Journal of Hockey Analytics: Volume I Issue 6


The Rangers deploying advanced systems tactics

Welcome to this week’s version of the Journal of Hockey Analytics. The final game of hockey for the 2013-14 season has been played, which means that all of your favourite bloggers and writers will now be turning their attention towards the future; the draft, free agency, and trades.

Advertisement - Continue Reading Below

So for your slow Monday at work, continue past the jump for analysis of draft picks, and lots of cool little nuggets regarding neutral zone data and luck!

  • Rhys J looks at adjusting scoring for age in the CHL [That’s Offside]
  • Kyle Alexander has an analytical review of the Syracuse Crunch [Raw Charge]
  • Jen LC has released her years worth of Chicago Blackhawks neutral zone data with some analysis [2nd City Hockey]
  • Manny also releases his neutral zone data and analysis for the Ottawa Senators asking some questions and bringing some wrinkles to what we think we know [NHLNumbers]
  • Rhys J thinks the best player in the draft is Sam Reinhart and he is in with some elite company [Canucks Army]
  • Rhys also looks at adjusting all CHL draft-year CHL PPG by age and by era [That’s Offside]
  • Luck got a bit of attention in hockey this week with Barry Petchesky saying it is a “bullshit dump” [Deadspin]
  • Travis Yost addresses luck as Random Variance [Hockey Buzz]
  • While Jason Strudwick doesn’t believe in Puck Luck [Oilers Nation]
  • Not directly on hockey but this post made it’s way around this week on the topic of luck by Ed Smith [Four Seasons]
  • Neil Paine wrote on the possibility of a Rangers comeback when they were down 3-0 in the SCF [FiveEightThirty]
  • Neil Greenberg had a similar thought [Washington Post]
  • Adam Gretz wrote on how the Kings won []
  • IJay Palansky looks at who should win the Conn Smythe [The Star]
  • Adam Gretz writes why Justin Williams has always been very good [SBNations]
  • With the Kings second cup win Mirtle asks if they are a dynasty [Globe and Mail]
  • Drop it like It’s Ott analyzes the value of draft picks and looks to see if it is worth it for the Canucks to move from 6th to 1st [Drop It Like It’s Ott]
  • A fan post by Scott13 also was much talked about early in the week when the author presented his thesis on a thorough indictment of PDO [2nd City Hockey]
  • Darren Kennedy writes about the idea of Point Shares [McKeen’s Hockey]
  • Ryan Lambert writes about the limitations of evaluating goalies [Puck Daddy]
  • Travis Yost writes about buying low on players with low on-ice ES Sv% [The Sporting News]
  • JP Nikota asks if a team can suppress shots without possession [Pension Plan Puppets]
  • Garret writes about Defensive Defencemen and their struggles in the playoffs this year [Hockey Graphs]
  • Garret writes about the realistic expectations of a 40 goal forward [Arctic Ice Hockey]
  • Neil Greenberg thinks Mike Green is the game changer for the Washington Capitals [Washington Post]
  • Garret Hohl writes about Evander Kane’s fancystats from this last season [Arctic Ice Hockey]
  • WingedOctopus looks at Matt Niskanen’s numbers and comes to the conclusion he’s going to get paid this summer [Winging it in MoTown]
  • Garret also looks at Mike Richards and why the Jets should not sign him [Arctic Ice Hockey]
  • Borisnikov writes about Brayden Shenn as a possible trade target [The Oilers Rig]
  • Garret continues analyzing players looking at Bryan Little [Arctic Ice Hockey]
  • Neil Greenberg asks what the Capitals should do with Brooks Laich [Washington Post]


  • Thomas Davenport writes about what businesses can learn from hockey analytics [MITSloan]
  • Garret Hohl produced a primer of early hockey analytics posts [Arctic Ice Hockey]

      • MattyFranchise

        the adjusting for age in the CHL article too!

        is there a way we can make sure no one in upper management of Florida, Buffalo or Edmonton reads either one of those articles?

        • DragonFlame

          I think the Flames will be looking at Bennett or Draisaitl when their turn comes around. I just can’t see either Reinhart or Ekblad still being available.

          • SmellOfVictory

            I’d be happy with either Sam but if Reinhart falls to us I will do cartwheels! I really think Reinhart is the class of this draft.

            …all that being said I think the ‘worst case’ scenario is winding up with Draisaitl and if thats the ‘worst case’ I’ll be happy either way.

            • piscera.infada

              I honestly believe Buffalo could take Bennett ahead of Reinhart. He seems to fit the mould that they’re trying to create there. So that just leaves Edmonton.

      • SmellOfVictory

        I think the Flames should be happy with either Sam at 4th overall, but Reinhart is my favourite for multiple reasons – many of which don’t matter a lot, some of which do (less risky playstyle, more proven, more NHL ready, collection of family members, right-hand shot).

