Beyond Corsi - Let's Dig a Little Deeper

[cccsberg]

This has been a very rewarding season to date as a Flames fan, both with the success of the team and also as a new student of the burgeoning Advanced Stats crowd.  Its been fun to watch the team's success, or lack thereof and see how it corresponds with what the Advanced Stats gurus believe is success.  The correlation of Stats' success(high Corsi, or 'possession') often doesn't jive with actual wins and losses, and unfortunately most refuse to look deeper than the simple answers of high Corsi=success, contrary results is due to Luck (high PDO), and everyone will Regress to the Mean if they happen to be doing better than average.  

 

This is disappointing in that I believe there is a lot more to be gleaned from the data, and learned regarding a team's success or failure and whether or not that is sustainable.  To that end I've looked at a few things from War_on_Ice's mid-season data of a couple days ago and here is my take on where the Flames are at and a little glimpse at what we can perhaps glean from the Advanced Stats data.

 

First, a couple definitions:

 

Corsi Events: Shots on Net & Missed Shots & Shots Blocked

 

PDO: Combination of Goalie Save % & Team Shooting %

 

This first chart is what I like to call Effectiveness of Team Offence.  What it shows is the number of Corsi Events required per Goal Scored for each team in the NHL.  What it means, basically is that a team like Edmonton requires 25 Corsi Events to score 1 goal, whereas a team like Calgary only requires 19 to do the same.  Everything else being equal in a game situation,  Calgary would have a relative Corsi of 19/44 or 0.43 and people would be saying Calgary was dominated when really it should be a tied game.  Slightly better Corsi and it should be a victory, even with a negative Corsi.  

 

Hmmm, interesting.

 

This second chart is just the reverse, or what I like to call Effectiveness of Team Defence.  What it shows is the number of Corsi Events Allowed per Goal Against, for every team in the NHL.  What this means, in my scenario from before, is that Calgary can give up 23 Corsi Events for each goal they allow, whereas Edmonton can only give up 17 before getting scored on.  

 

Now there's good reasons why Calgary ranks 5th & 9th Overall in the NHL in both measures, and also why Edmonton ranks 28th & 30th overall.  These would include Save %, Shooting %, Blocked Shots and Missed Shots for sure, and they probably also provide clues to something more, like Systems, Schemes and Personnel.  I've got a bunch of that data, and my thoughts on it too, and if there is an interest I'll add it later.  

 

In any case, when you look at actual things that matter directly to wins and losses, like Goals Scored, Calgary IS NOT the crappy 28th place team (defined by relative Corsi alone) that is baffling the Advanced Stat nerds like Lambert and others over at FlamesNation(i.e. its all Luck, Regression to follow), but a solid team in the upper middle of the pack, just about where they are sitting in the league standings....

 

Let me know what you think.  Any thoughts yourselves?

 

[Pyromancer]

Your advanced stats are scaring me.

I have avoided delving deeply into advanced stats as I may succumb to several of my inherent weaknesses & I would be all in.

 

I will peruse your musings with great interest but for now I will be on the sidelines to prevent from crossing over to the dark side of statistics.

[cccsberg]

Your advanced stats are scaring me.

I have avoided delving deeply into advanced stats as I may succumb to several of my inherent weaknesses & I would be all in.

 

I will peruse your musings with great interest but for now I will be on the sidelines to prevent from crossing over to the dark side of statistics.

Yeah I know what you mean, its easy to get in deep.  I'm just doing it because I'm fed up with the usual lame excuses I keep hearing and wanted to see if the data (which may or may not be good...lol) could tell us anything more.

[cccsberg]

Here's the next layer of the onion, so to speak...

 

Looking in a bit more depth into the components of Corsi Events For, we see Calgary bottom of the NHL in % of their own Shots Blocked, 5th best in fewest of their own Shots Missed and overall near the bottom in % of their own shots actually getting through to the net.  Edmonton is also bottom-league in Shots For Blocked, only middling in Missed Shots For and therefore bottom-league in getting their own shots through on net.

 

The thing that's really interesting to me is the Missed Shots For data shows a lot of the strong Western teams at or near the bottom, which I would interpret as them having a focus of throwing a lot of pucks towards the net, whether they missed or not.  Contrast that with Calgary, Pitt, Chicago and TB... teams that seem to have more of a Counterattack focus.  When you combine that with overall Shots For and Corsi Events For(just above), we see those same Western teams at the top of the NHL it seems to back that thought up.  The two big exceptions are Chicago and Montreal, who both seem to succeed but for very different reasons....

