PDO explained

Advanced statistics, fancy stats, underlying numbers—whatever you want to call them, the ones you can find on behindthenet.ca—are becoming more and more prominent. CBC’s first intermission panel for Saturday’s game between Anaheim and Vancouver had a discussion about offensive zone start deployment. TSN’s graphic guys showed a puck possession tracker at times last year with “shots for and against” as measurements. Sportsnet, is, uh, well, their Canucks broadcast team selected Zack Kassian as the first star of their Sunday night tilt against the Oilers after everybody else watched Ales Hemsky skate circles around everybody through the third period and scored a late tying goal and the clincher in the shootout.

Point being, there’s a lot more interest for analytics now than there was two years ago or so when I first started really getting into blogging. So much so that there are a lot of misconceptions about the way these statistics work or what they’re good for. For determining whether a player would be a good fantasy hockey pickup, his Corsi Relative number would be about as useful as his height and weight, yet I get questions all the time like that.

Forget Corsi Relative for now, though. The most important advanced statistic, or the most critical to understand, is the PDO number. I’ve been asked to write about it this morning.

Frequently Asked Question No. 1 – What does PDO stand for?

You’ll have to ask it’s creator, an Edmontonian who commented at old Oilers blogs under the Internet handle “PDO”. A lot of problems people have in understanding the new hockey statistics is that they don’t stand for anything. In baseball you have OBP, ISO and WAR, which stand for on-base percentage, isolated power and wins above replacement. In the same field of hockey, you have Corsi, PDO, and Fenwick, which are all named after the people who are generally credited with inventing the modern concept.

How is it pronounced? Well, whenever I’ve mentioned it in conversation (which is more often than you might imagine) I spell it out. “Pee” “Dee” “Oh”. It’s not pronounced “pedo” or “pidoo”.

Frequently Asked Question No. 2 – What does PDO count?

PDO, at a team-level, is the simple addition of shooting percentage and save percentage at even strength. If a team’s save percentage is .927 and they’re shooting at 8.0%, their “PDO” could be counted as 1007, 100.7% or 1.007. If a team’s save percentage is .900 and they’re shooting at 9.4%, their “PDO” would be recorded as 994, 99.4%, or 0.994. There’s no standardized notation for a PDO number.

Frequently Asked Question No. 3 – What does PDO have to do with trivia craps?

The first reference I can find to the PDO number is in August of 2008, when some Edmonton Oilers fans were dissecting the finish of the Oilers in the previous year. Though the team finished out of the playoffs for the second straight year, they closed with a 14-5-1 record, giving fans optimism for youngsters Ales Hemsky (then only 24) Sam Gagner (18) Andrew Cogliano (20) and Robert Nilsson (23).

Vic Ferrari, in an attempt to mitigate some of the optimism, wrote a fantastic post about trivia craps, and the effects of luck on winning and losing. In the comment section, a guy under the handle “PDO” made the following point about the current crop of Oilers:

Hows this for ugly? Lets pretend there was a stat called “blind luck.” Said stat was simply adding SH% and SV% together. I know there’s a way to check what this number should generally be, but I hate math so lets just say 100% for shits and giggles.

Oiler players who had over 101%:

Nilsson (103.9), GlenX (103.8), Cogliano (103.4), Stortini (102.8), Horcoff (101.6), Rourke (101.4), Moreau (101.4), Gilbert (101.1), Greene (101.1).

And Oilers who had under 99%:

Smid (98.7), Brodziak (98.5), Roy (98.5), Tarnstrom (98.00), Stoll (97.6), Visnovsky (97.3), Sanderson (96.9), Reasoner (96.8), Pouliot (96.7), Thoresen (95.5), Jacques (87.1).

You’ll notice the first group tended to get extensions while the second ground tended to get shipped out of town.

PDO’s point was that the addition of team shooting percentage and team save percentage when a player was on the ice could lead a lucky player to make management believe he was “good” and earn an extension. Shawn Horcoff signed that summer to an expensive extension with the Oilers. Since then, his “PDO” has been 101.3%, 96.0%, 100.8% and 97.8%.

That summer, Tom Gilbert was signed to an extension as well. His PDO numbers since? Well, it actually increased to 101.3% in 2009, but in the next three seasons dipped to 99.2%, 97.6% and 98.6%, the last season he was traded away.

Robert Nilsson? Also also re-signed. He only lasted two more years at the NHL level, with PDO numbers of 101.5% and 96.5% to his credit. After his last season where he didn’t have as much luck as his rookie year, he was gone.

Point is, you can’t really predict PDO numbers, other than say that they’ll be close to 1000, 100% or 1, depending on how you prefer your notation.

Frequently Asked Question No. 4 – So how does it work?

Ellen Etchingham isn’t a mathemagician, but wrote it out rather simply for us plebes in a post on regression at this blog: “Understanding regression involves nothing more sophisticated than knowing whether a given number is higher or lower than another.”

‘Regression’ here is the theory that since every shot taken in the NHL must result in a save or a shot, the mean PDO in the NHL is 1. The longer a player or team plays, the closer its PDO will get to 1. A player with a “hot start” that “cools off” is likely experiencing regression, because they can get wildly out of hand if a player has only played an hour of ice-time or so. One weak goal against the goalie and an unlucky bounce that results in a 2-on-1, and all of a sudden a player is minus-2 without really doing too many of the wrong things.

If a player has a PDO of over 1020 and producing well or playing with a PDO of below 980 and struggling, there’s a large, large chance that the struggles or the production is unsustainable, and it will normalize over the next handful of games

Frequently Asked Question No. 5 – Does this work with teams?

