Intro To Advanced Statistics - PDO

Editor's Note: Over the summer we're going to try to put together some primers on some of the statistical measures that fall under the "advanced statistics" umbrella in order to help lay the groundwork for a lot of the analytical discussion that takes places on the site and throughout a lot of the Bettmanosphere (trademark pending). This is a great chance to ask any questions, seek any clarifications, and share any links that can help increase everyone's undertanding (ie no "advanced stats are stupid" links).

PDO

What does the stat actually describe?

The Sum of On-Ice Shooting Percentage and On-Ice Save Percentage. It can be measured for individuals or teams. It's a very telling way of exploring a team's "puck luck" or which way the bounces have been going, though skill does heavily influence the result.

Where did the weird name come from?

Named after regular Irreverent Oiler Fans commenter PDO. Originally devised by statistical guru Vic Ferrari.

What it means and why it's useful

PDO is the sum of a team's On-Ice SV% and On-Ice SH%. typically quoted as 10 x the percentage. For reference we usually look at the Even Strength PDO as the vast majority of ice time is spent in this game state. Usually the number regresses towards the NHL average of 1000 over the course of a season, though for some teams their sustainable PDO will be higher due to excellent goaltending or consistently good shooters.

PDO = 10 x (On-Ice SV% + On-Ice SH%)

For instance, last season the Leafs shot 8.6% at 5v5, ranking 10th in the NHL in that regard. Their 5v5 SV% was .906, which ranked 27th. Their combined PDO score would thus be 10 x (90.6 + 8.6) = 992. 992 ranked 26th in the NHL in PDO last year, thus it could be argued that the bounces went against them - though the Leafs atrocious goaltending obviously contributed to such a result.

Teams like Boston and Vancouver consistently outperform PDO as a result of their excellent team goaltending and shooting. The Canucks in fact are the top team from a PDO perspective over the past 5 years, with Boston ranking 2nd.

Median SH% over the past 5 years in the NHL is 8.3% at 5v5 while SV% is .917. That makes the median PDO exactly 1000. Teams that are consistently above or below this level are proving that their results are likely more than just luck - particularly if they outperform by a wide margin.

Meanwhile, Toronto and the NY Islanders would be the worst under-performers over the past 5 years. Toronto's problems almost solely result from horrid goaltending (tell us something we don't already know) as the team's .906 5v5 SV% over the past 5 seasons ranks only slightly ahead of Tampa Bay's .905 SV%. New York on the other hand has gotten below average goaltending AND shooting for the past 5 years.

At the team level PDO will allow you to quickly spot a team unlikely to sustain its performance mid-season. Minnesota is a near dead on average team over the past 5 years, producing an average PDO of 1001.4 - largely due to excellent team defense and goaltending combined with mediocre offensive results. Last year in the early going that stretched to extremes when they were leading the NHL through the first 40 games of the season despite having an absurdly low Team Fenwick % of 44.2. The Wild had a team PDO of 1007, thanks to a very favourable .938 SV%, which was due to regress. By the end of the year it actually went pretty far the other way (as one would expect given their atrocious team fenwick%), and the Wild finished up the year with a team PDO of 989.

Compare that to the LA Kings, who through the first 37 games were out of the playoffs despite having a favourable Fenwick % of 51.5. After 37 games the Kings had a team PDO of 981, which was unsustainably low for a team dominating possession as they were. Despite having a solid team SV% of .923, there was virtually no luck going their way at the offensive end as the team was firing home only 5.8% of their Even Strength shots. As the year progressed, their goaltending improved even further to a .927 SV%, and the shooting bounced back to a poor, but more respectable 6.0%. Thus the Kings closed out the season with a PDO of 987. Still less lucky than Minnesota, but trending in the right direction, and more reflective of their quality of play.

In the NHL playoffs the Kings had a dominating PDO of 1038... they frankly destroyed the competition on their way to the Stanley Cup. The lowest playoff PDO this past season would go to the Detroit Red Wings at a paltry 904, who despite playing reasonably well, got no bounces at either end of the ice. The Wings actually outshot their playoff opposition by an average of 10 shots per game at Even Strength, but were outscored by almost 2 goals per game at Even Strength.

Player PDO

In a similar fashion, we can explore the PDO scores of individual players during the course of the year and get a read for who is benefiting from fortuitous bounces while they are on the ice. Perhaps that guy padding the score sheet regularly has been seeing the benefits of sieve like goaltending from the opposition, or maybe that guy that looks like the worst player you've ever seen (*cough*Brett Lebda*cough*) is actually being victimized by atrocious PDO as a result of his teammates not scoring, and the goalie not doing their job.

Again we can make excellent use of the data provided by TimeOnIce.com to examine the "luck" seen by players over a stretch of games. Conversely we can also pull the data from BehindTheNet.ca

For some Leaf related examples from the last couple of years I give you two players to examine: Darryl Boyce and Philippe Dupuis. Following the end of last year, Darryl Boyce looked like a lock to be a checking forward for the Leafs entering the 2011-12 season. Then the Leafs signed Dupuis to a 1 year contract and a battle ensued in training camp that saw Dupuis get the nod for the opening night roster.

What people remember about Boyce was his production in limited opportunity. The guy scored 5 goals and had 13 points in only 524 minutes of ice time, and he ended the season a +8 player. This despite facing tough competition, usually starting shifts in his own zone, and not being a flashy offensive player. So why should we have been cautious about his results?

