Random Thoughts: Reading tea leaves

random thoughts

The Flames’ rough January has fans brandishing pitchforks and leaping off the bandwagon in droves. 

After going 4-9-0 and suffering humiliating losses to the Oilers, Leafs and Habs, it seems self-evident that Calgary’s season is circling the drain. But despite the bad record and gut-punch losses, January actually contained a solid kernel of hope and optimism. 

Seems impossible, I know, but stay with me.

– As noted in my Tweet above, Not everything went badly in January. In fact, in some ways this month was the Flames’ best of the season, even though it didn’t show up in the W-L record. Whaaat? It’s true. Check out this chart:

Corsi   Shots   Expected Goals   Goals   Percentages  
CF 567 SF 316 XGF 26.01 GF 15 SH% 4.75%
CA 458 SA 237 XGA 23.38 GA 23 SV% 90.30%
Cdiff 109 Sdiff 79 XGDiff 2.63 GFDiff -8 PDO 95.04
CF% 55.32% SF% 57.14% XG% 52.66% GF% 39.47%  

Lots of stuff here, but it’s pretty simple. From left to right I have the Flames’ total shot attempts (corsi), shots on net, expected goals, actual goals and shooting/save percentage at even strength (5on5) for Janaury. 

As you can see, there is one key outlier: despite owning the majority of shot attempts, shots and expected goals, the Flames managed less than 40% of the actual goals at 5on5. How is that possible?

Take a look at the last column. Calgary’s percentages over the last 13 games were putrid – particularly their shooting percentage of just 4.75%. To put that number in perspective, Deryk Engelland’s career SH% is 4.6. So it was like the Flames’ entire roster transformed into Engelland for a month. It’s remarkable they won any games at all. 

– “Yeah but that low shooting percentage is reflective of the Flames’ inability to create scoring chances,” you might say. It may seem like the club spent the whole month just firing low percentage shots into the opposing goalie’s crest, but that’s more perception than reality. 

That’s why I included both total shots and Corsica Hockey’s expected goals stat. That takes into account shot factors like distance to the net, shot angle, whether it was a rebound, off the rush, etc. According to that model, the Flames should have scored 26 even strength goals this month (for a completely average 8.22 SH%). Instead, they managed just 15.

– For long time readers, the Flames PDO of just 95.04 is instructive. We know from years of work that PDO tends to regress to the league average of 100 over time. In other words, no NHL team is so bad as to shoot less than 5% over time or manage a 95 PDO forever.

– For further context, Calgary’s even strength shooting percentage from October 2014 to now (Jan. 31, 2017) is 8.04%. That’s more indicative of their shooting ability than a single, isolated 13-game sample. We should expect them to move back up to that finishing rate. 

– If you’re looking for a more lasting criticism of the club’s January, it’s on the defensive side of the puck: Calgary gave up 23 goals against at ES and their expected GA was actually 23.38. That means the Flames were surrendering some pretty high-quality chances against this past month. 

Even when generating 79 more shots at ES than the bad guys, they only had three more expected goals for in aggregate. That’s much better than what actually happened and we should expect improvement as a matter of course, but the club also needs to tighten up. 

Of course, a few more timely saves from the guys between the pipes would go a long way, too. 

– “The Flames spent the whole month trailing which would inflate all of their shot metrics.” You’re correct! But not so much that it invalidates the conclusion they were roughed by lady luck. If we use Corsica Hockey’s venue and score adjustments, Calgary’s CF%, SF% and XGF% each drop to 53.12%, 55.1%, and 50.42%. Less impressive, but still pretty healthy. 

– If none of this analysis seems satisfactory compared to explanations like a lack of leadership, heart or toughness, well…that’s because we are all pre-wired to jump to those kinds of conclusions. Certain narratives feel far more intuitive than “randomness”.

– The big bias here is one I’ve written about frequently in the past called the Fundamental Attribution Error (or correspondence bias). This is the tendency to overweight personal, internal factors when explaining other people’s results or behaviour. That means we also tend to underweight the influence of external factors. Of course, most people do the opposite when explaining their own behaviour. A related bias is the Just-world hypothesis. This heuristic means we assume that people are generally deserving of their outcomes. 

