HFBoards

HFBoards (http://hfboards.hockeysfuture.com/index.php)
-   By The Numbers (http://hfboards.hockeysfuture.com/forumdisplay.php?f=241)
-   -   Help explaining adv stats (http://hfboards.hockeysfuture.com/showthread.php?t=1362139)

Colton2233 02-27-2013 08:47 PM

Help explaining adv stats
 
Can someone explain Corsi and or Fenwick close. Looking for links to help but having trouble finding them. I read today Anaheim has a Fenwick close of (46.64). What does that mean?

SmellOfVictory 02-27-2013 11:44 PM

Corsi/Fenwick close are (someone correct me if I'm wrong) a team's shot differential per 60 min when the score in a game is either tied or off by a single goal. This is important as it can limit the score effects (when a team is up by a substantial number of goals, it tends to go on the defensive, and if down by a number of goals it will become more offensive, which skews the "real" outshooting ability of the teams involved).

A Fenwick close of 46.64 would mean that Anaheim is outshooting its competition by 46.64 shots/60 min when the score is close. That seems absurdly high to me, though. Like, it should be in the single digits.

edit: Just to clarify on Corsi/Fenwick, both include shots and misses in their differential, while Corsi also includes blocked shots.

LiveeviL 02-28-2013 12:56 AM

Perhaps the figure in the OP was a percentage vs opposition? It would at least be within reason.

GuineaPig 02-28-2013 03:30 AM

Quote:

Originally Posted by SmellOfVictory (Post 60633085)
Corsi/Fenwick close are (someone correct me if I'm wrong) a team's shot differential per 60 min when the score in a game is either tied or off by a single goal. This is important as it can limit the score effects (when a team is up by a substantial number of goals, it tends to go on the defensive, and if down by a number of goals it will become more offensive, which skews the "real" outshooting ability of the teams involved).

A Fenwick close of 46.64 would mean that Anaheim is outshooting its competition by 46.64 shots/60 min when the score is close. That seems absurdly high to me, though. Like, it should be in the single digits.

edit: Just to clarify on Corsi/Fenwick, both include shots and misses in their differential, while Corsi also includes blocked shots.

Uh, no. It means that when the score is "close" (within 2 goals), Anaheim takes 46.64% of all shot attempts 5v5.

It's not a good Fenwick score. It's a bad one. And it suggests that Anaheim is due for some major regression.

Anaheim is currently 24th in the league in Fenwick close.

Rutabaga 03-01-2013 03:21 AM

For the record, it is obviously possible to be playing in the post-season despite having a Fenwick ratio under 50%.
But they're not that many. (4 last year, 2 in 2011...)

Its stating the obvious, but you cant expect good and consistent results, if you're outshot that often at ES without some magic recipe. The shortened season is fooling some people, but right now, teams like Anaheim, Toronto and Nashville are good ideas if you want to bet about a "suprising" fall in the standings or a first-round exit if the regression to the mean doesnt happen early enough. Which is totally possible in a 48-game schedule. But its very likely that their seasons are just a fluke.

These teams (the ones under 50%, not those 3 in particular) clearly overachieved during the regular season, and they basically never repeat their success during the playoffs.

The ones that managed to win several (well, thats never more than 2) rounds always did so because they ended up playing against other team(s) under 50% (Except Pittsburgh under Therrien and Montréal with Halak in the nets. Not sure that models you can/want to replicate.)

Over 70% of the teams playing in the PO have a Fenwick score over 50% and thats probably the safest tool to see which teams are actually controlling the play, and which ones are not.

DatsyukToZetterberg 03-05-2013 11:22 AM

I tried looking at some articles explaining what Corsi, Corsi Rel & Qual comp but I'm still a bit confused as to what they actually represent and how I should interpret them.

Also what's the difference between Corsi Rel QoC & Corsi QoC? For example right now Chara has a Corsi Rel QoC of 1.761 while his Corsi QoC is -.0832?

SmellOfVictory 03-05-2013 11:41 AM

Quote:

Originally Posted by GuineaPig (Post 60638161)
Uh, no. It means that when the score is "close" (within 2 goals), Anaheim takes 46.64% of all shot attempts 5v5.

It's not a good Fenwick score. It's a bad one. And it suggests that Anaheim is due for some major regression.