 

The last piece of the puzzle (to this point) for Offence is the Shooting % stats, which are shown just above.  Calgary is 5th overall in Shooting % For and 7th overall in Goals For/ Corsi Event.  They come in at 10.1% and 5.2% respectively for those parameters.  Note again the California teams low down on both lists, reinforcing the quantity vs quality theory for them all.  I guess i'll have to dig further into Scoring Chances (i.e. Shot Quality) to see if that data supports my contention.

 

Thoughts again?  Next I'll look at Defence, e.g. Shots Against, which tells a pretty clear story as far as Calgary is concerned...

[cccsberg]

As you can see on the charts above, Calgary leads the NHL in Shots Against Blocked as % Corsi Events, is 4th in Missed Shots Against as % Corsi Events and also leads the NHL in lowest Shots Against as % Corsi Events.  The shot-blocking and fewest shots on net Calgary is at an Elite level, far surpassing the rest of the league.  In Missed Shots against they are top level.   When you also look at Montreal, it seems pretty clear that Shot Against Blocked is the more important factor, and this is where Calgary excels.

 

Just reflecting now, but this to me is the essence of the Flames success defensively this year, and the clear reason why they are able to withstand a rather poor Relative Corsi and still be successful.  Blocking shots and keeping teams to the outside (reasonable assumption re: high Missed Shots Against) is the cornerstone to the Flames.  They do it extremely well.  Even as their GAA has tapered off to just mediocre levels lately.  Even losing Corsi battles regularly, their effectiveness at deflecting and boxing out gives them a chance.

 

The other thing clear, especially if this is so, is that the individual Flames most responsible, and that would be Russel (1st in NHL) and Bouma (7th in NHL I believe) are extremely important pieces to the Flames line-up, far more than basic line stats would indicate...

[robrob1974]

What I think about hockey. What ever team makes good on quality chances, takes advantage and scores a pp goal, kills penalties are the teams that win games. Mostly, those rare 3-on-1's or2's are what counts the most.

The flames are a very young team and as they grow, so too will the advanced stats. To use the farming analogy, they're just seeds thus far. It's a good foundation to build on.

Most of the year we have relied on counterattack and making good on those chances, hard work and intensity had helped too.

I admit, I get scared often at how other teams spend their time in our end. But when we counter and I see Johnny coming up the ice, it's like art.

[cccsberg]

What I think about hockey. What ever team makes good on quality chances, takes advantage and scores a pp goal, kills penalties are the teams that win games. Mostly, those rare 3-on-1's or2's are what counts the most.

The flames are a very young team and as they grow, so too will the advanced stats. To use the farming analogy, they're just seeds thus far. It's a good foundation to build on.

Most of the year we have relied on counterattack and making good on those chances, hard work and intensity had helped too.

I admit, I get scared often at how other teams spend their time in our end. But when we counter and I see Johnny coming up the ice, it's like art.

Yes, it would be nicer if we spent more time in the other team's end, but at least if we're able to hold them at bay and capitalize when we get our chances we'll be doing better than not in wins and losses.

[KidKappa]

I think that the different stats movement has died down a bit, you don't see it getting the media attention it did at the beginning of the season.  I'm all for the different stats experiment, but I want to see, predictability, causality and concrete results.  If all it turns out to be is explanatory rather than predictability, then don't don't bother me.  The interesting thing would be if the different stats actually produce results and all the GM's and coaches embrace them if then further different stats are invented to un level the playing field again.  Lol. 

[kehatch]

If I can be a bit of a Debbie Downer here.....

 

We've had the arguments (for and against) the need/use of CORSI numbers for the general public before (right this way, ladies and gentlemen --> http://fans.flames.nhl.com/community/topic/21186-so-we-are-just-lucky).  I'm no fan of new-fangled stats either, but this topic is still a valid one for those who enjoy it.

 

If you wish to poke fun at/argue about/rage against the existence of CORSI, may I suggest we take it to the other forum, and leave this one for the "stats geeks".

 

 

* hides the rotten fruits and vegetables baskets *

* departs, stage left *

 

To add to this.  There is legitimate discussion around the relevance of advanced stats or the application of them.  Understandably there are people that don't like them.  They just want to enjoy the game and that is fine.  But there are lots of threads with those debates being had.  I have cleaned up this thread so those that enjoy statistics have a place to discuss them without being mocked or the discussion being derailed.  

[kehatch]

II am always interested in line data.  How players match up with one another.  "With or Without You" (WOWY) data can be really interesting.  It shows how performance changes based on who a player is on the ice with.  All data comes from: http://stats.hockeyanalysis.com/.  

