Joe, the big cheese at BlueshirtBanter, has an article today called Hayes and Miller's Growth Vital to Rangers Success. It's a good article and you've already probably read it. Here's the comment that got me re-thinking.
I say "re-thinking" because I've been keeping an eye on Hayes and Miller all season for a few reasons. (1) I love watching them play offense. (2) They have been terrrrrrrible from shot metric stats. I have a secret wish that Hayes manages to lead the NYR in overall scoring, and have the worst Corsi For% of the NYR forwards. I suspect that would be the first time that happened in the Behind The Net era (since 2007), and some day I might actually check to see if that's ever happened. Just the other day I posted a chart showing Hayes' and Miller's ranks on some shot metrics compared to other forwards having over 500 5v5 minutes. Joe's article also looked at those ranks, but used a higher time on ice cut off (660 minutes).
Getting back to my initial point - Joe states what I believe would be conventional wisdom among fancystat-friendly NYR fans: that Hayes, Miller and Grabner are doing such a great job creating transition chances, it is part of their overall strategy, and makes them willing to accept some weakness on the shot metric percentages. Call it the Rope-A-Dope strategy: Take a licking on the defensive zone side, because you are waiting for that one chance to break out for a great scoring opportunity; and hey, if you miss because you tried for a high-percentage shot, you might be one-and-done in the offensive zone. I don't know if I even agree that is that line's (or the team's) overall strategy, but let's assume that it is.
There are ways to see if this theory is true. For example, we could look at actual rush chances recorded (which we'll do below). Also, we could look at passing data to see how many plays develop from a defensive zone breakout, and result in a scoring chance without many intervening passes (which we won't do). We could look at other indications of "shot quality", like expected goals metrics, to see if these players stand out (which we'll do). And we could look at how many shots the players are getting from forechecking events like dump-ins, recoveries, and low-to-high shot sequences (which we won't do).
Using Corsica.hockey (from now on I should just link to the Patreon page), I took all 2016/17 forwards playing over 650 5v5 minutes - I wanted to match Joe's 660 minutes, but that Corsica TOI slider is just a pain, especially on mobile (my kingdom for an input box!). Anyway, close enough - that cut off gives us 155 forwards to look at, including 5 NYR (in descending TOI order: Zuccarello, Stepan, Miller, Kreider, and Hayes). This doesn't include Grabner or Nash.
First, some rankings for the team shot metrics:
Deep red something is bad (less than 20th percentile). Deep blue something is good (better than 80th percentile). My key takeaways (many of which match the prevailing wisdom):
- Kreider is having an outstanding season. Upper level percentiles in every metric.
- Hayes and Miller are having a terrible season by the team's Corsi/Fenwick/Shots on Goal while they are on ice, but the team is good to great in getting Scoring Chances with them out there. Out of the group of 155 forwards, Hayes is Bottom 5 for C/F/SOG, but Top 10 in Scoring Chances For /60. That fits the Rope-A-Dope narrative (though I'll address this more below). Hayes is only 67th percentile in Scoring Chance For % - which suggests they are getting a lot, but giving up a fair amount too. Miller is similar.
- Zuccarello and Stepan are more of a mixed bag. Relatively poor to middling C/F/SOG, but again decent to great Scoring Chance For % and per 60.
If the overall team strategy is Rope-A-Dope, all of these results might be demonstrating that. Only Kreider shows dominance at every facet of play driving here.
Next, we can look at team and individual level expected goals.
Remember that Expected Goals models (and specifically Corsica's model) are attempts to model the types of shots, mixed with location, to give some idea of which shots are better than others. Also remember that Expected Goals do not equal Scoring Chances. [Read this on why Scoring Chances are a dangerous way to analyze on-ice results]. If Hayes and Miller are "trading" shot "quantity" for shot "quality", we should see it somewhat in their Expected Goals numbers - and I don't think we do.
First, their team Expected Goals For /60 are only middle of the pack, and no better than Kreider/Stepan/ Zuccarello. And their team Expected Goals For % are still very bad. Hayes and Miller are bleeding shots against while on ice, but they are also bleeding "expected goals" against. If Hayes/Miller are trying to Rope-A-Dope, they are giving up waaaay too much rope. We also see that Hayes and Miller aren't particularly good compared to their peers on their Individual Expected Goals.
