clock menu more-arrow no yes

Filed under:

Prime Example of Not Getting it

New, comments

I hope you can follow this twitter conversation - our friends at Pension Plan Puppets wrote a piece on Mikhail Grabovski's performance:

MLSE: Grabovski Is The Man - Wednesday's FTB

Other Guy: meh, Grabovski is having a good season on a bad team. He's been Brian Bradley on the Lightning.

MLSE: Playing top competition he's doing better than any Hab.

Other Guy: top competition... you mean the backup goalies who face the leafs more than anyone else? and don't give me no QualComp stat... QualComp is a b.s. measure

MLSE: why do you say it's b.s.?

Other Guy: the Leafs are often trailing... teams that are trailing push to tie.. .teams that are winning, protect leads... offense is generated. there is a reason history of the league is littered with guys who scored on bad teams & did shit on good teams. He has more to prove. I've written a long post about the methods used to calculate it... its not a well thought out stat:

MLSE: NHL teams disagree with you though since they use a lot of these micro stats and are pursuing more

Behindthenet: First, use CorsiQC. Second, the point seems to be that +/- doesn't match Other Guy's view of who's good, not that QC gets comp wrong

Other Guy: CorsiQC; Corsi really isn't much better than +/-, stat still fails to differentiate b/wn shots & scoring chances as well as neglecting to adjust for many other things like score and situation, etc....I've just never bought the argument that hockey can be reduced to a statistical "moneyball" style analysis. its a team sport, where baseball has the 1 on 1 of the pitcher/batter matchup. none of them are ever going to replace a pro scouting department... it can compliment it, but watching the games is still far and away the best method to evaluate. Micro stats are taken with a grain of salt. is there a small amount of useful info in them? sure, and multi million dollar NHL teams look for every edge; but it will never be that they use them to the extent baseball did in moneyball. NHL teams will never follow stats blindly & will attempt to add context to them; that is mistake many stats guys make.


Brian Bradley...Now that's a trip down memory lane.  Through Age 27, Bradley wasn't a bad player - 2nd line PP, some PK, either solid defensive skills or the softest matchups in the league (he was routinely near the top in +/- on some very bad Canucks and Leafs teams), and a bit of trouble staying in the lineup.  But once he went to Tampa in the 1992 expansion draft, things changed for him:


85-92 80 18 33 50 -3 59 13 5 139 105 23 8
Tampa 92-96 80 29 48 77 -14 77 19 10 196 129 46 2


These stats are normalized to the offensive levels in 1992-93.  We don't have ice time numbers for that era, but based on the number of EV and PP goals Bradley was on the ice for, his ice time probably increased from 16.5 to 19.5 minutes per game, much of it on the power play.  Bradley was slightly more efficient in Tampa than earlier in his career, but fundamentally, he did what you'd expect a 2nd liner to do on the first line.

Does that resemble Mikhail Grabovski? Grabovski, MacArthur and Kulemin seem to have played together all season, and while they've faced middling competition compared to their teammates, they've been absolutely lights-out at getting the puck moving in the right direction:


MIKHAIL GRABOVSKI C  48 14.72 0.712 10 15.37 1 53.3 5 53.6
CLARKE MACARTHUR LW 49 14.08 0.716 9 13.48 2 52.6 4 51.4
NIKOLAI KULEMIN RW 49 13.57 0.827 8 8.84 3 53.6 7 53.5


This isn't a one-year thing either - Grabovski and Kulemin have matched up against middling competition for the last three years, and they've been in positive Corsi territory in each of those seasons.  No other Leafs can make that claim.

But I'm digressing.  What does this have to do with the back-and-forth between PPP and the other dude?  Basically, we see the anti-hockey analysis mindset in a nutshell:

1. Don't bother to actually understand advanced stats:  "Corsi really isn't much better than +/-"

2. Cling to the notion of shot quality: "stat still fails to differentiate b/wn shots & scoring chances"

3. Play up minor adjustments to a statistic that we already know about as through they invalidate the stat entirely: "neglect[s] to adjust for many other things like score and situation"

4. Go with their gut to beat on the 'stats' straw man: "I've just never bought the argument that hockey can be reduced to a statistical "moneyball" style analysis...NHL teams will never follow stats blindly"

5. Tell me to watch the game: "watching the games is still far and away the best method to evaluate"

6. And the key piece of the puzzle: lazy analysis.  Mikhail Grabovski is no good because he's essentially Brian Bradley? 

The irony, of course, is that hockey analysis is accused of lacking nuance, while our friend who "watches the games" gives us one comp from 20 years ago without acknowledging that a player's career can go in all different directions.  And it's not exactly a helpful comp - does anyone even remember anything about Brian Bradley? 

I've beaten this dead horse a billion times, and I guess I'm going to do it again.  What we do here is count shots and count scoring chances - if you think you can count those things without watching the games, you're nuts.  If we had more resources, we'd do what NHL teams do, which is count more qualitative micro-stats - forechecks, touches, successful passes, neutral zone counters - you name it. 

Claiming that NHL scouting doesn't involve analysis of micro-stats only shows your own ignorance.  So instead of getting angry at a Michael Lewis book written almost ten years ago, get angry at all those teams that do exactly what you think they don't do.