 

All data is even strength. The size of the circle indicates the ice time played together.  The color designates the players position.  I used GF% for this.  It shows shows relative goals for versus goals against.  (Goals For / (Goals For + Goals Against).  The more players on the bottom right the better.  

 

HUDLER GF% WOWY EV

 

I was interested where Gaudreau was.  Hudler's GF% when on the ice with Gaudreau is   51.61% which is pretty good.  But it increases to 66.67% when not on the ice with Gaudreau.  I was curious so I ran Gaudreau's numbers.

 

GAUDREAU GF% WOWY EV

 

This graph shows the same thing.  The on ice GF% is of course the same (51.61%).  But when he is away from Hudler it increases to a ridiculous 76.92.  Another interesting note is that when Gaudreau is on the ice with Backlund his GF% is 80.0.  Its still early but that looks like a really good match.  Meanwhile Hudler is really good with Monahan (and has been for two years).  This really supports what we have seen the last couple of games with Gaudreau-Backlund-XXXX and XXXX-Monahan-Hudler as lines.  

 

Here is one more just for fun:

 

ENGELLAND GF% WOWY EV

 

This shows you how terrible Engelland has been.  He has brought EVERY player he has played with down except for Russell (and there isn't much time with Russell).  It also shows you how bad Diaz, Smid, and Bollig have been.  They are worse when on the ice with Engelland.  But they are pretty terrible even when they aren't on the ice with him.  

 

It would take up too much space for me to do this for every player.  But if anyone has requests let me know. Or better yet go to the link above and run them yourself!!  If you need help figuring out how to run them let me know.  I used GF% but you can look at just goals for, goal against, or CORSI.  You can also look at the charts adjusted for zone starts or isolate it to "Games Close" situations.  

[kehatch]

Here is one more of my favorite charts.  These are player usage charts from the following site: http://somekindofninja.com/nhl/usage.php?f1=2014_s&f2=5v5&f3=&f5=CGY&f4=C%2BLW%2BRW&f7=5-&bubbleType=corsiRel&yAxis=qoc&update-filters=Update+Results

 

What you are seeing is how a player is performing based on how they are being deployed.  As you move towards the right the player is getting more offensive zone starts.  As you move towards the bottom the quality of competition the player is facing goes down.  So players in the top left are playing difficult minutes while the guy in the bottom right have it pretty easy. 

 

The circle is the players Relative CORSI.  Blue is positive, red is negative.  The size of the circle indicates how positive or negative.  For those that don't know CORSI is = (Shots + Missed Shots + Blocked Shots Against) - (Shots against + missed shots against + Blocked Shots).  Basically it shows shot attempts for vs against.

 

Relative CORSI is the players CORSI - The Teams CORSI when the player isn't on the ice. Basically it is the players CORSI relative to the team he is playing on.   

 

FORWARDS

 

Kudos to Monahan, Glencross, Stajan, and Backlund.  They are thrown to the wolves every night but are still driving possession relative to the rest of the team.  

 

DEFENCE

 

I doubt there are many surprises here.  Gio and Brodie are awesome.  The rest of the D really struggles 5v5

 

I showed the data 5v5.  But you can look at PP and PK data if you want to.  Some interesting stuff (at least IMO).   

[cccsberg]

I agree with you...  

Hmmm, interesting.  It clearly puts the emphasis on scoring goals, and being ahead, which after all is the point.  Also shows that the Corsi, Fenwick and Shots stats not very diagnostic to Wins , as the correlation of Corsi win to actual Win is never more than about 55%.  Also an interesting dip late in the game, I'm thinking that's Score Effects?

To add to this.  There is legitimate discussion around the relevance of advanced stats or the application of them.  Understandably there are people that don't like them.  They just want to enjoy the game and that is fine.  But there are lots of threads with those debates being had.  I have cleaned up this thread so those that enjoy statistics have a place to discuss them without being mocked or the discussion being derailed.  

OK Thanks.  Should provide a better environment to actually look at several things and get feedback, which would be useful.

[tmac70]

II am always interested in line data.  How players match up with one another.  "With or Without You" (WOWY) data can be really interesting.  It shows how performance changes based on who a player is on the ice with.  All data comes from: http://stats.hockeyanalysis.com/.  

 

All data is even strength. The size of the circle indicates the ice time played together.  The color designates the players position.  I used GF% for this.  It shows shows relative goals for versus goals against.  (Goals For / (Goals For + Goals Against).  The more players on the bottom right the better.  

