Best Stat to Determine a Player's Defensive Ability (Corsi, Fenwick, etc.)

[rodney dangerfield]

Title Speaks for itself. What stat best determines who is a good defensive player? Is it corsi? Whats the difference between CF% and CF60? Or is it SA60? Or Fenwick?

 

[Michael Farkas]

None the above.

I don't have a good answer really though. But I know that none of what you're listing, all of which involve shots for, except SA/60, which involves shots against (also, not the meat of the issue) which are not indicative of defensive play.

 

[ResilientBeast]

Once player tracking becomes a thing, I'd say TOI in defensive zone vs shot attempts would be a good start especially if you could add in zone starts.

 

[Trap Jesus]

Expected goals against and relative expected goals against seems to be the best idea and the best catch-all for defensive play, although it's a little unclear to me how it gets measured.

 

[fiveonfive]

Expected goals against and relative expected goals against seems to be the best idea and the best catch-all for defensive play, although it's a little unclear to me how it gets measured.

Two things:

1) Not all expected goals are the same. Different sources use slightly different methods.

2) Usually expected goals are like corsi, accept each shot attempt is adjusted to the shot distance, shot angle, shot type, proximity to other shots (to adjust for rebounds), ext... where Again, it depends on who's expected goals figures you are looking at. The factor of adjustment is based on the historical likelihood of similar shots ending up in the net.

If you are interested in a more detailed explanation, I suggest searching for the methodology of xG calculation within the source you are using.

To OPs question. Looking at score-adjusted (and zone) CA/60, FA/60, SA/60, xGA/60 and the rels for all 4 usually draws a pretty good picture of defensive impact. The score adjustment is a big key for those numbers to be normalized so that players can be compared. However, if you are looking for a WAR like number - there isn't really one that is currently accepted as the go to. If you are interested though there are some people online who have been developing their own GAR and WAR numbers ( goals above replacement and wins above replacement). @DTMAboutHeart has one on hockey graphs that you can look at:

https://hockey-graphs.com/2016/10/24/a-primer-on-dtmabouthearts-war-model/

 

[oilerbear]

Once player tracking becomes a thing, I'd say TOI in defensive zone vs shot attempts would be a good start especially if you could add in zone starts.

The whole basis for My High danger- Cummulative goal (what people are calling expected) theory presented at Lowetides blog in 2006.

to get rid of the stupid notions:
1. all shots are equal.
the range of bottom success low danger shots and most successful high danger shots is greater than 2000%. a high danger shot can be 20= times more successful than low danger shots.
the avg High Danger shot is 5.1 times more successful than the avg low danger shot.
2. Dmen had no affect on GA.
A study was done and found that Dman affect was +/- 1.25%
but the individual failed to realize that variance in save% .900 to .935
a range of 3.5% so +/11.25 which is 2.5% is a affect of 71.5%

3. offence from Dmen serves positive Goal diff value.
95% of goal diff production comes from forwards.
Dman have to abandon Defence of HD area to attack the opposition HD area.

4. Corsi against was a measure of dman play.
It is a function of turnovers from the attacking of oppositions net by the top offensive players. 95% which are forwards.
Forwards yield a counter attack rate.

5. every Data value needs an x,y (expected affect) value associated with it.
Corsi x,y; Fenwick x,y; Shot x,y:

Shot x,y that I created years ago is what Expected goals is from.

the video of Pit cycling the puck around for many minutes was the talk of the boards.

But I stated Bobby Clarke would have punched the pit forwards in the face if they had been teammates.
Cause they got zero high danger shots.
you have to have the stones to go into the high danger area.

PiT were perimeter coward on that whole cycle .

lots of D zone time.
Zero offensive value.

it was great high danger defence.
but
a whole collection of people with zero understanding of scoring thinking it was poor play.

Quite honestly though.
in my High danger theory.
it is a move to the final step of Data analysis.

Until the data is broken up into open and closed goal potential,
open goal potential (X% chance of goal based on x,y location of net elevation)
closed goal potential (0% chance of goal) hits goalie (not saved)
All measures are wrong!