– Put those things together and you have the natural inclination to make up stories that stress personal, internal failings that (deservingly) lead to bad outcomes. Bad record this month? That must mean the team doesn’t care, has a fractured dressing room, lacks internal fortitude or just isn’t tough enough, right?

– Of course, if that stuff was true, we would have also seen the Flames’ shot metrics take a dive down the toilet, rather than just their percentages. But because the bad results are mostly isolated to SH%, we can safely conclude natural variance is the primary suspect for Calgary’s terrible January. 

– tl;dr – The Flames outshot the bad guys to a nontrivial degree in January and were undermined by an uncharacteristically low shooting percentage. If they continue to put way more pucks on net than their opponents, we can expect their SH% to regress back to the league average, meaning they should win a lot more often moving forward.   

  • Derzie

    My expectations for the Flames is the same as it was in October: Above average goaltending, majority of players performing (points) at or above history, competing for a non-wildcard playoff spot. All 3 are getting a failing grade right now.

  • BobB

    Is this just coincidence or is there some kind of trend/correlation with less elite teams? Score effects? Voodoo? We, the Regehr-like Flames plastered the Hemsky-esque Oilers one too many times and their stink rubbed off on us? I remembered reading these back in the day, so I spent far too long finding them again:

    “Under Eakins so far this season, the Oilers’ puck-possession numbers were trending upward; they control 50.9 percent of all shot attempts at 5-on-5, which is 14th in the league and a solid indicator for future success. A .902 team even-strength save percentage (29th in the league) rendered that moot, as did a 6.5 shooting percentage (28th).”

    and

    “The Oilers shot in bad “luck” under Eakins, connecting on just 7.4% of their shots in all situations. The opposition were much more successful at 11.5%, a conversion rate > 1.5 times better than the Oil…If in your belief system PDO is an expression of “luck”, then Dallas Eakins was one unlucky coach.”

    • Greg

      This is my question too… are the flames actually improving? Or is GG just the second coming of Dallas “if you throw the puck at the net from the neutral zone we’ll improve our corsi and wins will follow” Eakins?

      Question for @Kent I guess 😉

      • Eakins went all in on analytics. The GM in Phoenix is now repeating the cycle again with little to no success. Corsi , blah blah, Fenwick blah blah, PDO blah blah. All of these can be used in some fashion as a guideline, but anyone going all in and designing their roster around these stats has seen nothing but failure. GG is Eakins 2.0, the sooner you get rid of him , the better for your franchise.

        • SmellOfVictory

          Toronto went all in on analytics and they’re doing really poorl-… Oh. Also, funny that you’re using the example of the GM in Arizona (note: they’re not called the Phoenix Coyotes anymore) who is literally in his first year of restarting the rebuild that was unsuccessfully implemented by the previous GM.

          • jakethesnail

            Didn’t the Leafs fire their analytics guy? Now they are doing much better!

            Didn’t the Habs fire their analytics guy culminating over his objection to the PK Subban trade? Now they are doing much better!

            Florida is going into the deep end analytics…they are a mess now!

          • SmellOfVictory

            No, the Leafs still have Kyle Dubas as an AGM. Edmonton fired their analytics guy, and Montreal fired theirs. Montreal has had a healthy Price this season, and Radulov has been very good. Weber for Subban was effectively a wash.

          • SmellOfVictory

            Care to provide any evidence for your incredibly vague retort? The slight edge in production that Weber currently has? Or the fact that he’s playing in front of the best goalie in the NHL? I’m sure you’ve watched both of them extensively this year.

          • Randaman

            The fact that Carey Price seems much more settled and Weber has been a rock. Subban has been out for 1/3 of the season and still is a give away machine and yes I have watched both extensively.

  • RedMan

    it is so reassuring to know the Flames are “visually better” and that they are actually winning games even though the league has yet to start to recognize this and stick with the outdated method of counting goals for and against. 😀

    but in all seriousness, joking aside, one needs to realize that this team is on track with the rebuild, and it just hurts to see a team like the Oilers beat them so embarrassingly.

