Mo’ Greens Please: Empty Net Goals and IPP
- Updated: June 23, 2014
Recently, I’ve taken a look at the Individual Points Perspective (IPP) of some of the top CHL prospects. Here, I looked at the top draft eligible forwards. And, here I looked at the top 20 forward scorers in each of the three CHL leagues. IPP is a record of the percentage of goals a player earns points on while on the ice at even strength. It can be an indicator of both a player’s contribution to scoring and a player’s true talent relative to his team’s strength. Simply put, it gives us a sense of an individual player’s scoring ability relative to his team.
These studies were occasioned by a great new resource provided by McKeen’s scouting service. This resource is not without flaws. One major flaw is the inclusion of empty net goals in the even strength data.
In what follows I’ll review why empty net goals can skew data and with the help of a brand new fancy tool from extraskater, update my original IPP posts.
Empty Net Goals
There are two sides to the empty net goal coin that are interesting.
The first interesting thing about empty net goals is the idea of pulling the goalie.
This year, Oilers fans watched Dallas Eakins employ an aggressive approach to pulling the goalie. Eakins pulled the goalie while down several goals, when conventional wisdom would typically concedes the loss. He also pulled the goalie earlier in games than conventional wisdom is comfortable with. This isn’t new to Eakins as a strategy. When coaching in the AHL with the Toronto Marlies, he pulled the now Oilers’ goaltender Ben Scrivens in a playoff game against the Oilers’ farm team the Oklahoma City Barons with 2:32 left in a 4-1 game.
He’s not alone in this regard. Patrick Roy, current coach of the Colorado Avalanche, also has a penchant for aggressive goalie pulling.
This is not simply a matter of coaching temperament either. Roy and Eakins may both prefer aggressive hockey, but what informs their decisions here is also the sound analytic evidence that suggests the conventional wisdom is too cautious and waits too long to pull goalies. Tyler Dellow has taken a long look at some of the research done in this area. The long and short of it is: pulling your goalie early is a smart way to maximize your potential for wins.
The second interesting thing about empty net goals is the problem we face when accounting for the players on the ice when empty net goals are scored.
The greatly lamented stat +/- has a lot of problems––the most glaring being that it tethers the performance (good or bad) of linemates and goaltending to individual players. It also, punishes players who play on the power play and rewards those who play on the penalty kill. This is something Jonathan Willis has written about here. Mostly, the +/- stat persists within the hockey world so as to foster pearl clutching narratives about how terrible great hockey players are. See here, here and some smart responses here, here, here and here.
Another not-entirely-trivial problem with the +/- stat is that it counts empty net goals for and against among its members. In terms of boxcars, empty net goals are fairly marginal. A player rarely scores enough (or assists on enough) goals through the course of the year to dramatically frustrate a reliable reading of their scoring record. Jonathan Willis has examined this question and found that only in rare cases does a player score enough empty net goals to bear much interest.
Empty net goals scored (for and against) while a player is on the ice, however, do add up over the course of a season and can negatively or positively effect a player’s +/- stats. This situation is especially striking insofar as the players effected tend to be a select few. Coaches, over the course of a season, tend to ice the same cluster of players when a net is empty, whether they are trying to hold on to a lead or gain a tie. This creates a situation whereby a team’s offensive players tend to get dinged in +/- for allowing empty net goals against, something they have little control over considering their net is empty. Jonathan Willis has also looked at this question and shown how empty net goals can significantly skew a player’s +/- stat. Regarding the Oilers’ top offensive players he found the following:
The six forwards most impacted in a negative way by this are exactly the guys one would expect: the kids, plus Smyth/Horcoff/Hemsky. Including empty net goals in plus/minus really acts as a tax on top offensive players – since they’re the guys out there to try and tie the game. All of the kids are plus players when empty net goals are taken out of the equation.
Individual Points Percentage
IPP relies on two stats: even strength points scored and even strength goals scored for while a player is on the ice. Both stats are liable to be distorted by empty net goals. This is why stat gathering sites like stats.hockeyanalysis.com strip empty net goals from their their published data.
Unfortunately, the McKeen’s CHL tool I used to generate IPP records for various CHL prospects included empty net goals. (Since I published my articles, they managed to strip empty net points out of their data set but not from the On-Ice stats).
Now, with the benefit of extraskater’s CHL database, which does not include empty net goals, we can see how big a difference empty net goals make on IPP [*see note at the bottom of the article].
Here’s our original IPP chart of the top 24 draft eligible CHL forwards side by side with a chart of the same players using extraskater’s data:
(Click all photos to embiggen)
By and large, the ordering and IPP number roughly stands pat. A few jump or drop a percentage point or two. However, there are a few notable movers. Ho-Sang enjoys an big 7% jump by this new perspective and Reinhart moves up 6%. Barbashev and Bleackley both jump 4.5%.
Assuming this data is correct, Reinhart looks even more impressive, while mid-to-late first round options Barbashev and Ho-Sang also come off looking good.
The Song Remains the Same
What the new extraskater tool offers is a wealth of information on CHL prospects. It’s a fantastic development, esp. the time on ice estimates. And, while it gives me an opportunity to re-visit some things I had lingering reservations about, it turns out we haven’t learned a great deal more about these prospects from an IPP perspective for excluding empty net goals (which is not to say doing so is an unworthy pursuit).
What we have confirmed is that a few players were slighted in a non-trivial way on the basis of empty net scoring and probably deserve a slight up-tick in our regards.
I have some questions about the accuracy of these numbers, just as I did about the McKeen’s numbers. Mostly this relates to the spotty rink-side reporting for on-ice stats. The discrepancy between the original (apparently including empty net goals) McKeen’s numbers and the extraskater numbers is quite stark in some cases.
For example, if we look at a chart for the three top ranked centers with the original McKeen’s numbers we get this.
And, here’s a chart with the extraskater numbers.
If both sets of numbers are accurate and the only difference is empty net goals, it means that 9 of Reinhart’s even strength points contributed to empty net goals and that he was on the ice for 15 empty net goals.
These numbers struck me as high. So, I went through Draisaitl’s game-by-game stats looking to see how many empty net goals he contributed to and how many he was on the ice for, to see if I could account for the differences (I figured any of the 3 players should work).
That makes for a total of 46 points scored not at even strength. If we take Draisaitl’s overall number of 105 points and minus 46, we get 59 even strength points. Yet extraskater has him listed at 56 even strength points.
Interestingly, when McKeen’s did manage to strip empty net goals from their database (though not their on-ice numbers), they also got 59 points.
(I’ve included a screen shot here in case McKeen’s takes their tool down now that extraskater is providing the information).
In conversing with extraskater and McKeen’s about this on twitter, no one seems to be able to square the circle. For what it’s worth, I have no idea what the issue is or how to resolve it. But, I think it is worth noting.