How Screwed are the Flames?

[JATOSS]

Ryan Lambert is no good.

[CASe333]

I love advanced stats and think they have their place. That said I think some people forget their relative importance.

As of tonight Calgary is +20 goal differential through 65 games ranking them 4th in the conference. To me 65 games doesn't seem like a small sample size that could be based on luck.

Honestly I don't know how anyone could possibly argue that a good Corsi is more important then goal differential in the long run. It just seems asinine at this point to not look at the big picture.

[GranteedEV]

[CroFlames]

Originally Posted by GranteedEV

Exceptionally good look suit today I must say. Very sharp.

[TorqueDog]

Next time, can we please use http://www.donotlink.com/ when posting such tripe?

[edslunch]

Originally Posted by AC

http://forum.calgarypuck.com/member.php?u=2865

Best part of that profile:

ryan_lambert has not made any friends yet

[Red Menace]

Are not all of these advanced stats predicated on the assumption that whoever has the most shots in a game is the better team?
I've never heard an argument on quality of shots, which is not really being measured yet.
I like advanced stats a lot actually, but they are in their infant stage in hockey and have a long way to go. it would help their cause if they admitted that the current state of hockey analytics is very imperfect, and did so without getting super aggressive against anyone or any team who doesn't fit their world view.

[Caged Great]

If you could track players like they did during the all star weekend, you would be better able to determine who is responsible for good plays/bad plays and who does things efficiently.

Shots for/against are not equal and it is too simplistic.

Calgary also doesn't play like any other team, where most of the offense is driven by the defensemen rather than the forward group. Because of the excellent quality of the passing by the defenders, it catches the other teams off guard because the Flames do play differently. That leads to more goals simply because of the quickness of the swapping from defense to offense. There is no metric for that.

Other teams that had poor corsi/fenwick numbers that collapsed were getting otherworldly goaltending. Calgary is not. Steady NHL average goaltending only from both Ramo and Hiller. The Avs lost their 2nd line C (Stastny) and replaced him with slower players (Iginla/Briere). Add that to a step back from Varlamov and it's not surprising that they are playing poorly. Carey Price though is continuing to be the best goalie in the NHL and the Canadiens are riding his coattails despite not being a very good team beyond him.

[bigehn]

Are these Flames for Real?
Full Disclaimer:
1. I am a huge Hockey fan and a huge Flames fan.
2. I am a believer in the usefulness of advance stats and I work in the analytics industry. I believe in the power of information.

Through much of the 2014-15 NHL season the Calgary Flames have been the little engine that could. At times looking outmatched and overwhelmed but, through determination, hard work and a strong belief in each other they continue to find a way to win. Along the way the advanced stats, in particular possession metrics, point to a team that is much worse than the win total beside its name in the standings.

The difference leads to some common reactions.
1. Advanced stats enthusiast point out the Flames are a bottom tier possession team fueled by an unsustainable shooting percentage. Once they regress to the NHL mean, which in the case of shooting percentage is between 7 and 8%, the bottom will fall out and the flames will fall to their rightful place in the NHL standings.
2. Advanced stats opponents point to the Flames as an example of why advanced stats are useless.
3. While others acknowledge the Flames must work to improve their puck possession if the Flames want to become a perennial Stanley Cup contender but, they are willing to sit back and enjoy the fun filled ride.

While some in the advanced stats community hold their collective breath for the Flames to fall, I prefer to try to understand what is making the Flames an exception.

In Hockey Abstract 2014 Tom Awad has some very interesting analysis that I think is very useful in understanding the success of the Calgary Flames. The first interesting point he makes is goal differential is driven mostly by possession differential and shooting percentage and the split is about 50/50. The second interesting point is players can be grouped into 4 tiers in the NHL based on ice time, and there is a significant difference in shooting percentage between the tiers. What this means is not every player will regress to the NHL mean for shooting percentage. Some players can just shoot the puck better than others, and it is much more likely that an NHL player will regress to his historic shooting percentage.

