A Graphic Guide To Advanced Hockey Stats

Graphic Comments
October 10 2013 01:20PM

I WATCH THE GAMES!!!!1

If you're a regular reader here at Canucks Army, you are likely familiar with most of the new stats that we use to analyze players and games. Sure, everyone likes a good story, but good stories don't win you the Cup. Not cinderella stories, not chronicles of heroism, and certainly not a tale of destiny. Sigh.

No, what wins you the Cup is good old fashioned luck.

Ok, maybe I'm oversimplifying that, but when you did down to the root of the matter, the NHL results are indistinguishable from a league in which 76% of the games are decided by random chance. In other words, the better team will win 24% of the time due a higher skill level. The rest of the time, it's a toss up.

That means in a seven game series, the better team will win one, maybe two, games due to a skill advantage. The rest is up to Lady Luck.

So keep that in mind next time you hear another storyline disguised as analysis from an expert on a TV panel:

Stuff that actually happened

And that's really at the heart of the work that has gone into tracking and analyzing advanced hockey statistics. It's just some pretty knowledgeable hockey fans trying to find ways of quantifying what actually happened.

Traditional Stats

Anyway, back to the role of luck in the NHL and why that even matters. If you consider that team level results are simply the sum total of individual player performances, it is not too great a stretch of logic to assume that luck plays a pretty significant role at the player level as well. 

The fact is that goals are pretty rare events in a hockey game. And given that luck plays such a big role in whether or not a puck actually gets into the net, the level of uncertainity around goal scoring is going to be very high. This is, I believe, the single biggest non-intuitive idea you need to grasp in order to understand advanced hockey stats.

Yes, skill clearly plays a part. But luck plays an even larger part. That's why it's not really useful to base any kind of meaningful analysis on goals, which are just a likely to bounce in off a leg as to be the result of some measurable skill on the part of the shooter. And if goals on their own are almost meaningless, then don't get me started on plus/minus:

The most useless of hockey stats

That's just one of the traditional hockey stats that is constantly referenced to fit some predetermined narrative or other. The worst example is when the so-called experts ADD +/- STATS TOGETHER to say that player A and player B are a combined -251 in the last year. This is especially frustrating because the players being aggregated are usually linemates or defense partners, which means that they were probably on the ice for many of the same goals for/against that go into that particularly heinous stat. GAAAAAAAAAH!

Ok, deep breaths. Let's not talk about +/- any more.

Of course, most of the other common stats that on-air analysts refer to are just as meaningless. You need look no further than Saturday's post-game show following the Canucks/Oilers game, where Glen Healy actually said these words:

"When you look at the stats, every stat was in Edmonton’s favour. Faceoffs, they were better in faceoffs, They were better in hits. They had more blocked shots. But they weren’t even in the game."

Let that sink in for a second.

This is a game that the Canucks clearly dominated from start to finish. Yet the stats that Healy counts on were ALL in Edmonton's favour. Shouldn't you think about looking a different stats then, Glen?

Here's what you would have seen if you looked at just one of the advanced stats that's available:

Canucks vs Oilers Fenwick 

Does that maybe match up a bit better with what your eyes told you on Saturday night? Here's the rest of the advanced stats on the game, which all show what a dominating performance that was by the Canucks.

That chart and data is from the excellent new Extra Skater website that is wrapping a really easy to use interface and navigation around stats that have been tracked for years on Behind The Net, Time On Ice and Hockey Analysis.

Shooting Stats

So what stats do matter? Well, there's two key components factors that go into making a stat useful. The first is that it should have some relationship to the outcome of hockey games. The second is that it should help us weed out the effects of luck in some way. Oh, and I guess there's one more rather important factor to consider: is it countable?

Probably should have started with that last one...

