Why are people so obsessed with so called advanced statistics?

[TT1]

Oh, this explains a lot. You don't understand how corsi works or how to use it. It's clear because you aren't making any sense.It's the team's corsi during the entire regular season that matters (actually the last 25 games of the regular season is even better), not their corsi during the series against the team they play. Also, score adjustments is a big thing for a lot of the individual mashups you have posted, because it's a known and measured effect, and it dictates that teams that are down during a game will start taking the greater share of shot attempts, and the greater the lead of their opponent the greater is the effect. So for a lot of those, the losing teams spent a lot of time "chasing games" theresfore their raw corsi during the series looks inflated - because it is due to score effects.

Pittsbugh was number 2 in Corsi last year, not number 8. SJ was number 9 not #12. Notice other theams like Dallas, Nashviellle, and Tampa, were all top 10 as well and made it deep in the playoffs. Florida was actually 20th, NYR were 26th and Chicago were 15th. Not sure why you thought they were top 4.See link below (need to sort it for CF at 5v5)
https://stats.hockeyanalysis.com/teamstats.php?disp=1&db=201516&sit=5v5&sort=CFPCT&sortdir=DESC

Now the year before, Chicago was number 2 and Tampa Bay was number 4, not 6th and 11th like you posted.Vancouver was 19th - not a top team. St. Louis was 11th and lost to 16th Mini, but their corsi deference wasn't that big, so of course there is no surprise that the favorite didn't win. After all, it lets you forecast probability, it does not grantee a result. See for your self below:
https://stats.hockeyanalysis.com/teamstats.php?disp=1&db=201415&sit=5v5&sort=CFPCT&sortdir=DESC

And your last paragraph is strait up silly. No, no team gives up shot attempts on purpose and forfeits shot attempts due to their system - that is a myth. Same with players. There are teams who play the "limit shots to the outside" systems and still have great shot metrics and end up succeeding as predicted, and there are the reverse who have terrible metrics and end up sucking.



So you are basically trying to claim "actually the statistical model that predicts them as a likely winner is wrong, because I can come up with arbitrary reasons that they won that aren't that statistic". You can hold on to that, but it's not very convincing. Their corsi indicated that they would be a likely favorite to win during those years, and they did win twice! That confirms the model, not the reverse. How can you not see that?

wait what?

I'm talking about the playoffs (and i compared teams in H2H matchups), thus i posted their corsi's in the playoffs. You're the one that makes absolutely no sense. Why are you talking about regular season stats when those games have 0 relevance in head to head matchups? Regular season corsi numbers is a tool that people use to predict winners/losers in the playoffs.. but we already have all the numbers from the playoffs, why would i care about their regular season stats?

_________

"And your last paragraph is strait up silly. No, no team gives up shot attempts on purpose and forfeits shot attempts due to their system - that is a myth. Same with players."

In regards to this statement, look at Weber/Josi's numbers. Josi/Weber are regarded as weak CF players, but if you look closely at HD (high danger) chances/shots you can see that Josi/Weber are top players at limiting HD chances/shot attempts (among other Nashville dmen). Right there you can see an example of players/a team who prioritize defending high scoring areas (thus reducing HD shots/chances) while being weak Corsi players/teams. Their system prioritizes quality of shots over quantity of shots. That's an example of how context has an impact on Corsi, if you do a bit of research and look up other numbers that have an impact on Corsi you might figure stuff out on your own..

QoT/usage are other variables that have an impact on corsi.. and there are others.

P.S: Historically HD shots has a ~7 to 1 ratio over Corsi shots (in shooting percentage shooting %). There's another important stat you should add to your toolbox.

http://www.tsn.ca/statistically-speaking-measuring-d-by-quality-of-shots-against-1.391447

 

[Acallabeth]

Because it's a quick way for a fan to feel knowledgeable on something they often don't even see.

