San Jose Sharks v Vancouver Canucks

Whether you’re into advanced stats or not, you should definitely give PDO, a terribly named measure of “luck,” a chance.

Our own Cam Charron (well, he’s everywhere, but you get the point) laid it out for us in plain english here, but in a nutshell, it’s a combination of a player (or team’s) on-ice shooting percentage plus his team’s save percentage while he’s on the ice. Basically, if a guy’s PDO is really high he’s had some good luck, really low and he hasn’t, and the idea is that everyone will regress back to 1.000 (or 1000, however you want to write it), because nobody can shoot 20% all year, and nobody is going to be on the ice while the goalie has a .600 save percentage all year. Those numbers are impossibly unsustainable in the NHL today barring them putting me in net and starting me on the daily with magazines taped to my shins for pads.

Only…a lot of good teams tend to have high PDOs, and a lot of bad teams tend to have low PDOs, and I’m unwilling to say that those teams are good and bad because of their luck. One of my current beefs with the advanced stat community in hockey is an over-attribution of luck to success. Hockey players are taught to create luck, and I think some players and teams are better at doing it than others. I’ll get more into that in a sec, but let’s take a look at the numbers.

Cam Charron does a weekly update over at NHLNumbers.com of team’s PDO’s (and more) and where they sit, which I’ve cribbed below:

teampdo

I know, I’m annoyed the teams/numbers don’t align perfectly too, but we’re all in this together, so let’s look past that.

Eight of the 10 teams with the highest PDO in the league are among those currently qualified for playoffs. Seven of the 10 teams with the lowest PDO are among those currently on the outskirts of the post-season looking in, with an eighth, the Islanders, also technically in the league’s bottom half (16th).

My (admittedly somewhat cursory) understanding of PDO is that folks are looking at this data and saying “the teams in the bottom portion are due to improve because they’ve had unsustainably bad luck, and the teams at the top are due to get less lucky going forward, and should stumble” (particularly the Ducks/Habs have been pointed at in this regard, from what I’ve seen). Or “that team has been bad and unlucky, what are the odds?”

Last year, some people saw the numbers and accurately predicted the Kings success and the Wild’s mid-season decline, thus solidifying PDO as one of our most useful predictive stats.

Most people who’ve played hockey at a decent level (not saying pro, just minor hockey into bantam and beyond) have heard a coach talk about creating luck. And I’m not talking about some silly cliche like “You gotta be good to be lucky,” or anything like that (though I may be arguing that in a roundabout way), I’m talking about the things that teams do to create offense.

You try to get screens in front of goalies, you try to walk off the wall and throw the puck into net-front traffic, you shoot at the pads for rebounds, you blindly throw the puck to the danger areas to create chaos, which results in the occasional extra goal. You create luck in hockey. Good players are better at creating themselves that extra inch of room and getting the puck into trouble, versus turning it over. Stuff like getting the puck to trouble zones is an actual, conscious effort from forwards, particularly depth guys, who aren’t going to beat anyone the way Jonathan Toews would (incidentally, I’ve had the Blackhawks in my head the entire time I’ve been writing this). So, having better depth guys should help you create more “lucky” goals, it seems to me.

Even you rec hockey players out there know you do this – you play the odds, and good players know the odds better than others, and will do more dangerous things (offensively) with the puck in a moment of panic.

Maybe one team’s depth forwards are slightly better at elevating their shots 12 inches off the ice so if the goalie is screened they go in instead of catching some butterfly’d leg pad. Maybe luck creation really is a thing in hockey and good teams can maintain “unsustainably high” numbers and and bad teams won’t always regress to getting better.

I understand there are extremes that have to move, and I could be badly missing something advanced stat people have thought of here, but I don’t think it’s fair to blindly look at a hockey team’s PDO and say “they’re due to climb/drop.” Teams try to create luck, and I think the creation of it is a replicable skill in hockey.

New York Rangers v New York Islanders New Jersey Devils v Philadelphia Flyers Minnesota Wild v Vancouver Canucks New Jersey Devils v Buffalo Sabres Montreal Canadiens v Ottawa Senators Tampa Bay Lightning v New Jersey Devils

Comments (14)

  1. Funny, because I just wrote about the same thing last week. I found that the PDO for the Wild when they were on the Win streak was very high, well above 1000, and the PDO during their latest crapfest was in the 800′s. They were outshooting the opposition yet the PDO was horrendous.

    From the numbers alone, you would think “oh they are just really unlucky right now”, but when you watch the games they were just awful. Turnovers, not breaking the puck out, letting up lots of odd man rushes, etc.

    My conclusion was that PDO might be a measurement of quality of scoring chances rather than just pure luck.

    If your shooting percentage is high, maybe you are lucky…or maybe you are great at intercepting pucks and getting breakaways, or taking the puck through the neutral zone for an odd man rush, or maybe you’ve just got a really damn good shot (is Stamkos just really lucky to have high teens shooting percentage?).

    Similarly, with save percentage (which I’ve always considered a team stat, not just goalie), a good team will box out the opposing shooters, grab rebounds and move them up ice, force bad shots, etc.

    I don’t know that it’s useful as a predictive stat though. There’s a high correlation between PDO and regulation wins/losses (less sure about points, I counted two points for a reg win, two points for a reg loss, and one point for OT either way). But as we saw with the Wild, PDO can shift quite a bit.

    Although specifically in the Wild’s case, you could argue that losing several top six players led to fewer quality scoring chances for, less offensive puck possession, which meant greater time spent on defense, better scoring chances against, and then you see how their PDO would completely fall apart.

    TL; DR – I think PDO measures a lot more than just luck.