It turns out that shooting at the net meets all of our requirements. It should be intuitively obvious that you are more likely to win a hockey game if you have the puck. Not only does that mean you have an opportunity to score, but the opposing team doesn't. The NHL used to track possession times, but they stopped a few years ago. However, shot attempts are a pretty good proxy for puck possession. Clearly the more shots you take at the net, the more you have the puck.

Don't believe me? Well, J.P. Nikota over at Pension Plan Puppets got out his trusty stopwatch and timed possession times for the Leafs' games last year. This is what he found when he plotted offensive zone time with shot attempt differential:

Now that's what I call correlation

That's a pretty good correlation, I'd say.

The other reason shots and shot attempts are a good statistical base for analyzing hockey is that unlike goals, they are not rare events. There are usually over 70,000 shots on goal in the NHL every year, and about double that number in shot attempts. Compare that with just under 7,000 goals.

Why is the number of events so important? Because the larger the sample, the less impact randomness and errors will have on the results. So looking at shot attempts helps to smooth out the effects of chance or luck that you get when focusing just on goals.

Finally, the NHL actually counts shots and shot attempts for us! So really the only advanced part of most advanced stats is that the use of addition and subtraction:

There's value in plus minus after all.

I guess plus minus is useful after all.

What The Hell Is A Corsi?

Now, you might have looked at that chart from J.P. Nikota above and thought to yourself, "What the hell is a Corsi?" Well, I'm not going to get into where the term came from, but suffice it to say that Corsi refers to shot attempt differential. At the team level, that just means you add up all shot attempts for one team, including those that were blocked or missed the net, and subtract the shot attempts for the other team. At the player level, you do the same thing for all shot attempts for and against while that player was on the ice. The result can be expressed as a positive or negative number, or more often as a percentage. The latter is calculated simply by dividing shot attempts for by total shot attempts by both teams.

There are a number of variations on Corsi that try to factor in things starting from a face-off in the offensive or defensive zone, quality of teammates, quality of opposition, etc. One of the more common is CorsiRelative. Note, this is has nothing to do with that drunk uncle that shows up to all your family events:

What the hell is a Corsi anyway? 

No, CorsiRelative is the difference between an individual player's Corsi while he is on the ice and the rest of the team's Corsi while he is off the ice. This helps to pick out players that might be driving possession even though they play on a horrible team.

Another key filter applied to shooting stats is the game state. First of all, most shooting stats only account for things that happen at even strength. Clearly if one team is on the power play, they are going to have the puck most of the time and get the majority of the shot attempts. So you don't want that to skew the overall results. The other thing that tends to affect possession and shot attempts is the score. There has been plenty of work to show that when a team is comfortably ahead, they tend to sit back and protect the lead.

So you will sometimes see people referring to score-tied or score-close Corsi. This just means that only shot attempts while the score is tied or close (within one goal) are counted. Score effects are something that Canucks fans are very, very familiar with:

PANIC!!!

Just don't tell Canucks fans that all teams in all sports go into a defensive shell when they're comfortably ahead.

Ok, I Get Corsi, But Fenwick? WTF Is That?

Fenwick is just like Corsi except it does not include blocked shots. Here's a handy chart to help you tell the difference between all the advanced shot stats: 

What about tequila shots?

If anybody was counting shots that bounced at least once on the way to the net, we'd have to call that a Lidster.

Anyway, based on quite a bit of statistical analysis over the last few years, Fenwick Close or Fenwick Adjusted (for score effects) are generally accepted as the best predictive stats over a given NHL season. That's why you'll often see Fenwick Close referenced when talking about a given team's performance to date and how you might expect them to perform in the future.

PDO: The Most Obscure Of All Advanced Stats

Look, I'm not going to explain this one. If you want and explanation, our very own @camcharron provided one over at Backhand Shelf earlier this year. Suffice it to say that over time the sum of a team's shooting percentage and save percentage will asymptotically approach a value of 1. Sometimes you will see it represented as values around 1,000 but that's just for people that hate decimal points.