Relative adjustments are a joke!

 

[OvermanKingGainer]

Title Speaks for itself. What stat best determines who is a good defensive player? Is it corsi? Whats the difference between CF% and CF60? Or is it SA60? Or Fenwick?

The best stats for defensive ability would be measurements of

- Blocks per shot attempt against
- puck retrievals per minute of ice time (in any zone, BTW)
- successful clean zone exits per shot attempt against
- successful zone entry denials per minute of ice time
- successful takeaways in the defensive slot area or neutral zone per minute of ice time
- successful stick lifts and rebound clears in the slot area

Since those aren't all ready at our disposal, the closest thing to these stats to look at each of

CA60
FA60.Rel
xGA60.Rel
xGA60
GA60.Rel

Which generally display the results of the first six things. They need to be supplemented with context - QoT, QoC, ZSR, and even simply who the off-ice teammates are.

Don't even bother with SA60/SA60.Rel, as you can't control whether a shot makes or misses, or raw GA60, as you can't control a goalie's raw ability level.

 

[ImporterExporter]

Watch the games. You have to actually watch the best play, because there aren't metrics that can give you the entire story. I'd wager 3/4 of the writers who vote on the Selke (and all awards) pretty much just look at +/-, reputation from years past, and team record.

It's sad, but the majority of folks don't watch enough hockey from multiple quality defensive forwards. Some of it is just impossible given you could have 10 games in a single night, many of which overlap. You can't get to them all.

 

[Doctor No]

Your post implies that watching the games can "give you the entire story".

It can't. Aside from the reasons that you correctly mention, your brain isn't good enough to catalog everything without introducing bias.

 

[Cursed Lemon]

Someone will eventually come up with some kind of aggregate stat that weighs in a myriad of defensive metrics. We'll all argue over it while simultaneously quoting it to support our viewpoints.

 

[Nalyd Psycho]

I would say that opponents changes in corsi and fenwick would be the best indicator.

 

[Dertell]

Expected goals against and relative expected goals against seems to be the best idea and the best catch-all for defensive play, although it's a little unclear to me how it gets measured.

The problem with xGA60 is that relSv% doesn't repeat very much, so it's fair to assume it's less about quality shots and more about shot against per minutes spent in the defensive zone (or something along the lines of it).

 

[Filthy Dangles]

What even is 'defensive ability' or 'being good defensively'? I think the best definition would be 'how impactful a player is at preventing goals (against).' That's a results oriented view, though. What exactly does a player 'do defensively' on the ice that allows him/his team to prevent those goals?

Take baseball for example. That question is a lot easier to define. A good defender in baseball can get to more batted balls vs a more average player at his position can (by virtue of athleticism, ability to read the ball of the bat, first step, route effeiciency etc.) and thus will get his team more outs and save more runs over a season. Even still, baseball defense stats are a long ways away and hockey is light years behind baseball statistically.

In the end, you win hockey games by outscoring the opposition not 'preventing more goals' than the opposition. With current advanced stats, the conclusion is that if you regularly drive possession for your team and are a good possession player, your team should outscore the other team(s) over the long haul when that player is out there and thus he'll positively contribute towards winning hockey games.

Sort of like 'the best defense is the best offense'. Or for another analogy, like managing the clock in football. You are preventing the other the other team from scoring simply by having possession of the ball; the other team clearly can't score in this situation.

 

[Henkka]

Title Speaks for itself. What stat best determines who is a good defensive player? Is it corsi? Whats the difference between CF% and CF60? Or is it SA60? Or Fenwick?

It's the Eye-tested Scoring chance data. You look the game and make notes who will get Quality Scoring chances against and who won't. Who'll prevent's those chances and puts opposite team shoot from bad angles = Goalie can cherrypick. That kind of Defenceman can lose the shot distribution, when he wins the percentages on danger areas. And put some Quality of Competition on the mix, if this guy can prevent Quality Chances against opposite TOP players, he is a damn good defender.