• The Avs, Probability and Five Heads In A Row

    by  • October 22, 2013 • Hockey • 20 Comments

    Nick Cotsonika hit Twitter with a cri de coeur about an hour ago. I’ve pulled it into a paragraph:

    I have a minor worldview disagreement with Nick here. If you’re in the business of talking about what’s going on, there’s going to be pushback from people who disagree with you. You may not like it but that’s the business. You can choose to embarrass yourself in front of a large audience by freaking out about it, like some do (not referring to Nick), or you can try and back up what you’re saying.

    What really interests me though is Nick’s assertion that Giguere’s unsustainable save percentage being amazing because it’s unsustainable. (Clarification: Nick has since clarified that he means that it’s amazing that Giguere is 36 and adjusting to smaller pads and had a great three games.) I don’t really understand this sort of thinking. Now, this may be a sort of “Where do you set the bar for ‘amazing’?” issue but something that’s common isn’t that amazing. Giguere has a .981 save percentage through three games this year. I don’t have the data readily available but that’s something that happens with some frequency. Pick any goalie you like who plays a fair bit and you’ll find runs like that.

    Let’s look at from the perspective of Colorado’s 5v5 team save percentage, which I happen to have handy. I got into it with the @TH2NSTATSGUY last night about the Avs. He’s of the view that Colorado is doing something that makes their .960 save percentage possible. I assume that Nick would say the same thing about Colorado’s save percentage at 5v5 as he says about Giguere: it’s remarkable because it’s unsustainable. (Clarification: again, the pad/age thing; I misunderstood Nick’s point. The @TH2NSTATSGUY is still without a defence.)

    Here’s the thing. I went through and looked at last year’s data for teams on a rolling nine game basis. So for Anaheim, games 1-9, 2-10, all the way through to games 40-48 for Winnipeg. That gave me 1200 nine game segments. You know how many of those nine game segments saw a team post a .960+ save percentage at 5v5? 30. 2.5% of the time, a team in the 2012-13 NHL was in the middle of a nine game run with a .960+ save percentage.

    No fewer than 11 different teams managed to post a stretch like this. It actually appears to be a relatively common thing. That’s how hockey works. I would have a lot more time for people like @TH2NSTATSGUY talking about how Colorado’s doing some special (even if I still didn’t believe him) if they occasionally said “Well, this is just chance.” Nobody ever says that.

    Just to provide some context to all of this: in the 2012-13 NHL, there was a 1/40 chance that any given nine game stretch played by a team saw them post a .960+ 5v5 save percentage. The odds of flipping a coin and getting five heads in a row are 1/32. The odds of six heads in a row are 1/64. You could quite reasonably say that observing a nine game stretch in which a team posts a .960+ 5v5 save percentage is like seeing a guy get five or six heads in a row flipping a coin (but closer to five).

    Is that remarkable? Well, it all depends on where we draw our personal lines for remarkable. It really doesn’t seem that amazing to me though. It’s something that the game churns out. I’d be more impressed with the “analyst” who had the gumption to say that Colorado’s probably not doing anything special right now, man defence or whatever be damned, and that they’ll fall apart: the probability of a professional hockey analyst doing such a thing seems considerably lower than the probability of seeing a .960+ 5v5 save percentage over a nine game span.

    Email Tyler Dellow at tyler@mc79hockey.com


    20 Responses to The Avs, Probability and Five Heads In A Row

    1. Colin (@SkinnyPPPhish)
      October 22, 2013 at

      After getting trade to Toronto from Anaheim, Giguere put up a .965 sv% in his first 3 games, including 2 shutouts. He finished the season in Toronto with a .916 sv% in only 15 games played.

    2. Triumph
      October 22, 2013 at

      We seem to have a breakdown in the skill/luck divide here – I imagine Costonika’s point is that not only does it take a good deal of luck to have a run like this, it takes lots and lots of skill, and that stats people lose the skill forest through the luck trees – Giguere will no doubt have a run of bad luck (or diminished skill) and his SV% will crash to earth, but it’s a nice run. I suppose if it hadn’t happened at the beginning of the season, few would take notice, but that’s what the start of the season is for – there’s not much else to talk about anyway until the 20 game mark.

    3. Tom Benjamin
      October 22, 2013 at

      No fewer than 11 different teams managed to post a stretch like this. It actually appears to be a relatively common thing. That’s how hockey works. I would have a lot more time for people like @TH2NSTATSGUY talking about how Colorado’s doing some special (even if I still didn’t believe him) if they occasionally said “Well, this is just chance.” Nobody ever says that.