Why is PDO useful, well, as Cam put it at the top of weekly tabulation of PDO numbers by team last season:

Basically, if a team is playing with a PDO number way higher than 1.000, they're producing above their expected output. If a team is playing with a PDO number below 1.000, they're producing below their expected output. Over the course of a long season, the number will generally correct itself.

Although PDO was developed as a team level stat, it applies at the individual player level as well. If your favourite player appears to be struggling, check out his PDO. If it's really low, he's probably just had a stretch of bad luck where he and his linemates have had a really low shooting percentage and/or the goaltending has been below par while he's on the ice. Over time, the bounces will even out.

As for what PDO actually stands for, nobody really knows, but I think it's pretty darn obvious:

Pretty darn obvious what PDO stands for

Extra Skater includes the PDO calculation in their advanced stats tables at the team and invididual player level.

But What About Shot Quality

Yes, yes. Not all shots are created equal. A screen shot from the blue line is not the same as an open look from between the hash marks. True.

But two things to consider. First, while shot quality is undoubtedly a thing, it's impact is actually quite small. You can find pages and pages online of pixels spilled in the shot quality debate. But at the macro level, any impact of shot quality is going to be lost in the noise and normal fluctuation produced by luck and any other factors that are not quantifiable. And that's the second thing, nobody has found a way to quantify shot quality.

Is it based on distance? Location? Just an innate ability? Well how would you measure that last one?

One of the most objective attempts to filter out some of the noise and focus on shots that should matter has been the work done to count scoring chances over the last few years. The recaps right here at Canucks Army include counts of chances for and against for most games. Last year, these results were combined with similar counts from around the league to allow for a meta analysis of scoring chance data.

The results showed that scoring chances correlated really, really closely with Fenwick. So, sure, you're much more likely to score on a shot from the roughly triangular region between the net and the two face-off circles, but you're also more like to get scoring chances proportionately to your overall shot attempts. Basically you should draw similar conclusions from scoring chances as you would from the shot counting stats, but counting chances is way more labour intensive while shot counting stats are readily available.

That being said, there is one factor that is very strongly correlated to scoring chances:

Experience may differ at the Roxy.

Hopefully this provided a little more insight into what the most common advanced stats are and why we use them as the basis for our analyses here at Canucks Army. There are plenty of resources around the web on the topic, and the NHL Numbers site on the Nations network is one of them.

One other site that you might find cool is ShiftChart.com, which lets you see who was on the ice at any point during a game in a super easy to use and very versatile graphical format.

We're on the verge of a breakthrough in the tools available to generate some real insights on what drives results in the NHL. So if you're at all interested, dig in and play with the numbers yourself. Either way, keep some of this in mind next time Glen Healy seems perplexed when he's looking at stats sheet.

Maybe he should stop worrying about face off percentage, hits and blocks shots and just watch the games.

 

RECENT GRAPHIC COMMENTS

C867930d1f1ba5cf2b078a93d3bd1a68
I'm not a hippie or on welfare. I don't live in Kits, wear Birkenstocks or own an umbrella. I've never been to the Capilano Suspension Bridge, but I'm sure it's very nice. I have a mayor, not a crack addict. I drink pale ale, not Blue. And I call it a cabin, not a cottage. I can proudly say my team's been to the Stanley Cup Final in the last 45 years. They may not have won, but at least they got there. I believe in sunshine, not haze; heat, not humidity. And that sushi is a healthy and tasty meal. A coho is a fish. A ski hill is a mountain. And the plural of leaf is leaves. Okay? Not leafs. Leaves! Vancouver is the country's third-largest city, certainly the most beautiful, and the best part of Canada! My name is petbugs and I am a Canucks fan! ... You can find me on Twitter @petbugs13 or send your hate mail to petbugs (at) gmail (dot) com but it better be funny or it's getting plonked.
Avatar
#1 Peachy
October 10 2013, 01:52PM
Trash it!
4
trashes
+1
14
props

Whatever you get paid, it's not enough. Nice work.