When you consider that Kreider/Stepan/Zuccarello are all better than Hayes/Miller in the C/F/SOG stats, and almost as good as Miller/Hayes in the Scoring Chances results, and if we assume that the team strategy is Rope-A-Dope, then Kreider/Stepan/Zuccarello are doing it much better than Hayes/Miller.
Now we can look at some specific individual stats.
First, a note regarding iRB (Individual Rebounds) and iRS (Individual Rush Shots), as they may not be as broad as you would think. Corsica's definitions are here:
A rebound is defined as any shot taken within two seconds of uninterrupted game time of any other shot by the same team.
A rush shot is defined as any shot taken within four seconds of uninterrupted game time of any event occurring in the defensive zone OR within four seconds of uninterrupted game time of any giveaway or takeaway.
I don't have a link to why those are the cut offs (or even if there is a predictive value to looking at them at all), but regardless, they treat all of the players the same, so I don't have an issue using them as a high-level comparison tool here.
Second, note the really small total numbers here. Hayes (76th percentile) and Miller (67th percentile) are "good" at generating Rush Shots, and Kreider/Stepan/Zuccarello are "bad" (tied for 26th percentile). But in raw numbers, its only a 5-6 Rush Shot difference over 2/3rd of a season.
Aside from that, mu only other takeaway is that Hayes does have a big improvement from his Individual Corsi For /60 (23rd percentile) to his Individual Scoring Chance For /60 (88th percentile). So he, individually, isn't taking a lot of shots, but he is getting a high percentage of Scoring Chances from those shots. Part of that may be a result of (or the reason for) his closer average shot distance. It also fits my assumed pattern for a playmaker - only take the shots when you really think it is the best option.
As Individual Expected Goals are based, in part, on location, I pulled the 5v5 heat maps for each player from HockeyViz.com. I added in the skater's Individual Expected Goals/60 just to highlight.
Kreider/Hayes/Miller are very similar in how close their shots are, but there's a big difference in Individual Expected Goals between Kreider and Hayes/Miller. Zuccarello and Stepan take fewer shots, from all over the place, and have poor Individual Expected Goals. Overall, I see nothing here that suggests either a team style of play, or whether certain individuals are better at that style of play.
Finally, shooting percentage results.
Here's where I wonder if that Rope-A-Dope narrative really gets ingrained. Miler and Hayes are kicking NHL ass in shooting percentages. Hayes is # 1 in Team Shooting Percentage and Team Fenwick Shooting Percentage, and # 2 in Team Expected Fenwick Shooting % Percentage. [xFSh% is the expected shooting percentage for the NYR unblocked shots with Hayes on the ice]. Miller is also way up there for the team results, and both are pretty darn good on the individual Shooting Percentage results too. Interesting to me that Miller's Individual Expected Fenwick Shooting Percentage isn't that great, but his Individual Corsi Shooting Percentage is. Maybe that, plus Miller's relatively poor shot distance, suggests that Miller's shooting percentage results are truly more luck-based than Hayes, for example, who's expected and actual results put him in a similar percentile - though of course we're not comparing apples here, since Corsi includes blocks.
So what are my key takeaways:
1. Hayes and Miller are getting destroyed by most of the shot-based metrics. They only show strength on Scoring Chance results. While that may be evidence of their ability to play the Rope-A-Dope style, it seems to be counter-balanced by their middling-to-poor Expected Goals results. And for me, when in doubt, I will place more value on the Expected Goals results than Scoring Chances.
2. Hayes and Miller's shooting percentage results are outstanding. The same comment applies - to the extent is shows their ability to play Rope-A-Dope and create better "shot quality", the other results suggest that we may just be seeing an inflated shooting percentage that is unsustainable.
3. I might not have as much faith as others do that Hayes/Miller can keep this up. Maybe these results will continue, and they will continue to shoot the lights out. But if the team looks at these results, and says "THIS IS FINE", then I have some serious concerns.
4. It would really be useful to look at passing data here - I think it would really highlight "transition" plays much better than these results do. Another time.
...Again I'm right in my analysis,