 

HUDLER GF% WOWY EV

 

I was interested where Gaudreau was.  Hudler's GF% when on the ice with Gaudreau is   51.61% which is pretty good.  But it increases to 66.67% when not on the ice with Gaudreau.  I was curious so I ran Gaudreau's numbers.

 

GAUDREAU GF% WOWY EV

 

This graph shows the same thing.  The on ice GF% is of course the same (51.61%).  But when he is away from Hudler it increases to a ridiculous 76.92.  Another interesting note is that when Gaudreau is on the ice with Backlund his GF% is 80.0.  Its still early but that looks like a really good match.  Meanwhile Hudler is really good with Monahan (and has been for two years).  This really supports what we have seen the last couple of games with Gaudreau-Backlund-XXXX and XXXX-Monahan-Hudler as lines.  

 

Here is one more just for fun:

 

ENGELLAND GF% WOWY EV

 

This shows you how terrible Engelland has been.  He has brought EVERY player he has played with down except for Russell (and there isn't much time with Russell).  It also shows you how bad Diaz, Smid, and Bollig have been.  They are worse when on the ice with Engelland.  But they are pretty terrible even when they aren't on the ice with him.  

 

It would take up too much space for me to do this for every player.  But if anyone has requests let me know. Or better yet go to the link above and run them yourself!!  If you need help figuring out how to run them let me know.  I used GF% but you can look at just goals for, goal against, or CORSI.  You can also look at the charts adjusted for zone starts or isolate it to "Games Close" situations.  

So just asking, not to create a fight. Engelland's numbers show how horrible he is( which I have indicated numerous times). You have made mention that BT is a advance guy and uses them. Then why sign Engelland to a contract if the stats clearly show that he is an inferior hockey player? Did his play decrease that much since leaving Pitts?

[cccsberg]

II am always interested in line data.  How players match up with one another.  "With or Without You" (WOWY) data can be really interesting.  It shows how performance changes based on who a player is on the ice with.  All data comes from: http://stats.hockeyanalysis.com/.  

 

All data is even strength. The size of the circle indicates the ice time played together.  The color designates the players position.  I used GF% for this.  It shows shows relative goals for versus goals against.  (Goals For / (Goals For + Goals Against).  The more players on the bottom right the better.  

 

HUDLER GF% WOWY EV

 

I was interested where Gaudreau was.  Hudler's GF% when on the ice with Gaudreau is   51.61% which is pretty good.  But it increases to 66.67% when not on the ice with Gaudreau.  I was curious so I ran Gaudreau's numbers.

 

GAUDREAU GF% WOWY EV

 

This graph shows the same thing.  The on ice GF% is of course the same (51.61%).  But when he is away from Hudler it increases to a ridiculous 76.92.  Another interesting note is that when Gaudreau is on the ice with Backlund his GF% is 80.0.  Its still early but that looks like a really good match.  Meanwhile Hudler is really good with Monahan (and has been for two years).  This really supports what we have seen the last couple of games with Gaudreau-Backlund-XXXX and XXXX-Monahan-Hudler as lines.  

 

This shows you how terrible Engelland has been.  He has brought EVERY player he has played with down except for Russell (and there isn't much time with Russell).  It also shows you how bad Diaz, Smid, and Bollig have been.  They are worse when on the ice with Engelland.  But they are pretty terrible even when they aren't on the ice with him.  

 

It would take up too much space for me to do this for every player.  But if anyone has requests let me know. Or better yet go to the link above and run them yourself!!  If you need help figuring out how to run them let me know.  I used GF% but you can look at just goals for, goal against, or CORSI.  You can also look at the charts adjusted for zone starts or isolate it to "Games Close" situations.  

Love the charts, all very interesting.  The Hudler-Gaudreau comparisons are intriguing.  They both do well together, but MUCH BETTER apart.  The only thing I can think of is when they are apart the Quality of Comp changes considerably, thereby the RELATIVE improvement.  

[cccsberg]

Here is one more of my favorite charts.  These are player usage charts from the following site: http://somekindofninja.com/nhl/usage.php?f1=2014_s&f2=5v5&f3=&f5=CGY&f4=C%2BLW%2BRW&f7=5-&bubbleType=corsiRel&yAxis=qoc&update-filters=Update+Results

 

What you are seeing is how a player is performing based on how they are being deployed.  As you move towards the right the player is getting more offensive zone starts.  As you move towards the bottom the quality of competition the player is facing goes down.  So players in the top left are playing difficult minutes while the guy in the bottom right have it pretty easy. 