    • Deef

      Agreed. We’re still in rebuild mode. Though I wish our young guys showed some more improvement this year over last year. The seem to have gone a wee bit stagnant.

  • Ogie Oglethorp

    This article is case and point on why those skeptical of advanced stats willingly throw the baby out with the bath water.

    I’m so tired of being told January was good. Several articles this month have told us silly game watching fans that the team is doing GREAT!

    I watched the games. All of them. They were NOT good. Embarrassing, pathetic, frustrating – yes. But not ‘optimistic’. GG ripping into the boys in a post game presser. Not exactly the skittles and rainbows the stats guys are showering us with.

    The PDO argument is perhaps the most insane of them all. Sure it gets close to 1. But looking at last years, all the terrible teams had PDO below 1. The best teams had PDO above 1. Fact. So while .95 may be an anomaly its probably not that far off if your team is getting slapped around the ice, can’t score and has horrendous goaltending, which you can see from watching the games – no spreadsheet required.

    Our team PDO is .98 right now for the season. About right where it will end for all bad teams. THe fact that it was .95 in January is utterly useless information.

      • SmellOfVictory

        How could a measurement be considered a “fallacy”? It’s just a combination of save percentage and shooting percentage. If you’re talking about the expectation of PDO regressing toward 1.00 (or 100%), it’s pretty straightforward: league-wide, every shot on net either enters the net (sh%) or is stopped (sv%), therefore league-wide, PDO must inherently be 1.00. It’s basic mathematics.

        Applied to individual teams, it’s recognized that PDO may vary based on team skill level, but parity is close enough, and the historical data have shown, that no team can successfully replicate a PDO above or below a certain percentage in successive seasons (I believe the upper and lower bounds are something around 0.985-1.015).

        To be a post hoc fallacy, regression toward the mean would have to imply that PDO actually drives anything, which is not the case. It’s just a measure of what’s occurring. It doesn’t happen before or after anything else. If your argument is that the PDO upper/lower bounds are based on inductive reasoning rather than deductive, then you’re right. That’s what the vast majority of factual (certainly statistical) information is based off of.

        • Ogie Oglethorp

          Your accurate explanation is EXACTLY why PDO is so dumb to ever talk about at any point of the season. ESPECIALLY when evaluating the month of January. Its completely irrelevant, utterly useless, and 100% pointless to talk about. Period.

          PDO is the EXACT same as the gamblers fallacy (wiki). Exact. So, I’m unclear on why it comes up in hockey discussions all the time.

          Flip a coin 100 times and it will probably be 50/50. But if you get 45/55 and then suddenly start saying things are gonna swing your way, then people would scoff at you.

          But this is exactly what this article does. Well, you see… we had bad luck in January, so things are bound to turn around! It’s ridiculous logic. Any good friend would drag their buddy out of the casino before he bets big on his luck turning. But somehow it’s logic, and PDO goodness to talk about in sports. Ridiculous.

          Especially when you watch the games, and its a total tire fire, Johnny visibly frustrated, goalie letting in soft goals and the coach publicly ripping on the team.

          But, hey, maybe we should put down a HUGE BET now. Because our luck is bound to turn! (sarcasm)

          • GodsGotSandals

            I don’t know if being unlucky now means your gonna be lucky later but at least one would expect to get back to league average in shooting percentage.

          • Kevin R

            If you are shooting perimeter shots 75% of the time & running low shooting percentages, you either get higher quality shots or expect your shooting percentage to stay the same. Most NHL goalies don’t let too many perimeter shots in, our goalies included. In 2014, I know we were the text book, the stats say we can’t maintain the success. Thing is, Hartley had this team playing a system where we had to block a lot of shots & spend a lot of time hemmed in our zone but when we got our chances, they were 10 bell chances. Other teams studied us in 2015 & shut us down but Hartley never adjusted, probably as we are seeing now, we don’t have the players to adjust to that side of the game.I would be curious to compare the number of high danger shots the Flames had in 2014 to this year.

            These stats are all analytical. They tell what is happening or happened in the games being played. Simple as that. At no point have I seen evidence that these stats can predict success. I see stats as being valuable for coaches working with players & for potentially developing players on the farm. They could also be used to profile prospects to assist in drafting decisions. But lets face it, maintaining peak performance is next to impossible for the majority of professional hockey player (with exception of elite players) Finding players that consistently bring their performance & play to their talents is the white unicorn all NHL teams search for.