Well - here are the Leafs Even-Strength Corsi, Fenwick, and Shooting and Save Percentages from the 2010-11 season. You can see that Boyce posted a Fenwick % of .438, a Corsi % of .430, and yet somehow... against all odds, Boyce enjoyed .923 goaltending, and the team shot 14.2% with him on the ice.

The goaltending isn't alarmingly good... at Even Strength, .923 goaltending is pretty close to NHL average, and in fact 19 Leafs had better ES SV% while they were on the ice... so his impact on the defense wasn't the reason for his stellar +/- numbers. But 14.2% On-Ice shooting? Was Boyce really responsible for that type of offensive production from his line mates? In a word - NO. He wasn't.

In fact, Boyce's On-Ice SH% was the highest in the NHL amongst forwards to play 30 games in 2010-11... by a margin over 2%!! His PDO of 1069 that resulted from the combination of the average SV% and ridiculously high SH% ranked him 2nd in the NHL. So let's just say, either he's one of the most skilled forwards in the NHL - or his PDO number from 2010-11 was unsustainably high, and he wasn't going to repeat.

After losing his job to Dupuis, Boyce was placed on waivers and picked up by Columbus... how did he fare there? Well his On-Ice SV% was .923 again... so that again wasn't really something he was ruining. But his On-Ice SH%? Yeah it dropped to 5.08%, and his PDO fell to 974. Indicating he's pretty far from stellar at the offensive end of the ice. Luck can make a big difference.

So who was the other guy? Oh right - Dupuis. Well in Colorado during the 2010-11 season, Dupuis worked a lot on the PK, and in 674 minutes of ice time produced 6 goals and 17 points. He looked like a comparable defensive player, but an offensive upgrade on Boyce - particularly when you factor in the likely drop off in Boyce's luck entering 2011-12. For the sake of discussion, Dupuis had a below average PDO of 984 in his last year with the Avalanche, mainly stemming from seeing .900 goaltending behind him at Even Strength. His On-Ice SH% of 8.4 is just under NHL average, so that implied he should be fine at helping put the puck in the net.

Logically we'd expect at least a slight improvement in the On-Ice SV% Dupuis would play in front of, and if he could replicate his On-Ice SH% then we might get at least a comparable 4th line C out of the exchange. Unfortunately for Dupuis that isn't quite how things worked out. His On-Ice SV% was fine... stellar in fact. At Even-Strength, the Leafs goaltenders posted a .966 SV% with Dupuis on the ice. He probably wasn't the driving force behind the stellar netminding, but he wasn't ruining the Leafs with defensive miscues apparently either. So... what was the problem? Offense - or a lack thereof. Dupuis' On-Ice SH% in 30 gp... was 0.00%... as in NO goals were scored with him on the ice.

Only 2 NHL players in the 2011-12 season played 30+ games and posted an On-Ice SH% of 0.00%. Dupuis and Eric Boulton of the NJ Devils. They are in fact the only NHL forwards to play 30 or more games in an NHL season in the past 5 years who were not on the ice for a goal for. So if you want to talk bad luck? I'd say Dupuis has a legitimate claim on it. Typically the guys with numbers that low are cement fisted enforcers, and checkers who see virtually no offensive zone time. This would be a case where Dupuis' bad luck prevented him from playing in the NHL.

So unsustainably bad or unsustainably good... there are often guys on every roster who are unlikely to repeat their stellar or horrid numbers from the year before. Then there are players who consistently outperform or underperform expectations. As I just mentioned, enforcers often have horribly low PDO scores, largely due to a complete lack of offensive contribution. Comparably, extremely skilled players often have above average PDO scores thanks to their positive influence on shooting., while top end goalies can sustain high PDO scores for their entire team thanks to high Even Strength SV% values. The following chart should effectively illustrate what I am referring to as the skill factor of PDO for skaters:

Obviously though, goaltending seems to vary wildly for both players, and it should be noted that it's extremely difficult for a forward to influence On-Ice SV% (which is completely logical). So when we examine PDO for players, it is not assumed that the PDO score should regress towards 1000 as is often suggested. What we should expect does depend on the context in which the player is used (usage metrics are a topic we'll get to later on in these postings), who they are playing with, and their own natural talent.

As Tom Awad of Hockey Prospectus lays out in this discussion of the results of the Columbus Blue Jackets, PDO is not JUST a measurement of luck.

"That's right, even after over 200 games, almost half of your shooting percentage and save percentage is luck...at the team level, slightly over half of team talent exhibits itself as puck possession, and slightly under half exhibits itself as finishing (and preventing finishing). It just so happens that the possession talent is much, much easier to measure." [emphasis added]

- Tom Awad, Jan 12th, 2002.

Similarly, David Johnson of HockeyAnalysis.com discusses the fact that luck is not the sole component of PDO here. To put it simply, we need to look at players long term patterns before assessing what's likely to happen in the future. At the individual level, a player's quality of team-mates, and offensive opportunities will likely influence their PDO one way or another, so when looking for future trends be sure to take this into consideration.

If you have any further questions regarding PDO, ask away. Similarly, if you're looking for any specific Adv. Stat insights, let me know in the comments, as this series is sort of open ended at this point and I'm filling time during August.

Follow @SteveBurtch