Yes. Tyler Dellow, in November of 2008, wrote a post about the best and worst first quarters, PDO-wise, between 2003 and 2007. He found that the Top 20 teams in PDO over the first quarter (combining for 103.1%) regressed to 100.5% in the final three quarters of the season. Similarly, the worst 20 teams in PDO (97.0%) regressed to 99.8%.

For a more recent example, I looked at PDO numbers through November 2011 for teams, taking the Top Five and Bottom Five. Look how they did between December and April, percentage-wise:

PDO After Nov. PDO Dec-Apr
NY Rangers 1041 1003
Boston 1026 1011
St. Louis 1021 1005
Philadelphia 1020 1007
Phoenix 1012 1009
Anaheim 974 991
NY Islanders 975 997
Ottawa 975 1008
Carolina 978 978
Columbus 979 1005

(Numbers derived from an app over at timeonice.com)

Columbus didn’t finish quite as bad as they start, New York didn’t finish as well as they started. Boston cooled off in the spring, the Senators came back to make the postseason.

Some teams can expect higher PDO numbers because they have better goaltending. New York has Henrik Lundqvist, Boston had Tim Thomas, St. Louis and Phoenix got crazy goaltending last season (although they aren’t likely to continue that this year) and Philadelphia… well, Ilya Bryzgalov had a better second half to the season than people give him credit for. Vancouver is typically up on these lists with Roberto Luongo, but they couldn’t shoot last season.

Frequently Asked Question No. 6 – Yes, but what about the quality of shots? Aren’t you forgetting that maybe some shots come from the outside and some come from better locations on the ice and are fired by better players?

Tyler explained that in the post linked above:

When I’m talking about luck, that’s the issue I’m talking about. It may be that there are other factors in play, the kinds of things that people love to talk about – Horcoff taking his shots from the outside, Smid making great defensive plays but it seems to me that in the long run, these things seem to disappear. If Shawn Horcoff will, in the fullness of time, start again getting his nose dirty in the crease and Laco will start losing the backdoor play again, I have a hard time getting worked up about their failures in that area for the time being. If the team percentages seem to snap back to much closer to the NHL average over the course of of an entire season, it seems silly to me to get too worked up one way or the other about what’s happening at any given point in time.

Recently, NHLNumbers.com writer Eric T. wrote that “shot quality factors tend to be small enough that they don’t grossly alter our understanding of the game, and they tend to be swamped by noise during in-season analysis.”

So the answer really here is that teams regress for a number of reasons. Adjustments could be part of it, but the randomness of shooting percentage absolutely swallows up any sort of “shot quality” difference between certain players or teams. Most of the affectation of shot quality by individual players or coaches can be easily matched by luck variance, so much so, that it’s almost worth it just to find players who can create a lot of shots at one end, and prevent a lot of shots at the other end. That’s obviously a very simplistic view of it, but if anybody attempts to navigate the forests of shot quality analysis and come up with a concrete, predictive vision of players’ or teams’ shooting percentage, they’ll go insane and probably get eaten by a Bengalese tiger.

Frequently Asked Question No. 7 – Why should I care about this? I just want to watch the game.

This is true as well. However, a basic understanding of PDO numbers and being able to notice outlying shooting percentages will help you make better trades in your fantasy hockey league, especially if +/- is a scoring category. PDO is found on the basic behindthenet.ca “Player Breakdown” page, along with individual components on-ice shooting percentage and on-ice save percentage.

Some people do like to be surprised by just waiting and seeing what will happen in a hockey season, but the variability of PDO and its overall effect on wins and losses, particularly over a short timeframe, is why most modern numbers analysts, including myself, pay closer attention to shot counters when trying to decide whether a team is any good.

For a recent, more scientific analysis of PDO, I recommend this post over at NHL Numbers.

Comments (5)

  1. “Lets pretend there was a stat called “blind luck.””

    I don’t think it’s blind luck. For comparison, I think applying PDO to a single player is as accurate as applying BMI to a single person to see if they’re overweight.

    I think it’s a very useful stat for teams, however. Interesting stuff.

  2. I can see this stat meaning something at the team level, but why would anyone calculate it for an individual player? Does it matter that a goalie’s SV% is higher than average when a certain player is on the ice? One thing a high SV% during a particular player’s time on the ice is that this player doesn’t block a lot of shots. What other meaningful information about a single player can we get from this stat?

    • Very little. Think of it like BABIP in baseball. Some players can expect to have slightly higher BABIP’s because they hit the ball a little harder, but generally every player has a mean that they can regress to.

      In hockey, the difference between “expected PDO” is fairly marginal. A player who plays 1600 minutes will almost assuredly have a PDO closer to 1 than a player that plays 400 minutes. There’s not enough information out there that shows that more than a handful of players can assuredly expect a significantly higher PDO than any other player. A few, Sidney Crosby, Henrik Sedin, Claude Giroux, can affect their teammates’ shooting rates, but the margin is slim, and so few players have the talent that from a practical standpoint it’s hardly worth worrying about.

    • Players who’s on-ice sv% is pretty low tend to get dumped on quite a bit, and players for whom its high seem like they become new defensive studs. Like Adam McQuaid got a .941 behind him, but I still don’t think he’s very good. Paul Martin got a .897 behind him, but since that’s his worst mark by far I’d expect him to sort some of his issues out and get back on track.

      • Gah, I meant *whose, not who’s.

        Or Jeff Schultz when he was +50 got .940+ goaltending from, mainly, Jose Theodore. And then he got an extension…(McPhee probably should have waited another year before deciding on term) John Carlson looked awful all season, but I trust he’ll get his defensive issues sorted out since he’s clearly an NHL player and NHL defensemen don’t stay in the league if they’re so bad their opponents shoot 11% on them consistently. (Carlson seems to have figured it out, if the playoffs are any indication)

Leave a Reply

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