What this means for the Calgary Flames is the team shooting percentage should regress to the collection of historic shooting percentages of all its players and not the league mean.

From the table below we see the total team shooting percentage differential between the 2014-15 season and the career shooting percentage is an increase by 0.88%. If you multiply the individual shots by the individual shooting percentage differential you get the individual goal differential for each player. When you sum this for the team it results in the Calgary Flames scoring just over 7 goals more this year than the expected number of goals based on their historic shooting percentages. While this would most likely result in a few less wins the difference is not enough to completely remove them from playoff contention.

[DJones]

Tracking corsi but not taking into account shot quality is the equivalent of keeping track of pitches but not taking into account the speed or spin of the pitches.

Just a science in its infancy

[DJones]

And your adjusting goals by the individuals historic mean still completely ignores the fact that the mere act of a goal disappearing would alter the rest of the game. If one of those shots didn't go in the Flames would be pressing harder rather than playing with a lead.

It's not baseball and until advanced stats makes a gigantic leap forward they will be hard pressed to ever be as effective as moderately educated eye test.

[topfiverecords]

No surprise. 0 Friends. How's that for an advanced stat?

[Blarg]

Looking at our current road trip record, I can safely say the Flames are so screwed

[Jay Random]

Originally Posted by bigehn

What this means for the Calgary Flames is the team shooting percentage should regress to the collection of historic shooting percentages of all its players and not the league mean.

Apologies in advance for encyclopaedia—

It's also worth considering that several of these players are rookies or sophomores, so that they really haven't established a very firm baseline (or any at all, in some cases) for their historical shooting percentages. Johnny Gaudreau, for instance, was basically an unknown quantity coming into this season, since a previous career total of 1 G and 1 SOG really doesn't give any useful information. Several of the team's best shooters are also among the youngest and least experienced players. So when people expect a regression (say) to the form the Flames showed in past years, they are overlooking a very large infusion of new talent.

Is Gaudreau the kind of player who can keep up a shooting percentage of 12.7? Quite likely. He's got superior offensive skills, and (what matters just as much) the habit of looking for a pass rather than taking a low-percentage shot. As his teammates adjust by learning to pass the puck to him and let him quarterback the play more often, I expect his offensive numbers to improve.

I'd like to see stats for ‘disposals’, as they call them in Aussie rules football (thanks, Imported_Aussie). A team that blasts the puck at the net from anywhere and everywhere, and generally loses possession as a result, will have better Fenwick and Corsi numbers than a team that frequently passes the puck looking for a better shot, even though both teams may have possession of the puck for identical lengths of time.

We know that shots vary widely in quality. According to one summary I've read (link not handy, sorry), a shot on which the goalie is square to the shooter has an expected save percentage of about .950, whereas if the shooter receives the puck and shoots before the goalie can change position, the expected save percentage can be as low as .650. And that is not even taking into account the differences in skill between different shooters!

Logically, a team with sufficient offensive skill and creativity should be able to compensate for taking fewer shots, up to a point, by moving the puck rapidly, finding skilled shooters in good scoring positions, and taking higher-quality shots. And in fact (I have read) the numbers show a slight negative correlation between SOG and winning percentage, even after accounting for score effects, so one suspects that this is not just logical, but actually the case.

One thing that ought to be obvious is that the more offensively skilled players a team has on the ice together, the better their average quality of shots should be. It's simple geometry: it's harder to defend against five angles of attack than three.

Usually a team's wingers are the weakest defensive players on the ice; but that doesn't matter so much, because the other team's defencemen are the weakest offensive players. When all five skaters are offensively dangerous, there will be opportunities to capitalize on any weakness shown by the opposing forwards – as the Flames have often done.

With that in mind, it should be no surprise that the Flames have good offensive production when their top four defencemen are playing, and look pretty awful when the bottom pairing is on. Those top four are all in the top 60 of scoring among NHL defencemen (as of Friday afternoon), and three of them are in the top 20.

