PDO: If you were going to understand just one NHL statistic
by Hawerchuk on Oct 28, 2011 7:00 AM EDT
in Shot Quality
Email
PrintPeople sometimes ask me what the single-most useful statistic is in hockey. Obviously if we had a reliable way to measuring how many wins a player contributed to his team, that would be our choice, but the computational complexity of such a measure limits its value to someone who's looking for a simple yet powerful stat.
My vote for the simplest yet most-useful statistic is PDO (the name is short for its inventor, also named PDO.) PDO is nothing more than shooting percentage plus save percentage. But this simple statistic captures several complex and powerful concepts:
1. Shooting percentage is primarily luck-driven
We've gone through this a billion times - a season's worth of shots, whether for a team, while a player's on the ice, or just those taken by an individual player, simply isn't a large enough sample to overcome the role of luck in putting pucks in the net.
2. Save percentage is primarily luck-driven
The spread of goaltending talent is much smaller than most people suspect, and 29-year-old goalies jumping from obscurity to the All-Star game are hardly uncommon. Again, we've got single-season sample size issues, backup goalies, and a whole lot of luck.
3. The apparent talent there is in these quantities, particularly offensive finishing ability, shows up as negative talent at the other end of the ice
Add #1 and #2, and you get a statistic that is almost 100% luck. How can this be? Surely there are players or lines who are higher-percentage finishers than others and can also play adequate defense? I'm not saying there aren't but for the vast majority of NHL regulars, a high PDO in one season comes crashing down the next. And many players with high one-ice shooting percentages get them by cheating offensively, which leaves them susceptible to higher-percentage opportunities against them at the other end of the ice.
If I could pick one statistic I absolutely wouldn't want to bet against, it would be this one.
30 comments
|
4 recs |
Do you like this story?
Comments
Gotta love PDO
Whats even worse – from an NYR standpoint – is that nearly half the team (9 this morning) has a PDO of over 1020 despite their horrible play. I know its early, but this is not good a team who was dead last as of 10/25 in 5v5 SF/60.
Blueshirt Banter - Muddling in Mediocrity - Fantasy Hockey Scouts
US Soccer/Red Sox Supporter
Where would you go for up-to-date PDO stats?
Down Goes Brown - Unapologetically nostalgic for the past. Brutally realistic about the present. Grudgingly optimistic about the future.
by Down Goes Brown on Oct 28, 2011 10:11 AM EDT reply actions
Gabe’s own site has the individual numbers. Go to the bottom of the page and select “on ice shooting”, then adjust the filters on the drop down menus as you see fit. Here are the Leafs’ numbers thus far.
by Robert Cleave on Oct 28, 2011 10:53 AM EDT up reply actions
Cool, thanks.
Down Goes Brown - Unapologetically nostalgic for the past. Brutally realistic about the present. Grudgingly optimistic about the future.
by Down Goes Brown on Oct 28, 2011 11:35 AM EDT up reply actions
Does anyone maintain it on a team basis?
"[Phil Kessel]'s as streaky as a flipped coin" - Shift
by The '67 Sound on Oct 28, 2011 2:38 PM EDT up reply actions
Not that I've seen
I go to the team pages and bring my calculator.
Drop at line at The Backhand Shelf or any of the fine Nations Network blogs: Canucks Army, Jets Nation, The Leafs Nation
@camcharron
It’s labour-intensive in its own way, but Vic Ferrari has his playershots script running again. Here’s the link to Toronto’s EV tied performance in 11/12. Note that Vic’s script uses percentages rather than rates/60 for the outshooting data. You’ll just have to do the math and add up the EVSV and EVSH % to obtain the PDO for players and teams. For other clubs, simply replace “TOR” in the URL with the proper abbreviation for your team. If you want the full data, simply delete the word “tied”. Be advised, the script takes a bit of time to load. You can also look at segments of games by changing the game numbers. A season’s worth of NHL regular season games are numbered from 20001 to 21230. You can find those numbers on the game summaries at the NHL web site.
by Robert Cleave on Oct 28, 2011 2:56 PM EDT up reply actions
A few questions
Not that I disagree with anything you wrote, but few thoughts jump to mind.
You’re talking about individual-level PDO, I’m curious whether it has a tendency to 1000, or just to a certain point – 1020 for some players, and 980 for others. What should interest us – PDOs that vary significantly from 1000, or PDOs that are far different from a player’s established levels.
I seem to recall you once concluding that many of the 80s Oilers, for instance, sustained high PDOs.
Another question – does arena bias in recording shots make any difference? You’d think the answer is “no” given that the same scorekeeper is counting shots for as shots against, but if there’s a large gap in the number of shots a team takes and shots a team faces then the arena bias could skew the over-all PDO.
