation between the two within the 95% confidence margin. Honestly, over even 82 games, I would be fairly confident that it would fall within 2 z-scores.
If that's the case, he can feel free to show the stats on it. It's not like it's a new concern, but none of the random blogs that the "hockey sabermetrics" types like to use as "proof" ever seem to post any of the important parts of the statistics.
The larger point that's being missed is that he's gone off the deep end in regards to conclusions based upon a single, low correlation stat which hasn't even been established to fall within an extremely generous α value.