Avatar
#2 Mironovs Nose
October 10 2013, 05:05PM
Trash it!
1
trashes
+1
12
props
Mac daddy wrote:

"No, what wins you the Cup is good old fashioned luck."

Lmao! Are you serious? If you ever want to discredit an entire article right off the bat, write something idiotic like that!

I guess hard work and dedication has nothing to do with it then?

For petes sake....

Clearly given any two teams, one team has worked harder and is more dedicated than the other, and that team always wins. Unless the other team just wants it more, in which case the team with the most intangibles wins.

Avatar
#3 Mac daddy
October 10 2013, 03:56PM
Trash it!
14
trashes
+1
9
props

"No, what wins you the Cup is good old fashioned luck."

Lmao! Are you serious? If you ever want to discredit an entire article right off the bat, write something idiotic like that!

I guess hard work and dedication has nothing to do with it then?

For petes sake....

Avatar
#4 Cam Charron
October 10 2013, 05:20PM
Trash it!
7
trashes
+1
8
props
Mac daddy wrote:

"No, what wins you the Cup is good old fashioned luck."

Lmao! Are you serious? If you ever want to discredit an entire article right off the bat, write something idiotic like that!

I guess hard work and dedication has nothing to do with it then?

For petes sake....

I often wonder what's going through commenters minds when they see something they don't like in the thesis, then scroll down through all those other words before hitting the comment box.

Avatar
#6 Joel
October 10 2013, 05:55PM
Trash it!
0
trashes
+1
7
props
Mac daddy wrote:

"No, what wins you the Cup is good old fashioned luck."

Lmao! Are you serious? If you ever want to discredit an entire article right off the bat, write something idiotic like that!

I guess hard work and dedication has nothing to do with it then?

For petes sake....

Well, sure you have to have "hard work and dedication" to win the Cup... because you need them to get into the playoffs, let alone win 3 rounds to be a position to win the Cup.

But that is around as useful as saying you need 18 skaters to win a game. Or that 100% of teams that win the Cup have a head coach.

What makes you think that one team is dedicated or hardworking, and the other team that just slogged through 6 weeks of playoff hockey is just a bunch of lazy slobs half-assing their way through the playoffs to get to the finals?

So it isn't a factor that determines the winner.

It does make for a nice narrative or sound bite though.

Avatar
#7 Unknown Comic
October 10 2013, 04:15PM
Trash it!
2
trashes
+1
6
props

Wait? Luck plays a big factor? But I want to be mad at Gillis for that small sample playoff defeat last year.

Avatar
#8 Local
October 10 2013, 03:14PM
Trash it!
0
trashes
+1
5
props

"As for what PDO actually stands for, nobody really knows, but I think it's pretty darn obvious"

"pretty darn obvious"

"p...d...o"

I hope this was purposeful. Anyways, fantastic work.

Avatar
#9 Joel
October 10 2013, 05:58PM
Trash it!
1
trashes
+1
5
props

@jb

And luck becomes even more of a factor once you get into the playoffs, and then increases in importance as more teams are eliminated and the potential skill differences between teams diminish.

Avatar
#10 Unknown Comic
October 10 2013, 04:18PM
Trash it!
3
trashes
+1
4
props
Mac daddy wrote:

"No, what wins you the Cup is good old fashioned luck."

Lmao! Are you serious? If you ever want to discredit an entire article right off the bat, write something idiotic like that!

I guess hard work and dedication has nothing to do with it then?

For petes sake....

Hard work and dedication?!?!?!

Is that you Mr Healy?

Avatar
#11 mattyc
October 10 2013, 03:35PM
Trash it!
1
trashes
+1
3
props

Awesome article.

For what it's worth - IMO the results of scoring chances vs. Fenwick are entirely predictable because of a flaw in the definition of 'scoring chance'. What's being used now is really just subsetting close range shots. How often are in-close shots flubbed or bouncing, or just haphazardly golf-swingging at a lose puck. Not necessarily a better 'scoring chance' than a point shot one-timer.

