Narratives, Damned Narratives and Statistical Anomalies

Never let stats get in the way of a good story.

Talking heads on TV sports panels have never met a statistical anomaly they didn’t like. No matter how meaningless. No matter how lacking in context. No matter how irrelevant.

No. The only thing that matters is whether they contribute to a narrative, or tell a story.

Well, there they were on December 1 and the Canucks sat three points out of a playoff spot.

PANIC!!!

Apparently December 1 is a magic date because only X teams in the last Y years that were out of the playoffs at this point in time have come back to make it into the post-season by the end of the year. Insert your own favourite numbers in there for X and Y, or heck, just make them up because either way it will have just as much bearing on the final standings this year.

By the way, is it points back coming in to Dec. 1 or at the end of Dec. 1? Either way, they were still out of a playoff spot, but I’m starting to see what they mean about stats being confusing. And there’s not even any math involved in this one.

Anyway, Travis Yost actually dug up the standings for the last five full seasons and found 19 teams that made the playoffs despite being 9th or worse in their conference on December 1. That’s less than four per season. The chances of making it must be really low, right?

Well, on December 1 there are 14 teams not in a playoff spot. So if you’re one of those teams, you would be 28% likely to get in, just based on historical percentages. But as I said above, historical percentages have just as much bearing on the final standings as made-up percentages.

What does have a bearing on final standings, however, are the possession metrics. In particular, a given team’s ratio of unblocked shot attempts for/against during score-close even strength play (FenClose in fancystats jargon) has been shown to have the greatest relationship to final standings points than any other statistic, including goal differential and even current points.

There are two things you should notice when you look at that list of teams that did come back to make the playoffs. First, the Canucks are on there twice. As we are well aware, this team has recovered from some pretty bad starts to a season. But again, that’s just a statistical anomaly. The important thing to take away is that the FenClose% for most of those teams is around 50% or higher.

This year, the Canucks are sitting at 53.7% FenClose. That’s currently sixth best in the league. It’s barely behind LA and San Jose, and ahead of four of the teams they’re chasing for playoff spots, including the Ducks. Now, they are seven points back of the Ducks and Anaheim isn’t too shabby at 52.5% FenClose, so it’s unlikely they’re going to catch them for the third seed in the Pacific. But Phoenix, Minnesota and Colorado are all catchable, even with games in hand.

Coming into this season, all the talk in the advanced stats community was about how Toronto would be a litmus test possession stats. Despite the hot start, the Leafs have been a very mediocre team of late and are starting to come back down to earth. Turns out that the Canucks are also turning out be a test, but from the opposite direction. They are a better team than where they sit in the standings, and barring an incredible run of bad luck, injuries or both, the meaningful stats are actually in their favour. 

So take another look at those standings. Sure, the Canucks came into Dec. 1 three points out of the playoffs, while over in the Leastern Conference, the Leafs sat four points up on ninth. But which team would YOU rather have? 

Somebody's going to fall into a deep regression.

Finally, I want to end with a shameless plug.

If you like these poorly drawn graphs and you like know what day it is, then have I got the perfect thing for you!

The Graphic Comments 2014 wall calendar is now available, featuring 12 of your favourite graphs from the last year or so. This would make the perfect stocking stuffer for your favourite hockey fan. Well, it would be if they have a size 29 shoe, but you know what I mean.

Plus, next year you’ll know exactly when December 1 rolls so you’ll know when you can stop looking at the standings.

(PS: The online publishing site I used for the calendar is offering free shipping on any orders placed by 11:59 PM on Dec. 3. Just use the FREESHIP coupon code.)

RECENT GRAPHIC COMMENTS

    • JCDavies

      I didn’t mean to imply that you were cherry picking data, that was not my intention. FenClose is entirely appropriate here. Apologies.

      It is interesting that NM00 is finally interested in the quality of the statistics used to draw conclusions, even suggesting that it might be useful to use multiple statistics and perspectives in the analysis.

      I look forward to NM00 incorporating this higher analytical threshold into his future posts.