          • McRib

            “At no point have I seen evidence that these stats can predict success.”

            Every team to win the Stanley Cup since 2011 has been a Top. 5 analytical team in the league (Vancouver was a Top. 5 analytical team in 2011 and deserved to win outside of Tim Thomas standing on his head, Boston also had a much easier road through a weak east). Advanced Stats also predicted two very unlikely wins by LA over that span and a Pittsburgh win last year. The data I have only goes back to 2007, but all those other winners after 2011 were also Top. 5 analytical teams. So essentially in the last decade only one team has gone on to win a Stanley Cup that wasn’t a Top. 5 analytical team, looking at the data closer six of the teams to win the Stanley Cup the last decade were Top. 3 analytical teams or higher. When Boston won the Stanely Cup they were around the 10-11th best analytical team in the league, so it wasn’t like they were awful according to advanced stats and anyone who knows anything who watched that series against Vancouver believes Boston would never have won except for Tim Thomas being out of this world.

            People can point to Florida and Arizona currently striving to be top analytical successes, but they aren’t right now so using them as a reason why “analytics means nothing” is rather nonsensical. The absolute best thing about all of this, is analytics predicted that the Flames would fall off a cliff last season after unsustainably making the playoffs the year prior and they have really predicted us to a “T” this year as well. We are a middle of the pack club that has a shot at making the playoffs and avanced statistics is predicting that. Refute advanced stats all you want, but it has basically predicted who is worthy of winning the Stanley Cup since it’s inception (Top. 5 analytical teams). There may be an outlier or two in the last couple decades like Boston, but Boston was still a positive posssesion team. Boston also had a possession beast in Patrice Bergeron. Analytics are here to stay whether people want to deny them or not (I tried for years, but the data was far too overwhelming to continue doing so).

            Florida and Arizona using analytics to try and get a couple of value signing this past offseason doesn’t mean they are top analytic teams. Who is the best player in the world? Sidney Crosby! Who is the best analytical player in the world? Sidney Crosby!

          • Kevin R

            Again, you are only evaluating past results of teams here & missing my point. Maybe we should be having this discussion in September. I want to see the analytics predict results based on their data of players that make up the team (based on their historical stats) & provide a statistical explanation for the predictions. Then lets revisit it at the end of the season.

            I like analytics & I think there is a value that every NHL team should be utilizing in how it runs it’s organization. I guess where I get my back up is when players like Brouwer get continually slammed & written off as being anchors & useless NHL players. Guess what, lets look at the successes Mr Brouwer has had in the NHL playoffs as compared to the NHL analytic darlings like Frolik & Backlund. You will find that performance by most NHL players is inconsistent year to year for many reasons & players all have a role to play within their teams. But I’m sure you know that already.

            Another thing that gets my back up is when results don’t go in accordance to the analytics, PDO is thrown as the luck factor as an excuse for the lack of predictability.

            As for predicting the Flames to fall off the cliff after 2014. Was that PDO or was that opposing teams studied what the Flames did & watched the films on the few key players & were able to shut down the Flames because they were more prepared. Leafs are kind of getting that youth bump this year. Once expectations raise, the Leafs & their talented rookies will find teams better prepared to defend against them.

            For the record, I am all for analytics. I just don’t think they are being applied properly, given the infancy, is totally reasonable.

          • Ogie Oglethorp

            I just did some math. We played 12 times in January. A PDO of .95 means we got 11.45 magical PDO points out of a possible 12. (so dumb but bear with me)

            6 games we were over 1. 6 games we were under. 2 games we got buried. 2-4 loss to Vancouver (PDO .74) , and 5-1 thumping to Montreal (PDO .78)

            Remove those 2 games and PDO = 0.99

            The rest of the games:

            1.06 (WIN), 1.03 (LOSS), 1.03 (WIN), 0.98 (LOSS), 1.0 (LOSS), 1.08 (WIN), 0.90 (LOSS), 0.96 (LOSS), 0.86 (LOSS), 0.78 (LOSS), 1.04 (WIN)

            So dumb. PDO is worse than +/-. The dumbest most useless stat in hockey. Period.

