• Learning From Xtapolapocetl

    by  • February 28, 2014 • Hockey • 8 Comments

    If you read or participate in Oilers game chatter on Twitter, you can’t help but notice that Sportsnet’s insistence on providing the home and road PP/PK information every game is a bit of a punchline. There’s a reason for this: it’s probably just white noise. If a team is doing far better at home or on the road in terms of the power play or penalty kill, it’s probably just because they’ve gotten some breaks at home or on the road.

    Let’s back up a second and talk about why we care about data like this. Some people care about it because they enjoy trivia. There’s nothing wrong with that – it’s cool that, for example, Taylor Hall set an Oilers record in New York this year, scoring goals eight seconds apart. It doesn’t tell us anything about what the future is likely to hold when the Oilers return to Long Island though. There is no moral or lesson. It’s just a bunch of stuff that happened.

    Not all data is like this. Some data is predictive. If the Oilers are travelling to Chicago to play the Hawks, it’s instructive if we know the records of each team because it gives us a sense of what might happen. People who are interested in hockey analytics are implicitly interested in this sort of data and in trying to sort it from the data that’s just a bunch of stuff that happened.

    So what are home and road power play records? Are they predictive or are they just things that happen? In my view, they’re just things that happen, subject to the caveat that every team has a different inherent level of ability. We’d expect Pittsburgh to produce more than Florida at home on the power play because Pittsburgh is better at power plays than Florida. If Pittsburgh’s been way better at home than away, or vice versa, we don’t necessarily expect that to continue because that’s the part of it that’s chance.

    In the big picture, teams do slightly better at home than they do away. So far this year, teams are 574/3116 on the PP at home (18.4%) and 502/2857 on the road (17.6%). You’ll note that teams get more power plays at home than away. At least some of the gap is due to an increased frequency of 5v3 at home. So far this year, the home team has played 120.7 minutes of 5v3 at home and scored 48 goals. Road teams have played 93.4 minutes of 5v3 and scored 29 goals. That gap in power play efficiency is actually smaller than it appears. There does seem to be a slight edge for the home team but it’s not very big.

    “Fine,” you might say “but some teams do do better at home than they do away.” That’s true. Here’s this year’s home and road PP% data.

    You can see that there are some massive differences in there between home and road results. My argument is that, in all likelihood, the vast majority of that is just chance – there simply isn’t enough power play time for home and road PP results to normalize. We cannot assume, just from the presence of a difference in home/road PP results that they mean anything and should assume that it’s just noise in the absence of a compelling argument to the contrary.

    How can I prove that this is a likely explanation? Well, what if we divided the games in half based on something so completely stupid that it couldn’t be driving things? Like, let’s say that we just went through our games and numbered every game a team played in order: 1,2,1,2. Then we looked at how teams did in the games numbered “1″ and how they did in the games numbered “2″.

    Here’s how teams have done in odd games and even games.

    I’ve had to switch from using PP% to using PPGF/60 due to NHL.com not presenting this information in an easily useable fashion but I think the point is pretty clear. Adam Oates can only be bothered to have a good power play in odd numbered games – possibly a character failing of his many Europeans? Maybe it’s a French thing because Vigneault’s team has the same problem. Hmm – but Julien’s team has the opposite problem, they’ve only got a good power play in even numbered games and he’s French too. It’s as if chance has a huge role in hockey games and even if you divide games on some completely arbitrary basis, you can get teams that look they’re dominant or struggling in some aspect of the game.

    Now imagine how stupid it would be if someone actually presented this data as if it told us anything other than the fact that chance can be a funny thing in hockey. Imagine if someone asked a coach “Your power play does better in odd numbered games than even numbered games – why do you think that is?” The questioner would (hopefully) be laughed out of the room. And yet this is precisely what happens every time home/road PP/PK numbers are discussed with no additional evidence being presented to suggest that a split is meaningful. It’s a funny old world.

    Email Tyler Dellow at tyler@mc79hockey.com


    8 Responses to Learning From Xtapolapocetl

    1. February 28, 2014 at

      If I could give you kudos or recs or +1s or whatever for the bizarre Simpsons reference in the post title I’d do it in a heartbeat.

      • Jay
        February 28, 2014 at

        No Maggie, not Aztec. Olmec. Olll-mec.

    2. Aaron Luchko
      February 28, 2014 at

      The tables undersell your case, the numbers actually look a lot more convincing if you use percent difference rather than absolute difference. ie

      (22.4 – 10.6) / 22.4 = 52.6%

      (11.5 – 5.9) / 11.5 = 48.7%

    3. boneshj
      February 28, 2014 at

      Great article.

      It’s like a comparison years ago I did for one of the forums showing how the Oilers played on certain days of the week. I can’t remember what it was specifically but it was pretty drastic as there were certain days of the week that they lost virtually every game and days of the week that they won every game. It didn’t tell us anything, it was just a fun observation.

      Now if a coach ever used that information to say they are doing fine (like Eakins did recently specifically on the road vs. home powerplay) I’d question his sanity.

    4. Redned
      March 1, 2014 at

      Methinks you slightly overstate conclusions. Isn’t what we mean is that it is *probably* luck which is determining things, rather than the X factor? In other words, an unknown unknown *could* still be at play within the numbers (but unlikely).

    5. Redned
      March 1, 2014 at

      Then again, maybe I overstated my criticism *upon rereading*.

    6. Benhur
      March 1, 2014 at

      Thanks I actually learned something….xtapolactocetal. Never heard of it before but i was never a Simpsons fan! Just joshing. I enjoyed your article.

    7. Art
      March 3, 2014 at

      I agree that you can toss out any outliers like Edmonton in the home/road splits, but the fact that home PP% is slightly higher every single year means there’s something at play–that sample size is large enough that it’s not just chance. And the 5 on 3 data doesn’t explain all of it.

      Throwing something like home/road splits out within a given season doesn’t tell you anything, but history says a home PP is (infinitesimally) more dangerous than a road PP.

    Leave a Reply

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