      • Graphic Comments

        Another abstraction devoid of anything concrete…

        It’s interesting, though, that JC is finally interested in deploying strawmen as his fallacy of choice.

        You’ll fit right in…

        • JCDavies

          I think I’ve written enough about your choice of data and statistics in the past six months to not need to rehash those here. If you want to take that as a lack of concrete evidence, then so be it.

      • Graphic Comments

        Haha, no worries.

        If NM00 wants to find his own data to try to refute a side point, more power to him. Good luck to him hanging his hat on a team sporting the highest PDO in the league.

        • Graphic Comments

          “Good luck to him hanging his hat on a team sporting the highest PDO in the league.”

          I am not hanging my hat on the Avs to maintain their lead on Vancouver BECAUSE of their high PDO.

          Good to see that one narrative is being replaced by another…

  • asdf

    “Phoenix, Minnesota and Colorado are all catchable, even with games in hand.”

    Is Colorado really catchable?

    In addition to the large buffer they have created for themselves, Colorado has been controlling play in certain situations.

    FF% Tied: 52.6, 265-239 (10th in NHL, 7th in West)

    FF% Down1: 60.5, 115-75 (2nd in NHL, 2nd in West)

    FF% Up1: 41.9, 220-305 (29th in NHL, 14th in West)

    Compare that to Vancouver.

    FF% Tied: 51.1, 391-374 (15th in NHL, 10th in West)

    FF% Down1: 59.2, 250-272 (4th in NHL, 3rd in West)

    FF% Up1: 51.6, 162-152 (5th in NHL, 4th in West)

    It should also be noted that the Canucks have spent a large portion of games down by a goal:

    http://stats.hockeyanalysis.com/teamstats.php?disp=1&db=201314&sit=5v5down1&sort=toi&sortdir=DESC

    4th in the NHL and 1st in the West spent down by a goal, though the Canucks have played more games than most every team in the league.

    Vancouver has been adept at controlling play while down by a goal.

    Not surprisingly, a young Colorado team has been inept at controlling play while up by a goal.

    Not sure that this makes Colorado “catchable”, though…

    • Graphic Comments

      Well, guess it depends on what your definition of catchable is. I didn’t they would be caught, just that they’re catchable. How? Well…

      Despite the fast start, I don’t think it’s out of the question for Colorado to go .500 the rest of the way. That would put them at 95 points.

      To catch them, Vancouver would need 62 points in 53 games. That’s a 0.58 pts%. They’ve been at 0.57 pts% so far this year.

      So yeah, I think Colorado is catchable.

      • Graphic Comments

        Is Colorado catchable? Well by factoring in the work of the Denver police department, then certainly their goaltending is “catchable”! And a team with no goaltending is the Philadelphia Flyers.

          • Graphic Comments

            Is it really up for debate that Colorado has more upside than Vancouver?

            It’s been a pretty long time since Vancouver has outscored their opponent (86 games and counting) with pretty much the same old core.

            It’s not even remotely comparable to Colorado even though UC continues to rely on strawmen for his punchlines…

          • Graphic Comments

            Again, depends on what you mean by upside.

            I’m still talking about this current season we’re in. And from that perspective, I see a Colorado team that has outperformed the underlying stats, and a Vancouver team that that has underperformed the underlying stats. So, yes, I think the Canucks have more upside in terms of standings points over the remainder of this season.

            Longer term? I’m definitely not arguing that Colorado has a young core with tons of upside compared to the Canucks.

            Now, about that second sentence…you’ve completely lost me.

          • Graphic Comments

            But why not consider the Fenwicks based on whether or not the team is tied, up by a goal or down by a goal?

            Why not look at whether or not Vancouver’s Fenwicks are being propped up largely by the fact that they have been down by a goal A LOT?

            Again, the article to which you link notes that at the 30 game mark the difference between Fenwick Tied & Close is marginal.

            At the least, shouldn’t both metrics be considered to give a more complete picture?

          • Graphic Comments

            Dude. I said catchable. As in, it’s not out of the question. And yes, I do rely on FenClose as backup, but I would still say it’s a possibility even based on FenTied, which, as sample sizes go, is 40% smaller than FenClose for the Canucks right now.