          • cjc

            Well, I suppose it depends on whether one assumes the coin is fair. If it is, then a person could justifiably bet that the NEXT 100 flips would result in a ratio of 50/50 (expecting 55/45, of course, would be falling prey to the gambler’s fallacy).

            Kent wasn’t suggesting that the team’s PDO would swing round to 105 in February because it was 95 in January. Only that teams have never sustained a PDO that low over the long term. It’s not that Calgary’s luck is bound to turn, only that it should return to normal.

            Of course one could also conclude that Calgary really IS a 95 PDO team and do something rash as a result.

            Kent also mentioned the team was lax defensively, giving up a lot of high quality scoring chances. He’s not saying things are perfect, only that they are not as bad as they may seem.

        • kittensandcookies

          Of course PDO doesn’t drive anything. That’s exactly what I’m saying. You’re saying that a low PDO will result in a higher PDO later, which is ridiculous. What’s even more ridiculous is the expectation that the higher PDO will happen in some expected time frame, such as “next month”, when in fact you can be unlucky *FOR YEARS*. Conversely, you can be lucky *FOR YEARS*.

          • Eggs Bennett

            Please show me a team that has been lucky *FOR YEARS* or one that has been unlucky *FOR YEARS*

            Think of PDO this way: if it is consistently less than 100, this means you are consistently shooting at a lower % than your opposition. This means that you can consistently lose games where you outshoot the opposition. Most people in this forum would agree that if you are consistently out shooting your opponent, you shouldn’t expect to keep losing (similar to you should not expect to keep winning if you’re consistently out shot by your opposition, I.e. pdo>100)

          • Ogie Oglethorp

            Of course. But you see, this article isn’t talking about PDO FOR YEARS. It’s talking about January PDO. A tiny sample size not based on *FOR YEARS*

            The stats guys love to use PDO as an indicator of why things will change. Which is a fallacy.

            Well not exactly… There is some logic there, buried beneath the BS.

            Let’s say half way through the season your team PDO is 0.95. The stats guys always say ‘wow, bad luck, don’t worry it’ll even out’. This is FALSE. If it truly was 50% through the season, then you should adjust your year end logical PDO target to 0.975. Banked bad luck is permanent. So PDO will likely be 1 in the last half, but that means it ends at 0.975

            Where people go off the rails is claiming we had bad luck in some tiny sample size, and then saying we will have good luck to even it back to 1. This is patently false.

            PDO should never be talked about. Its irrelevant, and utterly useless to discuss as any sort of predictor of future anything.

          • kittensandcookies

            You’re kidding me right?

            Calgary. Calgary’s had low or average PDO for many years, other than 2014.

            Now, maybe they’re unlucky, or maybe they’re just bad teams?

          • SmellOfVictory

            Where did I say that low PDO now results in high PDO later? That’s a pretty massive misinterpretation of what I wrote. Anyone who’s even vaguely familiar with stats knows that bad luck now doesn’t mean “balancing” good luck will come within x timeframe afterward. Luck still does regress toward the mean over time, which means that, while you won’t go from .95 PDO to 1.05, you are more likely to be higher than .95 PDO in future than not, because you’re always more likely to be closer to 1.00 than not.

            Same thing goes for Kent’s piece. Of all of the writers I follow, he’s probably most acutely aware of the various cognitive fallacies that befall people (e.g. gambler’s fallacy). Nothing about his article states or implies that the Flames will henceforth be “lucky” as a result of their bad luck; merely that luck this bad doesn’t normally last.

  • Eggs Bennett

    Geez I’m glad you guys aren’t my portfolio manager. Talk here makes me think that if the S&P 500 is low despite low p/e ratios and other strong market fundamentals that it’s going to stay low forever…

    For those who are hating on GG for calling out our players (I was one of those guys), do any of you actually know what the players actually think of him? Or is it all heresay from general sentiment on the forum and watching some post game press conferences? I actually heard on FAN960 a dressing room insider who commented on the fact that the guys love GG in the dressing room and hated Hartley. Being called out by a coach you love might be more motivating than you think…

  • Just.Visiting

    Perhaps there should be more advanced advanced stats.