P.S. To your second sentence – is GVT really that tricky?
GVT isn’t tricky, it’s just flawed. Any stat that tries to be all-encompassing without accounting for context and luck is going to have issues. And if GVT did account for context and luck, it would be a lot trickier than it is!
Or put another way, GVT seems to be more of results stat than a process one.
by Stephan Cooper on Oct 28, 2011 3:47 PM EDT up reply actions
I like PDO as a stat, I’ve seen mc79’s post where he shows that it always seems to regress to 1000 but I just can’t wrap my head around that “1000”.
As a fan of the Leafs who constantly have horrid goaltending it seems to me that players playing for Toronto would have a lower “normal” PDO than players for say Boston. There has to be something I’m missing.
Pension Plan Puppets*
* Blog contains less than 2% puppet content by weight.
This is a great stat that speak a lot. For me, that just tell me if a guy is ready to go one step higher or if he really fit in his job. Like all other hockey stats you need to put them in their context to really appreciate.
how do you measure this?
enlighten me.
Blueshirt Banter - Muddling in Mediocrity - Fantasy Hockey Scouts
US Soccer/Red Sox Supporter
how do you measure PDO? or how do you measure if a guy is ready to go one step higher?
by Justin Azevedo on Oct 28, 2011 8:59 PM EDT up reply actions
Quickly, my thinking is those number (like every other hockey stats) directly link to teammates and opponents.
So depending of how a player is use, ie if a PDO is above 1000, your player certainly can take more minutes against better opposition.
Its a good way of giving overall results to a player.
That’s my way of reading those numbers.
It’s really hard to say what’s a bad PDO or a good PDO. You need to look at the big picture.
Example
I will go for an easy example.
first thing: we do not have enough game plays to really get some meaning from those numbers.
second: I’ m a CH follower so here are our numbers (sorry, don’t know how to put this link cute)
href=“http://www.behindthenet.ca/nhl_statistics.php?ds=34&f1=2011_s&f2=5v5&f5=MTL&c=0+1+3+5+29+30+31+32+33+34” target="new">
Just look at Weber, 1082 just old us that he do the job that his coach wanted him to (keep in mind teammates and opposition).
Now, my thoughts maybe Weber is ready to play better opposition, maybe his teammates just made him better than he is, or maybe by now he is been lucky on bot sides of the ice.
That said, by now we DON’T have enough game play to really get those numbers rights…and I really think that for really put meaning on a player PDO it has to be on multiple years. (this way opposition and teammate, and luck disappear)
Stat Names
I wish the advanced hockey stats had better names that are easier to understand. Baseball has done a pretty good job of this, where the name of the stat leads to an idea of what that stat will be describing. Whereas in hockey, you’ve got CORSI and PDO and others that do nothing to give you start in figuring out what exactly you are looking at.
Stupid question, but...
What does PDO tell us? What do the numbers mean? You say:
If I could pick one statistic I absolutely wouldn’t want to bet against, it would be this one.
But what would the bet in question be?
Silver Seven | Twitter | Facebook | E-mail | eBay
by Peter Raaymakers on Nov 16, 2011 11:23 AM EST reply actions
PDO tells you whether the performance of a player is sustainable or not.
James Neal great start has a PDO of 988. That means he’s not relying on luck
Phil Kessel great start has a PDO of 1076. That means he’s been fairly lucky and will likely cool off to “normal” levels.
The bet would be that Kessel will continue to have a PDO of 1076 or that he would slowly inch closer to 1000 for the rest of the year.
Fair enough. But I don’t understand what bearing the play of the Toronto goaltenders has to do in measuring whether or not Phil Kessel’s point production (i.e. performance) is sustainable. My opinion is that shooting percentage is a simpler and more accurate gauge in determining what’s luck and what’s skill.
(This is especially the case because I don’t think Sv% is primarily luck-driven; instead, I’d wager that it’s secondarily luck-driven, and primarily skill-driven. But that’s a personal opinion I’d have no way of proving.)
Silver Seven | Twitter | Facebook | E-mail | eBay
by Peter Raaymakers on Nov 16, 2011 12:30 PM EST up reply actions
my guess would be that it has a much stronger correlation to +/- than to an individual player’s points.
That would probably be a fair guess. But then… I’d call plus-minus a highly suspect stat, and if PDO mostly just puts plus-minus into perspective, what’s the big deal?
Silver Seven | Twitter | Facebook | E-mail | eBay
by Peter Raaymakers on Nov 16, 2011 3:11 PM EST up reply actions
Comments For This Post Are Closed
Secondary Sidebar
User Tools
FanPosts
Community blog posts and discussion.