      One reason nobody ever says this is because they are paid to invent rational explanations for things that do occur just by chance.

      A second reason is that it is not merely chance unless any team was capable of producing a run like this. The save percentage is certainly unsustainable, but it is still damned good. So is the 8-0-1 record the Avalanche has posted. Last year the Avalanche posted a record that would produce a 67 point season over 82 games. What is the probability of a 67 point team having a nine game streak with that kind of record or that kind of save percentage? I daresay it will be a lot worse than the probability of flipping five heads in a row.

      The Avalanche record is not sustainable. They will come back to earth. But this streak tells me that the Avalanche “earth” is not at sea level.

      No fewer than 11 different teams managed to post a stretch like this.

      Which teams managed the trick? My guess is that Colorado joined some excellent teams. I don’t think you can draw any specific lessons from this run, but I do think you can generalize. “The Avalanche are not nearly as good as their record so far – nobody is that good – but it is clear they are a lot better than they’ve been for the past few years.”

      • Tyler Dellow
        October 23, 2013 at

        Tom –

        I doubt the Avs were a 67 point team on true talent last year – a short season doesn’t just allow for a Toronto Maple Leafs, it allows for a Colorado Avalanche. The Avs shot 7.1% at 5v5 last year – probably some misfortune in there.

        As far as the chances of a team going 8-1-0, it happens. Bad teams have good runs. The Avs aren’t doing well shots wise, so I don’t buy it. A .960 save percentage covers up a lot. I disagree with you on the Avalanche earth not being sea level because the whole thing depends on a .960 save percentage. Nobody does that. If they shots had changed substantially, I’d be open to this but I don’t see it.

        Which teams managed the trick? My guess is that Colorado joined some excellent teams. I don’t think you can draw any specific lessons from this run, but I do think you can generalize.

        Anaheim, Columbus, Chicago, Detroit, Nashville, Phoenix, Pittsburgh, St. Louis, Toronto, Vancouver and Washington. Some good teams, some bad teams, some team with notoriously good goaltending, some teams with notorious goaltending problems.

        We shall see.

    4. scott
      October 22, 2013 at

      I don’t think you’re comparing apples to apples here. You’re using 5vs5 save percentage data to point out that Colorado’s goaltending is relatively common. Except Colorado’s team save percentage is .961 right now, including 4 vs 5.

      I don’t have the data for teams that post 9 game stretches of .960+ goaltending, but I spsect thats its much less common and perhaps even more “amazing”, for lack of a better term.

      • scott
        October 22, 2013 at

        sorry, I meant that I don’t have the data for teams that post 9 game stretches of .960+ goaltending in all situations (as opposed to only 5vs5).

    5. Steven Olson
      October 22, 2013 at

      I think Tom brings up an interesting question. If you looked at a full season, does a 9 game stretch like this give any indication of the teams overall strength? I don’t know what the best way to divide would be, playoff/non playoff team, corsi, goal differential, but it would be interesting to see if/how much more likely good teams are at doing this compared to bad teams.

      • Tyler Dellow
        October 23, 2013 at

        If you look at hte list I posted in reply to Tom’s comment, you’ll see that there’s a mix of teams last year.

    6. nanodummy
      October 22, 2013 at

      I think you’re setting the amazing bar too high, especially considering the journeyman blueline employed by the avs. Mr. Benjamin’s point might be relevant: who are the teams hitting these marks and who are the goalies? Are they bottom dwelling, rookie and journeyman loaded dumpster fires with dubious and/or aging goaltenders, like Colorado has been for the last few seasons? The Floridas, Calgarys and Edmontons? Or are they Elite teams with real goaltenders?

      I’d also note that any goalie who makes the cut at 36 for the top 30 or 40 of the thousands of goalies in the world, like Giggy, and can post a 1/40 odds stretch is looking pretty amazing. Maybe you’re desensitized by watching too much elite hockey?

      • Tyler Dellow
        October 23, 2013 at

        The list is posted in reply above. Edmonton’s done this with Khabby in a year that they finished 29th, IIRC. When the inevitable happened and the team sank, Tambo bemoaned their inability to maintain their earlier level of play.

        Every NHL player is amazing in a sense, because he’s defied amazing odds to be an NHLer. I assume we’re setting the bar a bit higher than that. Giguere is a goalie who is good enough to be in the NHL at 36. If he’s a guy who just walked off the street, different story.

    7. Brent
      October 22, 2013 at

      I think Scott raises a good point as well. There’s a world of difference between ES save percentage and overall save percentage, but Tyler is using only 5vs5 data.