 

The circle is the players Relative CORSI.  Blue is positive, red is negative.  The size of the circle indicates how positive or negative.  For those that don't know CORSI is = (Shots + Missed Shots + Blocked Shots Against) - (Shots against + missed shots against + Blocked Shots).  Basically it shows shot attempts for vs against.

 

Relative CORSI is the players CORSI - The Teams CORSI when the player isn't on the ice. Basically it is the players CORSI relative to the team he is playing on.   

 

FORWARDS

 

Kudos to Monahan, Glencross, Stajan, and Backlund.  They are thrown to the wolves every night but are still driving possession relative to the rest of the team.  

 

DEFENCE

 

I doubt there are many surprises here.  Gio and Brodie are awesome.  The rest of the D really struggles 5v5

 

I showed the data 5v5.  But you can look at PP and PK data if you want to.  Some interesting stuff (at least IMO).   

Nice charts.  On both I think if you utilized %SOG(SF/(SF&SA)) vs Relative Corsi you'd see dramatically different results between players because both Russell(1st in NHL) and Bouma (7th? in NHL) block a tremendous amount of shots.  Some argue that's bad, i.e. just shows they don't have possession, but I'd argue they are relatively more effective defensively by the shot-blocking.  Also, with the defenders, I believe Smid & Engelland get a disproportionate share of minutes SH so I'd be curious to see the results for the same chart but Short-Handed minutes only.

[kehatch]

Hmmm, interesting.  It clearly puts the emphasis on scoring goals, and being ahead, which after all is the point.  Also shows that the Corsi, Fenwick and Shots stats not very diagnostic to Wins.  Also an interesting dip late in the game, I'm thinking that's Score Effects?

 

I don't think you can correlate any individual metric directly to wins.  The full article is here: http://rinkstats.blogspot.ca/2014/05/corsi-and-fenwick-suck-or-why-we-should.html and it is a good read.  

[kehatch]

So just asking, not to create a fight. Engelland's numbers show how horrible he is( which I have indicated numerous times). You have made mention that BT is a advance guy and uses them. Then why sign Engelland to a contract if the stats clearly show that he is an inferior hockey player? Did his play decrease that much since leaving Pitts?

 

Great question.  I should have written "how horrible has has been" vs "how horrible he is".  I don't watch a lot of Penguin's game so I don't know first hand, but by the numbers he was a little better last season. (Though still not very good).  He got to play with Orpik and Niskanen primarily.  Both are an upgrade to Smid or Diaz.  

 

But I think the simple answer is that the decision to sign him wasn't based on analytic's.  It was an old school decision to get bigger and bring someone in with "intangibles".  

 

No matter how you cut it, this was a bad contract and a bad decision.  

[cfreining]

So just asking, not to create a fight. Engelland's numbers show how horrible he is( which I have indicated numerous times). You have made mention that BT is a advance guy and uses them. Then why sign Engelland to a contract if the stats clearly show that he is an inferior hockey player? Did his play decrease that much since leaving Pitts?

Short contract, buys time to develop our D prospects.

We can even develop our F prospects and trade for a missing piece. We are in a rebuild, so forget about those 2 to 3 year contracts. They are only here to teach their experience.

[Khrox]

Kudos to Monahan, Glencross, Stajan, and Backlund.  They are thrown to the wolves every night but are still driving possession relative to the rest of the team.  

 

Also, feel we should give kudos to Byron as well, while his quality of competition is the lower than the 4 you mentioned (but not that much lower than Backlund), he also drives possession on a level similar with Monahan and better than Stajan or Glencross (though Glencross faces far harder competition so I still give give Glencross the edge there).

 

 

Also a little something I'm sure I can't be the only one noticing. Seto faced the hardest Quality of Competition on the team, and drove possession still. A Significantly higher Offensive Zone starts but still, that was a little surprising and goes to show you that Hartley wasn't just talking out his rear end when saying that Seto was working hard, even if he didn't get the points.

[cccsberg]

Bringing to the top in case any FN visitors wish to view...

[DirtyDeeds]

Bringing to the top in case any FN visitors wish to view...

FN??  

 

Fussy Nitpickers??

[mike_jones]

FN??  

 

Fussy Nitpickers??

FN = False numbers

[Wreckening]

I thought FN stood for Flames Nation 

[cccsberg]

Yes, FlamesNation, another site I follow.  They are doing an article on Corsi and etc, so instead of rewriting everything for that site, which is too much work, I just directed them here.  This site is way better, with better commentary and analysis IMHO, so good on you guys overseeing it all, and Thanks!

[kehatch]

The NHL will be putting advanced stats on the main NHL stats site effective February.