Anaheim is currently 24th in the league in Fenwick close.

Ah my bad; I got it confused with raw scores. Makes a lot more sense as a proportion.

SmellOfVictory 03-05-2013 11:51 AM

Quote:

Originally Posted by DatsyukToZetterberg (Post 61006689)
I tried looking at some articles explaining what Corsi, Corsi Rel & Qual comp but I'm still a bit confused as to what they actually represent and how I should interpret them.

Also what's the difference between Corsi Rel QoC & Corsi QoC? For example right now Chara has a Corsi Rel QoC of 1.761 while his Corsi QoC is -.0832?

Corsi is generally considered an analogue for puck possession. The better it is for a player or team, the more often that team has the puck (particularly in the offensive zone) when the player is on the ice (or simply the more often that team has the puck, period); in other words, the better at hockey the player/team is. Corsi Rel is the same deal except it's a differential between the player and the rest of his team (Corsi while player is on the ice minus Corsi while player is off the ice); it helps illustrate where a player ranks amongst his teammates in terms of puck possession.

Important things to note with Corsi & Corsi Rel are zonestarts and qualcomp/qualteam. They provide a lot of needed context, as a raw Corsi rating could easily make a 4th liner look better than a 1st liner in a vacuum.

QualComp confuses me a little, but it's some esoteric thing based on the ratings of a player's average opponents. Higher is tougher, although it is not the most reliable measure (better than nothing, in most cases).

DatsyukToZetterberg 03-05-2013 01:56 PM

Quote:

Originally Posted by SmellOfVictory (Post 61008347)
Corsi is generally considered an analogue for puck possession. The better it is for a player or team, the more often that team has the puck (particularly in the offensive zone) when the player is on the ice (or simply the more often that team has the puck, period); in other words, the better at hockey the player/team is. Corsi Rel is the same deal except it's a differential between the player and the rest of his team (Corsi while player is on the ice minus Corsi while player is off the ice); it helps illustrate where a player ranks amongst his teammates in terms of puck possession.

Important things to note with Corsi & Corsi Rel are zonestarts and qualcomp/qualteam. They provide a lot of needed context, as a raw Corsi rating could easily make a 4th liner look better than a 1st liner in a vacuum.

QualComp confuses me a little, but it's some esoteric thing based on the ratings of a player's average opponents. Higher is tougher, although it is not the most reliable measure (better than nothing, in most cases).

Ah makes more sense now, so if has a - corsi he is not as good at controlling the puck as someone with a + corsi? And what's the main difference between any of the stats with Relative in it? Ie: Corsi Rel QoC - Corsi QoC

Blue Blooded 03-05-2013 04:57 PM

Quote:

Originally Posted by DatsyukToZetterberg (Post 61016043)
Ah makes more sense now, so if has a - corsi he is not as good at controlling the puck as someone with a + corsi? And what's the main difference between any of the stats with Relative in it? Ie: Corsi Rel QoC - Corsi QoC

Corsi QoC only takes in the raw Corsi rating of the opponents, so if you match up against a unit that has averaged a Corsi of 5 during the season, you will have a Corsi QoC of 5 for the minutes you play against that unit. The Corsi QoC for the season will effectively be the average Corsi of all the players you have been facing weighted by the amount of time you have faced them.

Corsi Rel instead takes into account the relative Corsi of the opponents. Corsi Rel as earlier mentioned is only an intra-team comparison stat that shows which players are best at puck possession within the team. So a 3rd-liner in LA will have better Corsi than a 1st liner in TOR, but the TOR player will likely have a better Corsi Rel.

For example Jarret Stoll has a Corsi of 8.04 and a Corsi Rel of -12.6 while Phil Kessel has a Corsi of -6.34 and a Corsi Rel of 0.9.

I almost exclusively use Corsi Rel QoC since due to divisional schedules, everyone doesn't play the same teams the same number of times which can skew the raw Corsi QoC.

DatsyukToZetterberg 03-05-2013 05:15 PM

Quote:

Originally Posted by Blue Blooded (Post 61027201)
Corsi QoC only takes in the raw Corsi rating of the opponents, so if you match up against a unit that has averaged a Corsi of 5 during the season, you will have a Corsi QoC of 5 for the minutes you play against that unit. The Corsi QoC for the season will effectively be the average Corsi of all the players you have been facing weighted by the amount of time you have faced them.