    8. custard
      October 22, 2013 at

      I think the real reason behind this type of thinking is that human beings are just poor at intuitively estimating probabilities. Daniel Kahneman and Amos Tversky have demonstrated this even among professional mathematicians and statisticians.

    9. nanodummy
      October 22, 2013 at

      Fun fact: The median odds on winning the stanley cup are 30/1. 20/1 if you look at playoff contenders. Does that make winning a cup amazing or just something that happens in hockey? Mind you you have 8 more chances a year to post regular season numbers like this…

      *removes tongue from cheek*

      • Simon
        October 23, 2013 at

        Well actually, winning the Stanley Cup is highly dependent on luck, including injury luck, shooting luck, goaltending luck, etc. What’s amazing is being a contender several years in a row, which is what allows you to win Stanley Cups. That’s accomplished through strong possession, as measured by statistics such as Fenwick and Corsi. When a strong possession team gets enough luck in the playoffs, they win the Cup.

        • nanodummy
          October 23, 2013 at

          The point in my sarcasm was that reductivism like that ruins the narrative, and denies the part of the game that isn’t luck it’s due credit. Also being lucky can be considered amazing. You won $1000 dollars on a roulette wheel? Amazing! You won the lottery? Amazing! You found a Fifty Dollar bill outside the bar and now get to drink for free tonight? Amazing!

          Sports are entertainment and entertainment usually involves narrative. We want stories about sports because it makes them more enjoyable. This is what Cotsonika is defending. He also asserts that sports narrative based in reality is superior to cliche and fiction. To know that a team outplayed (outshot, out goaltended, out coached, out managed) the other and did it spectacularly is a fun narrative.

          To know that these Avs have been for the most part outplayed by the opposing teams skaters but have had their goalies carry them is also a great story.

          It’s the story of Varlamov gaining a big lead for contention as a Russian starter.

          It’s Giggy, a once top flight goalie in his twilight coming in and keeping up with the younger kid with something to prove, helping him stay in contention for the starter’s job should Varlamov come crashing back to earth.

          It’s a team that has no business being on top of the league sitting near the top of the league.

          It’s a rookie coach setting and NHL record, held by his former mentor.

          Now when this all comes crashing down, the story goes from heroic romance to tragedy-comedy, but for these 9 games, it’s been nothing but fun.

          If these 9 games happen during games 97-106, these Avalanche look like legends, because they’ve got a cup and a ring to go along with the five heads in a row.

          But instead of drama, glory and joy, you can just flip a coin and give the winner a trophy.

    10. Chippy
      October 23, 2013 at

      Well written, and I agree with your take except for this part:

      “I would have a lot more time for people like @TH2NSTATSGUY talking about how Colorado’s doing some special (even if I still didn’t believe him) if they occasionally said “Well, this is just chance.” Nobody ever says that.”

      I mean, I don’t know who that dude is but tons of people, everywhere, in the hockey world have been saying just that. It’s what the mean when they say it’s unsustainable. It’s what prompted Nick’s rant. It’s not limited to super advanced stat guy, it’s the vast majority of fans. Certainly its been said by Avs players, the Avs announcers, and national analysts. It’s been said 3 or 4 days in a row on MvsW.

    11. Cobbler
      October 24, 2013 at


      Do you know if anyone has calculated the standard deviation for team save percentage (5on5). I think it might be interesting to see how much natural fluctuation might be present for each club.

      After 9 or 10 games might be a bit too early to get a good handle on this but by the 20 game mark I would think the pattern would emerge.

      • October 24, 2013 at

        Basically there is way to many variables in a team sport, like hockey, to ever draw any reasonable conclusions, thereby establishing a pattern of sorts. If your looking at the statistics of a chess player for example, those values would be a true indication. Basically the scientific method in team sports is the equivalent of our current understanding of the universe. Just way to many variables to be certain of any presumed or prophesized outcomes.

    12. Stephan Cooper
      October 24, 2013 at

      A key thing that is also lost when this happens is that there is 30 teams in the league. Which makes it pretty likely that at some point somebody is doing something particularly unusual that results in success or failure. And given that there are many potential unusual things that could be happening (shooting 15%, saving one in twenty, getting 6 one goal victories in a row), Odds are something that intitally seems mindbogglingly unlikely to be simple chance is going on at a particular moment is actually pretty good.

    13. Pingback: [Bits] We’re Not Gonna Not Talk About the Goaltending | All New York Islanders

    Leave a Reply

    Your email address will not be published. Required fields are marked *