    • beloch

      Re: Scott Kennedy on PDO

      There’s some truth and some falsehood there.

      Kennedy argues that PDO does not measure luck, only skill. This is partly true, but demonstrably partly false. Good goalies typically have high save percentages and do fluff up their team’s PDO. However, predicting how goalies will perform from season to season is almost voodoo. Even Kipper’s world-beating seasons were interspersed with relatively mediocre seasons. Few goalies are reliably good year after year after year. What makes a good goalie’s sv% high one year and low the next? Does luck play no role at all? Player sh% also varies considerably from season to season. Do players have “lucky” or “unlucky” seasons also?

      Is a player or team performing at an unsustainable rate? Can this be measured and demonstrated with math? Sure. The first thing you do is to create a baseline for that player or team’s expected performance and then you compare it to current or recent performance and measure the variance. If you come up with a very high variance it’s likely unsustainable.

      Create a baseline for that player or team… Right. How do we do that? The ability of players changes considerably from year to year, and the composition of teams changes even more wildly. The sample size we can glean from a single season is too small to define a “baseline” for such rapidly moving targets. The baseline for an individual player is actually a curve, and not necessarily a simple one. Injuries, contracts, family events, and countless other factors all give it numerous bumps and dips. Even if the interaction of two player curves was simple to calculate (it isn’t), the complex interaction of all the players on a team would still result in a dazzlingly complex outcome that would likely resemble a chaotic system. In short, I reject the notion that this kind of analysis is likely to bear fruit.

      PDO is a flawed stat. It undeniably contains elements of skill, but it also contains heaping portions of random chance. For the sake of argument, let’s say that a player’s career sh% is set in stone the moment he is drafted. This is his inate ability and will never change. Already, this is an unrealistic simplification, but bear with me. That sh% applies to his career, not an individual season or game. Let’s say he plays 3 games, gets 6 shots, does not score, is injured, and misses the remainder of the season. His sh% for that season is 0%. Consider, alternatively, that he had just one shot in those 3 games, but scored. Then it would be 100%! This is the result of an insufficient sample size not reflecting his career average. More colloquially, it’s luck.

      PDO on the team level does give some idea of how repeatable a team’s performance is. If it’s high because of great goal-tending, as it was last season for the Oilers, then it’s reasonable to conclude that the team might not do as well if you take away the great goal-tending, as we saw this season. It’s not impossible to assemble a team of players whose skill is such that their innate shooting percentage will be significantly above league average, but it’s also not easy. Most teams will be pretty close to league average, so a team that is drastically outperforming the league average, such as Toronto was earlier this season, can be expected to regress.

      Personally, I’d like to see shooting and save percentages treated seperately. Save percentage is heavily influenced by goalies, and can change radically from season to season even for the same goalie. Team shooting percentage rarely deviates much from league average for long. That being said, PDO is still loosely correlated with “luck”, and it’s simple, so it will continue to be used.

      To those who would like to see a better measure of “luck”: please propose a formula.

      • Very true. Eric T took to the comments in that article with similar points.

        Team level PDO regresses to the mean 87% over the long-term. Now, if it were pure randomness, we’d expect 100%, which is why some teams can do the 101 thing adn some are more 99. But the functional difference between 87% regression and 100%, particulary over the course of small samples like a season or less isn’t practical.

      • SmellOfVictory

        Depends what you’re measuring. Realistically, for a player, on-ice SH% is all you need. The only thing that the SV% aspect of PDO affects in terms of counting stats is +/-. So SH% for individual skaters.

        Then on the team level, PDO is still useful to a degree even though there are teams that consistently outperform to a small degree (e.g. Penguins on SH%, Bruins on SV%, etc.).

        • mattyc

          @beloch, @Kent Wilson

          Problem with SV% comes down to causality. Does a team have a high SV% because they have good goaltending, or that they play strong defence, or that they play worse teams?

          Somewhere someone did a cumulative PDO graph with quantiles, that was pretty instructional. I’d link it if I could find it… (This one kind of… but using 3 years seems a little sketchy to me…). Regardless, it basically shows that there’s a range that can be expected as ‘normal’ for a certain team, but I find PDO is most useful for finding the huge outliers (like the Ducks or the Avs this year).

    • mattyc

      Given that Calgary is in the process of a real interesting rebuild and the fact that Calgary has some great MSM guys I keep coming to this site in hopes of finding something interesting and attention grabbing to read.

      Instead I find this. Lots of Fancy stats.

      Want to Kill an interesting read…talk about fancy stats.

      Want to make a really good game dull… about fancy stats.

      Want to sound like a geek that lives in your parents basement….Talk about fancy stats.

      All of you guys from this site need to have a talk with Jonathan Willis. He uses fancy stats when required, but doesn’t over load his blogs with them, thus by making his blogs readable and entertaining.

      I’ll stop by in another month and see if you have anything worth while reading.