Corsi Rel instead takes into account the relative Corsi of the opponents. Corsi Rel as earlier mentioned is only an intra-team comparison stat that shows which players are best at puck possession within the team. So a 3rd-liner in LA will have better Corsi than a 1st liner in TOR, but the TOR player will likely have a better Corsi Rel.

For example Jarret Stoll has a Corsi of 8.04 and a Corsi Rel of -12.6 while Phil Kessel has a Corsi of -6.34 and a Corsi Rel of 0.9.

I almost exclusively use Corsi Rel QoC since due to divisional schedules, everyone doesn't play the same teams the same number of times which can skew the raw Corsi QoC.

Ahhhhh, okay. So if you want to compare players on the same team you use the regular one but if you want to compare players across the league you would use the Rel one?

Blue Blooded 03-05-2013 05:25 PM

Quote:

Originally Posted by DatsyukToZetterberg (Post 61028197)
Ahhhhh, okay. So if you want to compare players on the same team you use the regular one but if you want to compare players across the league you would use the Rel one?

I guess you could look at it that way, but I'd generally use Corsi Rel for QoC no matter what.

If a player gets put in a shutdown role while the team plays high Corsi teams and then gets demoted afterwards it might skew his Corsi QoC greatly upwards, but it wouldn't affect his Corsi Rel QoC that much.

The Corsi differences between players are generally much greater than the differences in Corsi Rel which gives a higher probability of outliers skewing the average. Higher variance.

SmellOfVictory 03-05-2013 09:06 PM

Quote:

Originally Posted by DatsyukToZetterberg (Post 61016043)
Ah makes more sense now, so if has a - corsi he is not as good at controlling the puck as someone with a + corsi? And what's the main difference between any of the stats with Relative in it? Ie: Corsi Rel QoC - Corsi QoC

Yup, that's right. Controlling the puck being anything from hitting and taking it away, to Datsyuk-style controlling, to winning faceoffs, making good outlet passes, getting into shooting areas, etc.

In the simplest terms, it would be fair to say that any player (with a 50% offensive zonestart rate) with a raw Corsi of exactly zero could be considered a completely average player at whatever level he's playing (e.g. a guy playing against top 6 players regularly with a Corsi of 0 would be a dead average top 6 player). Positive Corsi rates would indicate progressively better than average players, and vice versa for the negative rates.

When you get used to looking at the numbers and comparing them obsessively, you can make all kinds of fun inferences and determinations, but that's a good starting point. There are also very few players who play in "normal" (50% zonestarts, etc) circumstances, though, so it's definitely just a starting point. You can find Corsi rates normalized for things like zonestart, but I don't know if there's a consistently available resource or if it's more just individuals doing the dirty work when they feel like it.

Hammer Time 03-05-2013 09:15 PM

Quote:

Originally Posted by Rutabaga (Post 60723093)
For the record, it is obviously possible to be playing in the post-season despite having a Fenwick ratio under 50%.
But they're not that many. (4 last year, 2 in 2011...)

Its stating the obvious, but you cant expect good and consistent results, if you're outshot that often at ES without some magic recipe. The shortened season is fooling some people, but right now, teams like Anaheim, Toronto and Nashville are good ideas if you want to bet about a "suprising" fall in the standings or a first-round exit if the regression to the mean doesnt happen early enough. Which is totally possible in a 48-game schedule. But its very likely that their seasons are just a fluke.

These teams (the ones under 50%, not those 3 in particular) clearly overachieved during the regular season, and they basically never repeat their success during the playoffs.

The ones that managed to win several (well, thats never more than 2) rounds always did so because they ended up playing against other team(s) under 50% (Except Pittsburgh under Therrien and Montréal with Halak in the nets. Not sure that models you can/want to replicate.)

Over 70% of the teams playing in the PO have a Fenwick score over 50% and thats probably the safest tool to see which teams are actually controlling the play, and which ones are not.

Doesn't Hiller have one of the highest career EV Sv% in the NHL? That would make him more likely than most other goalies to go on a Halak-like run, and therefore the Ducks could "get away" with having a low Fenwick.

