I think Copp-Lowry-Tanev should be reunited now.bUt LaInE wOuLd ImPrOvE wItH sChEiFeLe.
I'm genuinely surprised CLT hasn't been utilised regularly this year. I think they have proven themselves as a top tier shutdown unit already. Ehlers is a boon, whereas Laine's season is just astonishingly FUBAR. That has to change.
It's not a good sign for the Jets but there is a ton of noise / statistical error in those models. I won't be that surprised if the Jets or any team other than the Bolts get knocked out first round, but I also won't be that surprised if the Jets make it through at least a couple of rounds. Goaltending is really going to be a key along with the big questions around whether Scheifele returns to his form from last year's playoffs and whether Morrissey comes back healthy and sharp enough to transform how the Jets' top D pair plays.Disconcerting Travis Yost interview yesterday on 1290. He says a team's underlying stats for the last 25 games of the season are more predictive of playoff performance than the season as a whole. If that pans out for the Jets, people will be wondering who let the Senators into the playoffs.
GP | xSV% | dSV% | GSAA/60 | |
CHICAGO | ||||
Crawford | 37 | 91.39 | 0.07 | 0.02 |
Ward | 31 | 90.92 | 0.68 | 0.23 |
Delia | 15 | 91.41 | 1.03 | 0.41 |
COLORADO | ||||
Varlamov | 47 | 92.04 | 0.17 | 0.05 |
Grubauer | 35 | 92.44 | -0.02 | -0.01 |
DALLAS | ||||
Bishop | 45 | 92.10 | 1.54 | 0.46 |
Khudobin | 40 | 92.59 | 1.08 | 0.34 |
MINNESOTA | ||||
Dubnyk | 67 | 93.14 | -1.31 | -0.37 |
Stalock | 19 | 92.52 | -1.86 | -0.50 |
NASHVILLE | ||||
Rinne | 53 | 92.00 | 0.75 | 0.22 |
Saros | 31 | 92.11 | 0.23 | 0.07 |
ST. LOUIS | ||||
Allen | 46 | 92.48 | -0.88 | -0.26 |
Binnington | 30 | 92.26 | 2.03 | 0.51 |
Johnson | 19 | 91.55 | -3.42 | -1.00 |
WINNIPEG | ||||
Hellebuyck | 61 | 92.34 | -0.42 | -0.14 |
Brossoit | 21 | 92.13 | 1.87 | 0.60 |
bUt LaInE wOuLd ImPrOvE wItH sChEiFeLe.
I'm genuinely surprised CLT hasn't been utilised regularly this year. I think they have proven themselves as a top tier shutdown unit already. Ehlers is a boon, whereas Laine's season is just astonishingly FUBAR. That has to change.
I have blocked what seems like the last 30 or 40 Laine threads but I can’t imagine who’s fault it is now because we know it’s never Pate’s fault.
It's only been 4 games so small sample size caveats apply- in the month of April, the Jets have been #1 in the league in Corsi and #5 in Expected Goals. Wheeler and Scheifele's uptick and Buff has been a big part of it. Hopefully this carries on into the playoffs.
There is a definite large coaching contribution to Laine's play. Watch where he goes on the ice now. He backs his ass into the opposing goalie like he's Dave Andreychuk and hangs out on the half boards waiting to lose a inconsequential puck battle. He used to skate into the zone with the puck a lot. Now, that never happens. Laine didn't forget how to play hockey. I'm afraid our coaching staff turned a flawed superstar into a ****ty 4th liner. I ****ing hate Paul Maurice.
The latest narrative among some is that it's the coach's fault and that the coach has held him back.
Whatever is going on with Scheifele and Wheeler at 5 on 5 should be like a horror movie for the fans. This season is what it is but Wheeler/Scheifele stuff has long term implications for the future. I see two players who are having one hell of a time breaking 50% in their underlying metrics game after game and it has been going on for a long time. If this is the new normal then we are truly ****ed. If you go on Corsica and look at CSW line's stats from last season and compare them to this season, it is truly depressing how far they have fallen, I just hope that it's not one of these two falling off a cliff.
