Expected PDO or How Sustainable is Getting Outshot?

Can he help stop PDO regression? - Jerome Miron-USA TODAY Sports

The Leafs all situations (including even strength and special teams) shots for (SF%) is 43.8% or said differently for every 0.78 shots the Leafs generate they give up 1 shot against. For contrast, their Fenwick Close is lower at 42.7%. With fair luck, if all NHL teams had the same league average team goaltending (91.5%) and team shooting percentage (8.5%) [which would result in every team having the same PDO] then the team that outshoots the other would win. Again, with fair luck, if NHL teams had varied goaltending skill (for example from 90% to 93%) and team shooting percentage (from 6% to 10%) then the team that outshoots the other would not necessarily win.

For example, if a team gets mightily outshot, the ability (or luck) to convert a shot into a goal at an above average rate is critical to winning (and similarly the ability to save shots at an above rate). In a sense, the team with the better SV% can tolerate more shots against than the average team with average goaltending. Further, a team with better team SH% can tolerate fewer shots for than the average shooting team. Or said differently, all other things being held constant, we expect an average shooting team to need to generate more shots for if they expect to win against a team with better team SH%.

PDO Required To Sustain Winning

For example, with fair luck, a team like the Leafs that is wildly outshot in all situations with a 43.8% SF%, will need to sustain a PDO of 1019 on average to win their games. [This PDO includes even strength and special teams situations but not shoot outs]. More generally, the chart below is the PDO required to win given various SF% (shot differentials). That is, these PDOs on average will give an all situation positive goal differential and thus winning record before shootouts for the given shot differential.

Shot For % PDO To Win
40% 1,028
41% 1,026
42% 1,024
43.80% 1,019
45% 1,015
46% 1,013
47% 1,010
48% 1,007
49% 1,003
50% 1,000
51% 997

PDO Regression to What?

That said, how do we differentiate between skill and luck when it comes to goaltending and team shooting (SH%)?Clearly, not all teams have a 1000 PDO as goaltending varies from team to team (i.e. Tuukka Rask vs Ondrej Pavalec). And if a team has many players with above average career SH%, then the team SH% can be elevated above league average. Much work has been developed to show that teams regress over the season and tend towards 1000. The choice of 1000 is somewhat arbitrary as many have pointed out that a specific team may have a different PDO mean than the league average value of 1000. Tango's variance formula tell us that after 60 to 70 games the skill components overtakes the luck components in PDO as referenced and described in PatrickD's PDO work.

Expected PDO

That said, another method to find "true PDO" based on skill components is to create an expected PDO for the team based on the career average SH% and SV% of each individual player/goalie. I calculated a simplistic version for all situations version of expected PDO for the Leafs by first finding the expected SH% of the individual players. That is, the summation for all players of [career SH% x (player 2013 actual shots/total 2013 team shots)]. And similarly the team SV% was calculated using the sum of the individual goalie’s career SV% x 2013 shots faced/2013 team shots against. Career SH% was used because the larger career sample size will have a higher probability of regressing to the mean skill value and thus better reflect the true skill rather than the luck variation of a small samples from any given years SH% or SV%. The chart below captures for the leafs, the actual PDO (1018) and expected PDO (1014) and the excess PDO which is the amount we might expect the team to regress (4.3 or 0.43%). All source data is from hockey reference (shots, SH%, career SH%) and actual team SH%, SV% and PDO is from extra skater. The spreadsheet for calculated expected PDO is found here.

Actual Expected Based On Career Average Excess (Actual Minus Expected)
Team SH% 9.8% 9.58% 0.22%
Team SV% 92% 91.79% 0.21%
Team PDO 101.8% 101.37% 0.43%

The Bottom Line

If the expected PDO of 1014 is a truer measure of the Leafs PDO, then the expected regression is not as large as we expect compared to a PDO of 1000. That is, because the Leaf skater have a slightly above average individual SH% and the goalies also are above average in SV% then the Leaf team could be expected to sustain a 1014 PDO. However, the Leafs may struggle at times to win because they still fall short of the 1019 PDO required to win based on how much they get outshot. The Leafs will need to improve their shots differential towards 46% or need the benefit of some small PDO luck in net and/or team shooting or a few extra shoot out wins.

Limitations and Future Work

This method is not perfect as player usage (PP, PK, ES, checking role, score effects, empty net) varies over the career but I suggest this method with a large enough sample size is more accurate then using an arbitrary regression to the league average 1000. The calculation can be cumbersome to derive but it can be extended to other teams beyond the Leafs to understand what is a fair PDO value for team to regress towards during the season. For example, a low SH% team like NJD or the Sabres would be expected to have a below average expected SH% and thus low PDO. Further injuries to a player can impact both shot differential and SH% (or SV% especially in the case of goalie) throughout the season. This was the case for the Leafs in replacing David Bolland who had a 14.8% SH% with Jay McClement whose SH% is 8.5%. And estimating career average for players with little NHL experience and few shots is also error prone and can be improved further (Peter Holland or Carter Ashton). This method shown maybe most useful for intra-season analysis as actual shot totals generated by the team and faced by the goalie are used. However calculating the expected PDO at the beginning of the season is more difficult because shot totals (and sometimes SH% and SV%) will vary under system/coach, player usage, roster turnover and injuries. Finally, an approach that separates out special teams from even strength and uses the greater shot sample size that corsi or fenwick provide may improve accuracy further.

More from Pension Plan Puppets is a fan community that allows members to post their own thoughts and opinions on the Toronto Maple Leafs and hockey in general. These views and thoughts may not be shared by the editor of