Sidney the Kidney 03-06-2013 05:18 PM

Quote:

Originally Posted by SmellOfVictory (Post 61008347)
Corsi is generally considered an analogue for puck possession. The better it is for a player or team, the more often that team has the puck (particularly in the offensive zone) when the player is on the ice (or simply the more often that team has the puck, period); in other words, the better at hockey the player/team is. Corsi Rel is the same deal except it's a differential between the player and the rest of his team (Corsi while player is on the ice minus Corsi while player is off the ice); it helps illustrate where a player ranks amongst his teammates in terms of puck possession.

Important things to note with Corsi & Corsi Rel are zonestarts and qualcomp/qualteam. They provide a lot of needed context, as a raw Corsi rating could easily make a 4th liner look better than a 1st liner in a vacuum.

QualComp confuses me a little, but it's some esoteric thing based on the ratings of a player's average opponents. Higher is tougher, although it is not the most reliable measure (better than nothing, in most cases).

The QualComp usually confuses me because guys who are basically attached at the hip each shift don't always have the same (or even similar) QualComp ratings.

For instance, Pascal Dupuis and Chris Kunitz are Crosby's every game linemates. Their current QoC score is 0.005, each. Meanwhile, Crosby's QoC is -0.019. This confuses me for two reasons (that aren't really connected)

1-How do their scores differ so significantly from Crosby's, when they're regular linemates? Keep in mind the above is their QoC at 5 on 5, so it doesn't take into account any PP time or PK time.

2-How is Crosby's score so low when he seems to always be matched up head-to-head with the other team's top line? Every time the Pens play Tampa, he's head to head with Stamkos. Same with Philly and Giroux. Not only that, but the opposition coach always puts their best defensive pairing on the ice against him. So why does he have the 5th worst QoC on the Pens (for players with 10 or more games played)?

This is what confuses me about these advanced stats. Sometimes the numbers don't match up with the "eyeball" test.

Blue Blooded 03-07-2013 10:49 AM

Quote:

Originally Posted by Sidney the Kidney (Post 61115861)
The QualComp usually confuses me because guys who are basically attached at the hip each shift don't always have the same (or even similar) QualComp ratings.

For instance, Pascal Dupuis and Chris Kunitz are Crosby's every game linemates. Their current QoC score is 0.005, each. Meanwhile, Crosby's QoC is -0.019. This confuses me for two reasons (that aren't really connected)

1-How do their scores differ so significantly from Crosby's, when they're regular linemates? Keep in mind the above is their QoC at 5 on 5, so it doesn't take into account any PP time or PK time.

2-How is Crosby's score so low when he seems to always be matched up head-to-head with the other team's top line? Every time the Pens play Tampa, he's head to head with Stamkos. Same with Philly and Giroux. Not only that, but the opposition coach always puts their best defensive pairing on the ice against him. So why does he have the 5th worst QoC on the Pens (for players with 10 or more games played)?

This is what confuses me about these advanced stats. Sometimes the numbers don't match up with the "eyeball" test.

1. Sid may take longer shifts, or he may be double-shifted in special situations. Plus the difference is very small, I wouldn't call it significant.

2. While Bylsma might want Crosby out against Stamkos, Boucher hardly wants Stamkos out against Crosby. Opposing coaches are likely sending out their shutdown-units against him and these usually have very low Corsi Rel.

SmellOfVictory 03-07-2013 05:17 PM

To add to #2: top line players will always end up facing each other at some point as a function of the amount of time they spend on the ice, which can give a deceptive appearance, even if they're often getting a checking line instead.

Also, QUALCOMP does have its issues (especially over relatively short sample sizes); my understanding of it is that it may rank an excellent 4th liner the same as a good top liner because it's basing its rating off those players' Corsi ratings. Now that should sort of work itself out when you're looking at a league-wide sample of a player's opponents, but it's far from perfect. That's why it's good to look at it with a substantial grain of salt.


All times are GMT -5. The time now is 09:16 PM.

vBulletin Copyright ©2000 - 2015, Jelsoft Enterprises Ltd.
HFBoards.com, A property of CraveOnline, a division of AtomicOnline LLC ©2009 CraveOnline Media, LLC. All Rights Reserved.