Pittsbugh was number 2 in Corsi last year, not number 8. SJ was number 9 not #12. Notice other theams like Dallas, Nashviellle, and Tampa, were all top 10 as well and made it deep in the playoffs. Florida was actually 20th, NYR were 26th and Chicago were 15th. Not sure why you thought they were top 4.See link below (need to sort it for CF at 5v5)
https://stats.hockeyanalysis.com/teamstats.php?disp=1&db=201516&sit=5v5&sort=CFPCT&sortdir=DESC

CF 5v5 actually gives a data a bit different from what you've posted, check it please, so we're discussing the same numberz.

And LAK were 1st in Corsi and won 1 game in playoffs, and SJS who won the West were 5th among Western teams, and Detroit (that won 1 game) had almost the same Corsi numbers as the Eastern finalist Tampa, and Toronto (#1 OA pick) had better Corsi than Washington. Corsi's hard to interpret with any kind of certainty, it has a high sensitivity with low specificity.

their corsi deference wasn't that big, so of course there is no surprise that the favorite didn't win. After all, it lets you forecast probability, it does not grantee a result.

Yeah, so it can be used for splitting good teams from bad teams, and that's mostly it

 

[fiveonfive]

wait what?

I'm talking about the playoffs (and i compared teams in H2H matchups), thus i posted their corsi's in the playoffs. You're the one that makes absolutely no sense. Why are you talking about regular season stats when those games have 0 relevance in head to head matchups? Regular season corsi numbers is a tool that people use to predict winners/losers in the playoffs.. but we already have all the numbers from the playoffs, why would i care about their regular season stats?

The value of corsi is that you you look at at a sample of it, and then predict the probability of future outcomes. You can't predict future outcomes of past events based on the stats of those past events. Regular season corsi forcasted future winners of the Stanley cup, how can you not see that? Also, did you miss the part about score effects? A lot of your post seams oblivious to them.

Comparing H2H match-ups (while ignoring obvious things like score effects) that are too small of sample sizes is the definition of posting non-sense that shows clear misunderstanding of the application of the statistic.

And to your edit. I don't know how where you are getting your info, but Nashville was 4th in the league in Corsi percentage last year. I don't know how anyone can call that a weak corsi team. How anyone can be hypocritical enough to claim that they were a weak corsi team and then say "if you look at the numbers, you can figure it out on your own" is really something. Weber and Josi are "known as poor possession players" in a sense that Josi has shot suppression trouble without Weber, and Weber has transition and shot generation trouble without Josi. But that's brought up when they are compared to the very elites of the game. It's not like they have bottom of the barrel possession stats like you are making it out to be. Also, I'm not sure what about "having a smaller share of HD shot attempts against" makes you believe that they give up shot attempts on purpose or that it unfairly skewed their possession metrics as a whole.

 

[fiveonfive]

Because it's a quick way for a fan to feel knowledgeable on something they often don't even see.


CF 5v5 actually gives a data a bit different from what you've posted, check it please, so we're discussing the same numberz.

And LAK were 1st in Corsi and won 1 game in playoffs, and SJS who won the West were 5th among Western teams, and Detroit (that won 1 game) had almost the same Corsi numbers as the Eastern finalist Tampa, and Toronto (#1 OA pick) had better Corsi than Washington. Corsi's hard to interpret with any kind of certainty, it has a high sensitivity with low specificity.


Yeah, so it can be used for splitting good teams from bad teams, and that's mostly it

Maybe I wasn't clear, but the link I posted sorted by CF% at 5v5 is exactly the numbers I described.

Corsi helps you decipher which teams' skaters are good at playing 5v5 hockey. No one is stopping you from looking at other indicators for special teams. goal tending, and general execution skill of offensive players. Posting some outliers without examining the entire data set does not suggest that the shot attempt based analysis can't be helpful. Also, every year there is a team that is actually playing way over their heads, where everybody keeps denying they are about to drop off - corsi exposes those teams way sooner than anti stats people care to admit. Same with under-performing teams that are about to get better.

 

[Elvs]

Maybe I wasn't clear, but the link I posted sorted by CF% at 5v5 is exactly the numbers I described.

Corsi helps you decipher which teams' skaters are good at playing 5v5 hockey. No one is stopping you from looking at other indicators for special teams. goal tending, and general execution skill of offensive players. Posting some outliers without examining the entire data set does not suggest that the shot attempt based analysis can't be helpful.