If the stats folk started making a serious effort to measure ‘disposals’, which I suppose would include passes, recovered shoot-ins, and shots on goal where the attacking team regains possession of the puck, I imagine that would be a better proxy for possession than either Corsi or Fenwick.

In one respect, it might actually be better than a direct measure of time of possession, because it would measure velocity of possession – the number of discrete plays that a team makes while keeping control of the puck. Time spent carrying the puck doesn't do much good if you just hang onto it and never make a play.

At any rate, such a stat ought to be worth looking into.

I remember that Dave King, when he coached the Flames, kept track of the number of passes made by the team, and set a hard target of so many passes per period; and the offence improved significantly when the players made an effort to meet those expectations. A pass never shows up on either Corsi or Fenwick, except by accident; but it can certainly show up on the scoreboard, where it counts.

[Freeway]

If talent is evenly distributed, PDO predicts that over an INFINITE number of games, every team will have 100 Corsi (e.g., shooting/save percentages will regress/progress to that point). There is no consensus over what a suitable sample size is in order to say somebody was "right," nor is there a consensus over whether NHL talent is evenly distributed (it probably isn't).

[Jay Random]

Originally Posted by Freeway

nor is there a consensus over whether NHL talent is evenly distributed (it probably isn't).

Uh, I would say there should be a pretty easy consensus that NHL talent is not evenly distributed.

[EldrickOnIce]

That's the argument I love.
If the Flames went to the Stanley Cup finals and lost in OT in the 7th game, or whether they miss the playoffs - the same see we told you would result.
In fact, if the Flames won the Cup this year and faltered next, I expect these guru's would say see we told you so.

That's the perfect part of predicting someone will lose... You are eventually right 100% of the time.
Do something interesting, and predict who will win instead?

[Phil Russell]

"Statistics don't bleed" is a saying used to describe the failings of statistics-obsession in the human sciences. You can't model something as complex as human behavior or, in this case, a hockey game. Statistics don't bleed, but the Flames do and that's why they win. Sacrifice, team-work, opportunism, confidence etc. Where do these things fit in the model? And if they don't fit, then you are assuming they have no impact on winning or losing a hockey game. Which is, of course, ridiculous.

[ToraToraTora]

This guy is essentially a troll. Read his twitter, he trills trolls right back. Real journalists or professionals dont respo.d to criticism like he does.

[Resolute 14]

Originally Posted by DJones

Tracking corsi but not taking into account shot quality is the equivalent of keeping track of pitches but not taking into account the speed or spin of the pitches.

Just a science in its infancy

Yiup. Though Corsi/SAT and shot quality are two different things. The problem isn't the stats, but the level of faith placed in them. In general terms, teams with better possession numbers do better. This is rather obvious as you can't score without the puck and you can't be scored upon when you do have it (unless you've pulled your goalie on a delayed penalty call...)

Low Corsi is largely overcome in two ways: Great goaltending or great offence. The Flames have actually slid a bit in terms of offence, but are still fifth in terms of ESSH%. Whether that is sustainable or not speaks to shot quality. And since that isn't easily measured yet, it gets written off as "luck".

It is easier to continually buck the Corsi trend with a great goaltender (Carey Price), so the question of long-term sustainability for the Flames is relevant. As our team develops and gets better, that should matter less and less. Also, if players like Sean Monahan demonstrate that a 15%+ shooting percentage is not unrealistic, that pulls up our sustainability.

But no matter what happens in our final 17 games, the fancystat guys should realize that they were wrong about the 2014-15 Flames. But as last year's Avs demonstrate, they won't because of their emotional need to be right. Just as the goalposts were moved to point to a first round playoff loss as validating their argument that Colorado never should have been as good as they were last year, they would point to the Flames missing the playoffs as validation now. Except that the singular stat they use with godlike reverence says we should have been as bad as Toronto and Colorado are this year.

Whether or not we make the playoffs, the Flames have proven themselves to be one of this year's significant Corsi exception teams. What we do next year will be an entirely new story.