Edit: I should add that I don't have a better definition of a 'scoring chance'

Avatar
#12 Cole
October 10 2013, 03:46PM
Trash it!
3
trashes
+1
3
props

"Maybe he should stop worrying about face off percentage, hits and blocks shots and just watch the games."

More like "Maybe he should stop worrying about face off percentage, hits and blocks shots and just use our stats instead."

Avatar
#13 Kent Wilson
October 11 2013, 11:17AM
Trash it!
3
trashes
+1
3
props

@Billyb0y

It's amazing how folks like yourself completely miss the forest for the trees. If you would read and comprehend the message very clearly delivered in this article, you'd understand how "the Leafs could get destroyed in corsi in one game and still win 4-0".

Avatar
#14 Kent Wilson
October 11 2013, 11:19AM
Trash it!
0
trashes
+1
3
props

@mattyc

The scoring chance area was based on the NHL heat maps that show where the highest percentage shots tend to come from on the ice.

Personally, I don't count shots that are "flubbed" from in close as a chance. It has to be meaningfully directed towards the net and it can't be blocked close to the source of the shot.

Avatar
#15 Jeff
October 10 2013, 02:39PM
Trash it!
3
trashes
+1
2
props

Good ole' Glen Healy. Mainstream media drive me nuts, especially the CBC.

Avatar
#16 Ruprecht
October 10 2013, 03:26PM
Trash it!
1
trashes
+1
2
props

Thanks for this. Could have really used a piece like this a little while ago to save a lot of time when I was wading into the shaded portions of the circles. Still having the answers here helps. Would definitely recommend this to anybody that was looking to educate themselves on some of the deeper stats that provide deeper (beyond Healy) analysis.

Avatar
#17 Peachy
October 10 2013, 04:22PM
Trash it!
2
trashes
+1
2
props
Unknown Comic wrote:

Wait? Luck plays a big factor? But I want to be mad at Gillis for that small sample playoff defeat last year.

You still can be. Everything that happens to the Canucks, luck included, is GMMG's responsibility. Bad luck is only more evidence of his unsuitability for the position. Fire Gillis.

Avatar
#18 jb
October 10 2013, 05:49PM
Trash it!
1
trashes
+1
2
props

Saying that "luck is the biggest factor" is true based on the sample set.

When 2 NHL teams play each other, they usually pretty evenly matched, so luck dominates over the relatively small skill difference.

If an NHL team played an AHL team, skill would definitely dominate over luck.

Avatar
#19 SkinnyFish
October 10 2013, 06:03PM
Trash it!
4
trashes
+1
2
props

I want to have your babies.

Avatar
#20 TMS71
October 14 2013, 12:14PM
Trash it!
1
trashes
+1
2
props

@Billyb0y

The article explained quite clearly why Corsi and Fenwick are better than +/-. Maybe you missed it. I suggest you go back and reread the post.

Avatar
#21 Cale
October 10 2013, 02:02PM
Trash it!
0
trashes
+1
1
props

If you change your green text (links & headings) from #006240 to #00865C It'd be *way* easier to pick out links. See? http://i.imgur.com/dwGPucI.png

#justsayin

Great post, though. Far more content than we usually get from GC.

Avatar
#22 wes
October 10 2013, 02:15PM
Trash it!
2
trashes
+1
1
props

So much sense. Thank you

Avatar
#24 Joel
October 10 2013, 05:59PM
Trash it!
0
trashes
+1
1
props
Mironovs Nose wrote:

Clearly given any two teams, one team has worked harder and is more dedicated than the other, and that team always wins. Unless the other team just wants it more, in which case the team with the most intangibles wins.

*We measure intangibles by wins, right?