Not really. Not when the "outliers" are so many they can't be defined as outliers. Is every Kings player better than every Avs player at 5 on 5 hockey? I think not. Put any Kings player on last seaons Avs team and that guy is almost guaranteed to be another minus player (corsi). Hence why corsi is a flawed individual stat.

Also, every year there is a team that is actually playing way over their heads, where everybody keeps denying they are about to drop off - corsi exposes those teams way sooner than anti stats people care to admit. Same with under-performing teams that are about to get better.

Such as the Florida Panthers last season. Anyone watching them enough could see they were actually really good in transition. But the stats will tell you otherwise.

 

[4thline]

I have no idea what you are trying to say.

This is called building a statistical model

The base model is pretty established wrt to corsi being good. It's the "why" and "how do I make a better team" that current practices currently ignore, and it doesn't require building a model to fix, just a more case by case method of analysis, actual situational data rather than relTM and QoC adj. Stratify a players results by the aggregate of ice time with quality of competition, sub stratify that with quality of teammates based on the aggregate of icetime, qualify it for zone and score factors, find the competitive imbalance driving that players statistics either good or bad, either exploit it or correct it as your roster allows.

Spent a lot of time on McCurdy's site last night and its by far the best around, it's one step away from being IMO the perfect tool for the statistical side of player evaluation

 

[fiveonfive]

Not really. Not when the "outliers" are so many they can't be defined as outliers. Is every Kings player better than every Avs player at 5 on 5 hockey? I think not. Put any Kings player on last seaons Avs team and that guy is almost guaranteed to be another minus player (corsi). Hence why corsi is a flawed individual stat.

I'm not sure where you got the idea that corsi means that "every Kings player is better than every Avs player". That would be a clear misunderstanding of what the statistic is, how it behaves, and how you read it.

Corsi tells you that the kings as a group (as in the ENTIRE LA KINGS TEAM - all skaters that ever step on the ice collectively) are better than the Avs are as a group (as in the ENTR at 5v5 play (as in the ENTIRE AVS TEAM - all skaters that ever step on the ice collectively). With that reference point, you can look at every playr's individual corsi compared to your teammates and see a pretty clear picture of who drives that play. It shows-up very clearly in relTM and rel numbers. There is nothing flawed with the stat, it's your application of it that is flawed.

Such as the Florida Panthers last season. Anyone watching them enough could see they were actually really good in transition. But the stats will tell you otherwise.

Yeah, the same way the Leafs and the Avs used to be good at keeping shots to the outside right? As in, not very good when compared to teams that are actually good at it.

What you're saying is similar to "Alex Ovechkin is actually not a very good shooter and scorer if you watch him, but stats will tell you other wise". Nope, that's not how that works.

 

[Analyst365]

It's a way for mansplainers to have something to mansplain about.

 

[LiveeviL]

It's a way for mansplainers to have something to mansplain about.

Seems as total overkill to go out discussing e.g. relative corsi if you want to mansplain.

 

[WesMcCauley]

And your last paragraph is strait up silly. No, no team gives up shot attempts on purpose and forfeits shot attempts due to their system - that is a myth.

Players do and they should do it more often. Giving up low % shots and not blocking them/screening your own goalie is the way to go alot of the times!
http://www.msgnetworks.com/videos/vallys-view-the-true-value-of-blocking-shots/

 

[fiveonfive]

Players do and they should do it more often. Giving up low % shots and not blocking them/screening your own goalie is the way to go alot of the times!
http://www.msgnetworks.com/videos/vallys-view-the-true-value-of-blocking-shots/

I enjoyed that video, it's a great look at what types of shot blocks are actually beneficial. Thanks for posting it!

But I don't think it showed what you though it showed, or at least it defiantly didn't take away from my point very much. I was stating that teams do not provide other teams higher shot attempt volumes on purpose, and in that video the opposition is always already in a position to attempt a shot - so it addresses the question of "When they attempt to score, do you try block or not block the shot attempt?" In all cases, none of those players would have granted an unopposed or undefended opportunity to shoot in the first place if they could strip the puck away or win a battle or deny a zone entry - so the effect on shot volume is the same.

