daver
Registered User
I think that even strength scoring is more valuable than power play scoring when evaluating individual players, and it is absolutely worth differentiating between the two game states in any analysis.
First of all, there are a number of additional variables I could discuss here (many already mentioned in this thread), such as:
- adaptability to the playoffs
- decreasing marginal value of depth at a specific PP role compared to ES (imagine Ovechkin, Stamkos and Pastrnak all on the same team and how that impacts their PPG relative to their ESG)
- relative PP ice time varying even among elite players across teams and eras
- PP team effects being more strongly concentrated than at ES, since forward lines don't play with the same D-pairs all the time at 5-on-5 and few teams stack their top lines in the way that PP1s tend to do with their 5-man units
- PP stats being more role dependent, because roles are more specialized than at ES, and we see how a player's goals/assists split will change when they switch PP roles (this is particularly important if you're focusing on goal scoring or playmaking in isolation)
- ES play is more chaotic whereas a PP unit can practice and run specific patterns, suggesting it is more coaching/systems-dependent and less reflective of the individual player's skill
However, I'll focus on the two main points:
1. The replacement value of power play scoring is higher.
2. ES scoring tends to be a more reliable indicator of individual offensive contribution than PP scoring, especially over shorter time frames such as a single season
Replacement Level of PP vs. ES Scoring
Here are all of the forwards in the NHL in 2021-22, sorted by their average TOI/GP at each game state (source: Natural Stat Trick), and then separated into bins of 100:
5-on-5 Scoring, 2021-22 (Forwards Only), grouped by 5-on-5 TOI/GP:
Group Avg 5-on-5 TOI Avg 5-on-5 Pts/60 Exp 5-on-5 Pts/82 GP Top 100 Next 100 Next 100 Next 100 Next 100 Rest
Power Play Scoring, 2021-22 (Forwards Only), grouped by PP TOI/GP:
Group Avg PP TOI Avg PPP/60 Exp PPP/82 GP Top 100 Next 100 Next 100 Next 100 Rest
There are 32 teams and 3 forwards per line, so this is effectively giving us the typical scoring rates for players skating on each line, albeit fudged a bit because of all the usual call-ups and injury replacements over a long season.
It's important to note the selection effects here. Coaches will naturally give better players more ice time, which means that playing on the first line gives better linemates to work with (and vice versa for playing with 4th liners), creating a bit of a feedback loop. We can't assume that players will maintain exactly equal rates of production if they get shifted up or down the lineup, but it gives us a baseline to start from.
The key takeaways are that PP ice time is divided very unevenly compared to 5-on-5 ice time, and that first power play units tend to be very efficient. Individual teams will vary, but there is not always a huge difference between being on the 1st and 2nd lines at even strength (especially on a per-minute basis). On the other hand, there is usually a pretty major gap between being on the 1st or 2nd PP unit (or no PP unit at all). The replacement scoring rate at 5-on-5 appears to be slightly over 1 point per 60 minutes, while on the power play it looks to be around 2.5 (and it's really closer to 3.0 per 60 for players who get any sort of regular PP deployment). In addition, for 1st unit PP players, the per-game ice time with the man advantage is so low and the average scoring rate so high that a player has to be scoring at an off-the-charts rate to create a significant value advantage.
If a top line player misses a game you can bump everyone up on the depth chart to cover for him, but you are still going to end up needing to find a guy who previously wasn't good enough to make your 4th line and give him about 10 minutes of action at 5-on-5. However, that replacement likely won't even get a sniff on the power play, where most of your team's ice time will still involve 80% of the regular top unit plus someone called up from the second group (a player who is likely good enough offensively to be a regular in your top 6).
Even Strength Scoring Has a Higher Ratio of Signal to Noise
It is important to remember that points are only a proxy for offensive performance. If you make a great cross-ice pass to set up a teammate shooting at an open net, you made the exact same individual offensive contribution whether they scored or whether they whiffed, even though those end up being very binary outcomes in the points column. Over a long period of time that randomness will tend to balance out (subject to things like teammate quality), but the shorter the time frame the less that is guaranteed to happen. That's why it's foolish to say things like "Player X disappeared in that big game" simply because they scored 0 points. They might have been shut down, or they might have created 8 high-danger chances and hit 2 posts while keeping the other team out of their end all night, but simply never managed to be one of the last 3 players to touch a puck before it went in.
The key factor here is sample size. As we see in the numbers above, a typical first liner will get 4-5 times as much TOI at 5-on-5 as they will on the power play over a single season, making those stats less vulnerable to hot and cold streaks.
A recent example of an elite season created by unsustainable power play scoring is Patrick Kane's Hart and Art Ross winning 2015-16. Look at it in the context of his career:
Patrick Kane Scoring Results, 2014-2018:
Season GP ESP PPP ESP/60 PPP/60 PGF* PP IPP*
(*: PGF - Power play goals scored while Kane was on the ice, PP IPP - The percentage of PGF that Kane got a point on)
There is zero chance that Kane suddenly got ~60% better on the power play just for one 82 game stretch. Compared to his usual benchmarks, Kane scored an extra 6-7 points because he was involved in an unusually high percentage of scoring plays, and he picked up an additional 6-7 points because his team got hot on the PP overall, which is a tide that tends to lift all boats (especially for the main puckhandlers on any given unit). It is reasonable that he was peaking that year at least to some degree (as his much smaller improvement in ES scoring suggests), but I'm quite sure it's nowhere close to accounting for the majority of that PP discrepancy.
I think it's worthwhile to think conceptually of each player having a hypothetical "true offensive performance" which is separate from their actual goals and assists, i.e. what their point total would be after perfectly adjusting for extraneous factors like puck luck, usage and team effects. Since there is more randomness in the PP sample, we're less confident that it reflects a player's "true" output. Kane may have been credited with 106 points in 2015-16, but I'd estimate his "true" performance was something like 96 points, and I'm far more confident that 69 ES points over 1,413 minutes of ES ice time reflected his individual offensive contribution than his 37 power play points scored in just 256 minutes. Applying similar logic to all players leads to the conclusion that ES scoring tends to be a more reliable indicator than PP scoring when evaluating individual seasons.
We can avoid this problem by looking at a larger sample size (e.g. the entire 5 year period for Kane above), which will end up removing a lot of the randomness and giving us a better estimate of the player's offensive level. After all, the very best power play performers do deliver a lot of value over the long term by consistently creating lots of PP offence year after year (although even they are absolutely still subject to contextual factors like teammates and PP deployment, their skill simply transcends those factors more clearly than for typical players). However, a lot of player analysis and all awards voting focuses on 82 games or less, and that's why it is important to keep reliability in mind when looking at individual player seasons, especially those involving scoring outliers on special teams.
In regards to Kane, his PP TOI in 15/16 was lower than in the years surrounding. Maybe because the PP clicked that year than in other years. However, you can certainly point to his PP points being an anomaly that season that propelled him to his best regular season but the irony is that he has been much more of an elite ES player throughout his career while his PP points are surprisingly low compared to the other top ES scorers of his era.
So I don't see this as example of a player whose overall offensive resume is boosted by being a "PP specialist" or that year for that matter, he just clicked on the PP that year maybe due to Panarin.
I see players like Crosby and McDavid producing era best numbers at ES in one season then era best PP numbers in other seasons but their overall offensive production is the same. Ovechkin kept his overall goalscoring up even as his PP TOI went 25% as league PP opportunities decreased through the 2010s.