    Corsi seems to indicate that this year’s team is a materially better team than last year’s.

    My eyes tell me that I’m not nearly as entertained, that I’m often bored and that the results are disappointing on many nights when Corsi said we were the better team.

    I think it would be interesting to see time of possession in the offensive zone relative to time of the other team’s possession in my defensive zone and something where there’s a comparison of shots from within the danger zone, vs something in which all shots are regarded as equal.

  • freethe flames

    Is there any plan to have a round table where you discuss what the Flames should do as the trade deadline approaches?

    Here’s my thoughts and on what we might get:
    Expiring contracts: Versteeg a 3rd, Eng’s a 3rd, Wideman and Chiasson likely a 5th each. Elliot a third and Johnson a 4th.

    Versteeg would give a team in the hunt depth for their top 9 and playoff experience. Eng’s provides depth for a team that wants a stay at home depth defender. Wideman and Chiasson both add depth to teams but further down the depth chart. Either Elliot or Johnson could give a team like Pittsburg a back up if they move MAF.

    I could also see teams looking at Stajan or Bouma as character guys. Of course if we could unload Brouwer I would be extremely pleased but I doubt anyone wants him and his remaining three years at $4.5m.

    If we could move a couple of these guys and then use these picks for a test drive on an upgrade at RW I would be pleased.

    • Primo

      1) Are you assuming the Flames who are currently in the hunt with 30 games left that they will be sellers at the trade deadline? If so reminds me of an Oilers 10 year loser model.
      2) If we sell the return for the assets that you project are underwhelming and would get any general manager immediately fired…..

      • freethe flames

        By Feb 18th we will have a much better idea of if we are still in the hunt; if not by then for sure by the actual deadline we will know. Nothing gets done in a vacum you trade those assets and other things begin to happen. I would also suggest that if you don’t get something for these assets then you are likely fired as well.

        None of the guys I listed should have a long term future with the Flames and quite frankly losing 1 or 2 of them might even help the team be better for the run at a trying to make the playoffs.

    • everton fc

      We are where we are for the following reasons:

      1. Sub-par play from Gaudreau and Bennett.

      2. Erratic goaltending.

      3. Sub-par play of Brodie. Maybe even Gio. No offence outside Hamilton, from the backend.

      4. No true first line. No true first line RW.

      Versteeg is a 3rd line wing who can play “up” the roster when needed. A good guy (like Stajan) to have around. Why does everyone want to move Stajan – he’s been one of our best players, albeit in his own, humble way? If Engelland could be re-signed for 1 year/low wage… He, too, adds value as a 5/6/7 option.

      Wideman ain’t moving. Though I hope they can move him. Chiasson isn’t hurting us. He’s a low cap hit. He’s a perfect 4th line RW (though so is Hathaway – someone’s gotta go).

      Bouma you move. We have Ferland. Brouwer is one of the most disappointing signings – he’s basically a 3rd line RW at best these days, like Versteeg he can play “up” the lineup a bit, but I think Versteeg’s simply better. Brouwer’s contract may be one of BT’s worst moves (his prior signing of GIo at big bucks “forever” is another). Hathaway may have more offensive upside potentially, than Brouwer, who incidentally hasn’t scored a point since December 19th (in fairness, Chiasson’s only scored 3 points in the same stretch, Ferland 1 point).

      I’m sort of optimistic, but I will add our farm team ain’t that deep. Jankowski, maybe Shinkaruk, maybe Klimchuk… Maybe Mangiapane… All four “maybes”… On the defensive side, Andersson looks god, but can he skate at the next level? Will Kylington ever put it all together (I think he will). Is Gillies another problem goals once he gets to Calgary (if he gets to Calgary). I am more prone to see Rittich as the better “prospect” between the pipes.

  • jakethesnail

    Next Poll Question:

    At the end of the next 4 game stretch and the 5 day break the Flames will be:

    – in a wildcard position

    – out of a wildcard position, but still within 2 points

    – out of the playoff hunt, realistically speaking