The screening your goalie part actually corroborates that corsi is evidence of good play though, because as shown in the video - screened shots (with screens from your own teammates, or deflections, or rebounds (of the goalie or the screening players) - all complicate things for preventing goals against. So in any case, some shot attempt is better than no shot attempt.

 

[Elvs]

Corsi tells you that the kings as a group (as in the ENTIRE LA KINGS TEAM - all skaters that ever step on the ice collectively) are better than the Avs are as a group (as in the ENTR at 5v5 play (as in the ENTIRE AVS TEAM - all skaters that ever step on the ice collectively)

That's my point. Corsi is much more of a team stat than it is an individual stat.

Yeah, the same way the Leafs and the Avs used to be good at keeping shots to the outside right? As in, not very good when compared to teams that are actually good at it.

No, in the way Florida were fast, fluid and shifty through the neutral zone. Other factors could lead to their corsi being below .50, such as often being in the lead or not being aggressive enough in their own zone and thus not retrieving pucks very quickly.

 

[fiveonfive]

That's my point. Corsi is much more of a team stat than it is an individual stat.




No, in the way Florida were fast, fluid and shifty through the neutral zone. Other factors could lead to their corsi being below .50, such as often being in the lead or not being aggressive enough in their own zone and thus not retrieving pucks very quickly.

Yes, Corsi is a team stat. But players have control over the variation of it within their team - obviously team stats aren't independent of the players on the team. You can tell which players are improving the team stat and which are not, and by how much.

And fair enough, that may be a fair take. They were 18th in score adjusted corsi, so scored played some effect for sure - not enough to take them from "meh" to good though. I do agree they looked amazing in transition at the end of the year and during their first round loss to NY - but that was showing up in corsi at that point

 

[Kirikanoir]

Mark Twain said it best "There are three kinds of lies: lies, damned lies, and statistics."

 

[fiveonfive]

Mark Twain said it best "There are three kinds of lies: lies, damned lies, and statistics."

Mark Twain watched the games

 

[Lafleurs Guy]

The real question is why are people so obsessed with trying to show these numbers don't have value?

 

[6 Karlsson 5]

Not really. Not when the "outliers" are so many they can't be defined as outliers. Is every Kings player better than every Avs player at 5 on 5 hockey? I think not. Put any Kings player on last seaons Avs team and that guy is almost guaranteed to be another minus player (corsi). Hence why corsi is a flawed individual stat.




Such as the Florida Panthers last season. Anyone watching them enough could see they were actually really good in transition. But the stats will tell you otherwise.

no one else does either, you keep trying to bring down corsi by showing how idiots can misread the numbers, and no, I'm not saying that to sound smart. It is just painfully obvious that that is not how to interpret the numbers

 

[xxxx]

The one big problem with Corsi and every advanced stat that is based around shooting attempts (of course if you use that as the only argument without context and especially without watching the games) is that shots are not alike. Not every shot is dangerous the same way. If you shoot from a certain area, you are more likely to score than from a different area, for example. Another thing is if the shot is relatively easy - no body in front of the net, no players in the crease, no tip-ins, etc., hell, even a player's name is big difference. It's a difference to get outshot by Steven Stamkos or a player who scores 5-10 goals per season.

It is therefore very flawed. People tend to say that the more shoots you generally give up on your net, the more likely is the opposite team scores on you. But that's all "generally" "likely" "should be", it's all very relative. But the fact is, the advanced stats don't measure how much relatively dangerous a shot is. That's one of the reasons you need to watch the games. I don't care about probability, because if you want to "predict" something, that's very different from evaluating the actual players. Why are some teams generally good at outshooting the opposition (Toronto this year for example), but they don't win the games?

 

[Chojin]

If you've ever had an argument about Gretzky vs Lemieux, tried to predict the next year's standings, or evaluated a trade, then you should understand. It's fun to talk about hockey and argue your opinions, and so-called advanced stats are another tool in the toolbox.