I'll try to help:
Shot Metric Types
Corsi - shot attempts (goals, saves, misses, blocks)
Fenwick - unblocked shot attempts (goals, saves, misses)
Shots (on net) - as implied (goals, saves)
Goals - as implied (goals)
xGoals - expected goals (There are many different formulas but most are Fenwick adjusted for shot location and other factors. Note: reasoning why it's Fenwick adjusted and not Corsi adjusted is because the locational data in NHL pxp uses where the shot was blocked and not where the shot was taken)
Metric Forms
_F - events for team or for team with player on ice, ex: CF, xGF
_A - events against team or against team with player on ice, ex: CA, xGA
_D or _PM - differential or plus minus, ex: CD = CF - CA
_F% - events for team relative to all events, ex: CF% = CF / (CF + CA)
_/60 - events relative to an hour of ice time, ex: CF/60
Type of Adjustments
Score Adjustments - adjusts the volume of shots based on the score of the game and period
Venue Adjustments - adjusts for home team advantage
Arena Adjustments - adjusts for arena biases in how some arenas tend to count more conservatively/liberally or tend to push shots one way or another
RAPM - adjusts using ridge regression for players on the ice (QoC and QoT), zone starts, schedule (B2B or not), and the previously mentioned adjustments
Now to try and answer your questions directly...
1) Generally not. xGF is to account for shot quality not finishing talent. These are considered two separate skillsets and so they are typically separated. There was on xGF model that accounted for shooter talent but that model no longer exists.
2) xGoals use things like rush shots vs sustained pressure shots to adjust for shot quality. Again there is no adjustment for player shooting, but that is intentional. You can think of things like:
GF => xGF x luck x finishing talent
GA => xGF x luck x goaltending talent
3) xGF still matters for a player like Laine. Yes the average shot for Laine at the same location is worth more than the average player, but that doesn't mean that more shots or closer shots are not better for Laine to have. It's merely the shot finishing translator would be different. Ex: Laine has his worst career xGD/60 this season and his worst GD/60 this season (both when RAPM adjusted). That's not coincidental. That said, with Laine and Lemieux both having the worst xGD/60 (RAPM adjusted) on the Jets, I give Laine more leeway than a player like Lemieux with less finishing talent.
They both had good metrics in 2017/18,and sparkling metrics in the playoffs last season. I don't think it's necessarily a long-term issue. They just got into a lull where they weren't executing well defensively. Their Corsi numbers are pretty good in the playoffs so far.Whatever is going on with Scheifele and Wheeler at 5 on 5 should be like a horror movie for the fans. This season is what it is but Wheeler/Scheifele stuff has long term implications for the future. I see two players who are having one hell of a time breaking 50% in their underlying metrics game after game and it has been going on for a long time. If this is the new normal then we are truly ****ed. If you go on Corsica and look at CSW line's stats from last season and compare them to this season, it is truly depressing how far they have fallen, I just hope that it's not one of these two falling off a cliff.
Garret, thanks for the above. One other question for you.
I see some people use xGF to assess how well goaltenders and individual players did in a particular game. That just seems completely inappropriate.
I thought this stat and other shot metric stats gain more meaning over a large block of time instead of using this on a game level. To use it at a game level would demand quite a bit more input data to assess the play than what is currently used to compile this data.
Can you please clarify?
Agreed, and well-stated. I would add that very large sample sizes can make even the most minute difference statistically significant, even if the effect size is unimportant. In most applied statistics, it is customary to share / display both the effect sizes and the statistical error (e.g. confidence intervals). We seldom see that in hockey analytics... it looks less elegant, I suppose.The more data you get, the more confidence you gain. There is no black or white, just change in confidence intervals.
That's all data, including that from eyetest.
So, yes, statistics like xG and Corsi become more meaningful as "good performance" and "good player" align together as sample increases. How confident you are depends on the sample size, but also what you are using to measure as well as what question you are trying to answer.
I'd argue that using those statistics are good tools to measure how well individuals perform in a particular game, as they will paint part of the picture... but not the whole picture... but enough that it cannot and should not be ignored... but I'm a lot more confident in those measures at 15 games than 1.
Interesting to see that so far this post season there has been zero correlation with regular season CF% and playoff success with the teams having the better regular season CF% getting eliminated by the poorer performing CF% team.