Avatar
#25 NM00
October 10 2013, 06:46PM
Trash it!
1
trashes
+1
1
props
Peachy wrote:

You still can be. Everything that happens to the Canucks, luck included, is GMMG's responsibility. Bad luck is only more evidence of his unsuitability for the position. Fire Gillis.

Gillis shouldn't be fired until 'his' winning percentage slips dramatically.

Until then he is the best GM in Canucks history and one of the best active GMs in the NHL...

Avatar
#26 Cale
October 10 2013, 11:03PM
Trash it!
0
trashes
+1
1
props
Cam Charron wrote:

I often wonder what's going through commenters minds when they see something they don't like in the thesis, then scroll down through all those other words before hitting the comment box.

Not much.

Avatar
#27 Billyb0y
October 11 2013, 10:47AM
Trash it!
4
trashes
+1
1
props

All this time spent. Imagine the possibilities if you started writing about relevant things instead of why your stats are better then the other stats.

Writers at "theleafsnation" are baffled on how Toronto got destroyed in Corsi numbers yet won the game 4-0. Proving that stats, advanced or not, are generally useless on most nights.

Avatar
#28 Billyb0y
October 11 2013, 01:23PM
Trash it!
6
trashes
+1
1
props
Kent Wilson wrote:

It's amazing how folks like yourself completely miss the forest for the trees. If you would read and comprehend the message very clearly delivered in this article, you'd understand how "the Leafs could get destroyed in corsi in one game and still win 4-0".

The point is that your "advanced stats" are about as useful as plus/minus.

No more. No less.

Avatar
#30 mattyc
October 10 2013, 04:27PM
Trash it!
0
trashes
+1
0
props

@Mac daddy

as my man Ludacris once said

"Dedication, hard work, prayers and belief And a good team {*ding ding*}" - Undisputed

Avatar
#31 Cale
October 10 2013, 11:18PM
Trash it!
1
trashes
+1
0
props
Graphic Comments wrote:

I agree! Unfortunately, the colours are built in to the god awful editor we use here. I supposed I could go through and put the colour code into each url tag but...

And, um, I'm not really sure how to take that last sentence. :P

You'd have to ask whoever handles the site's maintenance to change the css. And I only meant that this post was meatier than most of yours, which trend to the lighter and laughier side of the spectrum.

Avatar
#32 BCLeafFan
October 11 2013, 08:16AM
Trash it!
0
trashes
+1
0
props

Good read. Made me think about a scout's comment about an NBA player - this kid has everything, size, quickness, skills, everything - everything but a clue. I find myself thinking about that comment more and more as I hear again about advanced statistics. Oh, and by the way, how do you factor in hot goaltending statistically? Nice work on this.

Avatar
#33 Prairie Chicken by-the-Sea
October 11 2013, 12:56PM
Trash it!
1
trashes
+1
0
props

@Kent Wilson

Great analysis. Except for Sean Monanhan. He'll get 100 points for sure.

Avatar
#34 mattyc
October 11 2013, 04:38PM
Trash it!
0
trashes
+1
0
props
Kent Wilson wrote:

The scoring chance area was based on the NHL heat maps that show where the highest percentage shots tend to come from on the ice.

Personally, I don't count shots that are "flubbed" from in close as a chance. It has to be meaningfully directed towards the net and it can't be blocked close to the source of the shot.

My issue isn't really the methodology per se - I think you guys are doing commendable work. More the point is that I take that Flames - Habs game the other night. The Subban goal from the point would have been a great scoring chance, even if MacDonald had somehow managed to save it. Contrast that with some 'garbage goal' in close. Intuitively we can see that one requires more skill (to get open like Subban or Ovechkin or Stamkos only can, and also to make the pass to that open man), but that skill - which would undoubtedly contribute to 'shot quality' doesn't bear out because the data doesn't capture a distinction between 'skill goal/chance' and 'luck/bounces/bad goaltending goal/chance'.

Comments are closed for this article.