Zanzibar Buck-Buck McFate said:
Thanks herman ? you know this stuff backwards and forwards, if you could only have 3 advanced stats to show a player's (let's limit it to skaters) value what would they be?
It's pretty hard to answer with 3 stats for all skaters so it's going to be long
different roles and playstyles profiles lend themselves to different emphases. It's kind of why Wins Above Replacement/Goals Above Replacement models are created to try to be profile agnostic.
The basic 'advanced' stats are just a zoom out of the standard shots/goals/points/+- stats.
Consider the sales funnel of a sales organization: at the mouth, the widest portion, you have your passive advertisement, marketing, social engagement efforts; once you get to direct sales you have sales calls, then quotations, then sales that are actually agreed to and delivered on.
In hockey 'advanced stats' you have players' on-ice decisions (and puck bounces) within two competing schemes (and how well executed they are at any given instance) happening; the stats are just results logged of those micro events. At the mouth is having the puck on your team's stick: currently untracked on NHL.com without the chip trackers. Within the funnel proper, you have
- shot attempts (Corsi): CF/CA
- shot attempts that don't get blocked (Fenwick): FF/FA
- shot attempts that make it on net (shots): SF/SA
- shot attempts that make it into the net (goals): GF/GA
More advanced is simply taking those stats and basically doing a +/- for those events, and then calculating it over a rate of time (per 60 min), either isolating to the individual or logging when that individual was on the ice (and with whom). Even more advanced is taking an aggregation of shot metrics crossed with shot locations and developing an Expected Goals model, which applies an average success rate for a shot from any position on the ice given a league average shooter against a league average goaltender. This is the xGF/xGA you might see.
The unfortunate thing is we currently do not have public data on events that lead up to shots (where the passes come from and go to, where the shots target on the net).
Cross-referencing actual goal results against 'effort' results (upstream events) sort of gives a clearer picture of what's happening. Standard shooting percentage is a simplistic version of this (goals/shots on goal for). Auston Matthews, for example, has fairly above average shot attempts share, somewhat average xG, but a GF% that outperforms his xG. This means he controls play decently, takes shots from mid-range, but has either really good shooting talent or very good setups (he has both!). Zach Hyman has great CF%, exceptional xGF, but not the greatest number of goals (historically speaking). As you know from watching him, he frees up pucks for his teammates to get chances, goes hard to the net for rebounds, but his hands are somewhat granitey.
So to finally summarize, the stats that seem most pertinent to me at the moment are the ones that players have direct control over, be it because of their talent, adherence to a good system, or some other micro-decision that yields results (Matthews' toe-drag release). It'll be a bit of a different focus per player type if we also mix in microstats that are currently hand-tracked (zone exits/entries).
Very basic forward stats
- Shot attempts share (CF%)
- Expected goals share (xGF%)
- Goals share (GF%)
Very basic defenseman stats
- Shot attempts against (CA60): do they give up a lot of shot attempts
- Expected goals against (xGA60): do they help prevent dangerous chances
- Zone exits/assists: do they help move the puck out of the zone