            But since you’re still on this, are you saying that you are 100% sure that the Canucks cannot catch Colorado? If so, what odds will you give me?

          • asdf

            Again, the article to which you link suggests that even with the smaller sample size there is a marginal difference between FenClose & FenTied at the 30 game mark.

            If your goal is to dispel narratives as the title implies, wouldn’t it be more useful to use both metrics as opposed to the metric that makes the Canucks look better?

            No I am not saying (and have not said) that the Canucks cannot catch Colorado.

            Edmonton and Calgary can still catch Colorado…

            Realistically, 12 of 14 teams have a chance in the West at the playoffs.

            But sure, I’ll go $100 to $50 that Colorado will end up with more points than Vancouver this season…

          • JCDavies

            “If the goal is to dispel narratives, perhaps these things shouldn’t be ignored…”

            “Again, the article to which you link suggests that even with the smaller sample size there is a marginal difference between FenClose & FenTied at the 30 game mark.

            If your goal is to dispel narratives as the title implies, wouldn’t it be more useful to use both metrics as opposed to the metric that makes the Canucks look better?”

            Ha! Pot meet Kettle!

      • Graphic Comments

        Sure they’re “catchable” by a relaxed definition.

        Of course, Dallas or Nashville may very well be the team that does the catching…

        I’m not sure, though, that the Canucks are actually “better” than Colorado or that the difference between the Fenwicks can’t largely be attributed to two things:

        1. One team has had the lead a lot and the other team has been behind a lot.

        2. One team is young and haven’t quite learned the finer (defense) points of the game yet while the veteran team, with all its limitations, knows how to play defense.

        Sure, the Canucks are controlling 5 on 5 a lot. But they have also been down a lot and have not been doing an espescially good job of controlling play with the score tied…

  • pdx

    The article to which you link also notes that the difference between Fenwick Tied & Fenwick Close% after 30 games is marginal.

    The Canucks are 15th in Fenwick Tied and 10th in the Western Conference.

    These numbers have also been accumulated without a significant injury while playing a disproportionate number of games against the inferior conference.

    If the goal is to dispel narratives, perhaps these things shouldn’t be ignored…

  • pdx

    The article to which you link also notes that the difference between Fenwick Tied & Fenwick Close% after 30 games is marginal.

    The Canucks are 15th in Fenwick Tied and 10th in the Western Conference.

    These numbers have also been accumulated without a significant injury while playing a disproportionate number of games against the inferior conference.

    If the goal is to dispel narratives, perhaps these things shouldn’t be ignored…

  • pdx

    The article to which you link also notes that the difference between Fenwick Tied & Fenwick Close% after 30 games is marginal.

    The Canucks are 15th in Fenwick Tied and 10th in the Western Conference.

    These numbers have also been accumulated without a significant injury while playing a disproportionate number of games against the inferior conference.

    If the goal is to dispel narratives, perhaps these things shouldn’t be ignored…

  • pdx

    The article to which you link also notes that the difference between Fenwick Tied & Fenwick Close% after 30 games is marginal.

    The Canucks are 15th in Fenwick Tied and 10th in the Western Conference.

    These numbers have also been accumulated without a significant injury while playing a disproportionate number of games against the inferior conference.

    If the goal is to dispel narratives, perhaps these things shouldn’t be ignored…

  • pdx

    The article to which you link also notes that the difference between Fenwick Tied & Fenwick Close% after 30 games is marginal.

    The Canucks are 15th in Fenwick Tied and 10th in the Western Conference.

    These numbers have also been accumulated without a significant injury while playing a disproportionate number of games against the inferior conference.

    If the goal is to dispel narratives, perhaps these things shouldn’t be ignored…

  • pdx

    The article to which you link also notes that the difference between Fenwick Tied & Fenwick Close% after 30 games is marginal.

    The Canucks are 15th in Fenwick Tied and 10th in the Western Conference.

    These numbers have also been accumulated without a significant injury while playing a disproportionate number of games against the inferior conference.

    If the goal is to dispel narratives, perhaps these things shouldn’t be ignored…