  2. Well written. I hate this PDO stat, I refuse to believe that players shooting % and goalies save % are a result of only luck. My biggest beef with the advanced stats group is many of them don’t seem to understand that correlation doesn’t mean causation.

  3. I dislike the term “luck”. I think it denotes that the performance we are seeing on the ice is attributable to chance or the fortunes of the Hockey Gods smiling (or frowning) on a particular player.

    The “luck” factor in most advanced stats is really just a bunch of variance that hasn’t yet been accounted for. Perhaps the tool we are using doesn’t accurately measure that variance. Maybe the tools for measuring what needs to be tracked simply haven’t been developed yet. Good teams with high PDOs are doing something to keep those numbers up there, but we aren’t able yet to figure out what that is with the tools we presently have at our disposal.

    I think this is why people are attracted to the idea of tracking quality of scoring chance data. It seems intuitive, but the data we can track and the tools we use for tracking it are only in their nascent stages and are simply too unreliable at this time. I think a portion of the luck or variance in performance we see will someday be attributable to the quality of scoring chances generated by the players on the ice. Creating quality scoring chances IS something that some players do better than others, as illustrated anecdotally by Bourne in this piece.

  4. So, wait, is this a post about Luck, or an ode to Jump by Van Halen?

  5. As a hockey player of over 20 years (nothing near professional; highest level was D3 ACHA), I can say that it was difficult to grasp the concept of “luck” until I started to view it as a ton of stuff going on in a hockey game that’s outside your control. There are so, so many moving pieces and no one player is ever aware of everything at play. That’s the basic concept of what I’ve come to view as “luck”. Things that are too random or remote to properly model or track. A guy losing his edge on a breakaway or somebody playing with a broken wrist would be good examples of luck or chance as single events.

    I think after a while I just came to realize that these kind of things happen all the time in hockey. Over time, they tend to even out. But it also makes sense that certain players could have a run where they are the beneficiaries of some incredibly good luck and vice-versa. Even as a recreational hockey player I’ve seen that. I guess that’s how I was able to put luck into a context that I could grasp.

    • And as a follow up, you could look at it this way:

      Every event in a hockey cause has a cause and result. If you were to ask someone why such and such happened at a single point in time, they could go back to the tape and explain what happened. But if you tried to track that over the course of a season, you have hundreds if not thousands very small events that accumulate over time. I think that would make explaining why a run of good or bad play took place very difficult. So it’s simpler just to refer to these minor events as luck or randomness or chance. Sort of a catch-all term.

  6. Is it “luck” when a team purposely practices (around-the-glass shots) using stanchions, and (figuring bounces off) lively boards to their advantage? Not really.

    It’s “detail-oriented”: playing slim odds to gain an edge. Everything counts when you’re out to create opportunities; these are subtle tools that contribute to success.

    Just like firing a shot off a defenseman’s butt, from the corner (behind the goal-line) and seeing it go in. You make your own luck, playing the percentages that favor you in any given instance.

  7. This is right on and something Sharks fans have been saying for years. The assumption that all team/systems will have the same shooting % and save % is ridiculous.

    In particular, the Sharks consistently generate shots but not scoring chances. All systems, shooters, and goaltenders are not created equal, the fatal flaw in PDO.

  8. PDO is a very misunderstood stat. Basically, the average of all player’s shooting percentages for all shots plus the league average for all goalie save percentages equals 100%. Well, duh. PDO of 1 is the league average, it does NOT mean each team will get there.

    Individual team’s average shooting percentages over the course of a year generally average between 8.5 and 10.5 percent. With outliers every couple of years.

    Teams goalies save percentages are, what?, in the .900 to .920 range. With outliers as high as .940. So a team’s actual PDO can easily very between .985 and 1.025 simply for skill differences between teams.

    So what is PDO useful for? It simply points you at which teams you want to look at in more detail. Can Anehiem maintain that really high shooting percentage? Since it was way above the career averages of its players, probably not. And they have come back some since then. Can Chicago maintain their .940 goalie average for two goalies? Uh ,no and they too have come back from their streak once that number started to drop.

    And to be honest PDO is more useful, imo, when comparing individual players. It helps take the goalie and team aspects out of a players individual performances.

    • This is one of the more useful explanations i’ve seen.

      Thanks.

    • Hit the nail right on the head.

      Unless the PDO is over 1.025 or under 0.985 it is possible the team isn’t lucky. They have an elite goalie/ elite shooters (or both). if you check historical PDO you can see nashville 1st in the league for 2-3 seasons in a row, which means if it does regress to 1 eventually, it takes longer than 1 season in which case there may not be a fall.

    • Good response Neal. Very few stats are all that useful taken in a vacuum, and as they get adopted we’ll learn how to use them more effectively and gain an intuition what sort of insight they can offer us.

  9. High PDO doesn’t mean you are due for a streak of low PDO. That is the gambler’s fallacy. It just means that you should consider the really High PDO going forward and go, “gee, maybe this guy won’t keep shooting at 37%”

    Exhibit A: Mike Ribereo at the beginning on this year. He has since regressed to a more believable shooting % and his goals/game have dropped. (and his corsi is terrible…. which should have been expected)

  10. It’s also worth noting that a team like Chicago has a Fenwick % above 50%, which is part of the reason they have a high PDO — the have the puck more than half the time at even strength.

    A team like Anaheim with a bottom-tier Fenwick sporting an astoundingly high PDO is absurdly unsustainable. Basically, everything they’ve shot towards the net has gone in. It’s therefore much more likely that the Ducks will regress than Chicago because Chicago has more time with the puck, ie. more chances to score.

Leave a Reply

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