Statistics that indicate defensive prowess

[TheKingSlayer]

Hey all,

I am looking to put together a rating set for a SIM league, and suffice to say there has been some dispute over what stats we should use to create the defense rating.

Past rating sets have used stats like +/- and this does not appeal to me.

I am thinking a long the lines of short handed TOI, blocked shots, takeaways/giveaways, maybe defensive zone starts vs offensive zone starts.

Others have suggested that perhaps we include hits as a portion of defense because a good hit can eliminate the player from the play therefore reducing the potential threat.

What stats would you use, and what percentage of the whole would you attribute to the stats (i.e out of 100%).

Click to read more...

 

[LeafShark]

The closest thing I've come up with was +/- per game adjusted for team save percentage. It's not perfect, but it's the easiest solution. Might not be the best. Real time stats are not a good statistical indicator for anything really. Only other things you can really look at are Shot-Dif, Fenwick, or Corsi, or time the amount of possession across the blue line for each team. In general the NHL is way behind when it keeps track of stats.

 

[Michael Farkas]

Best advice, is base as much as you can on coach's decisions. It's a coach's game now, everything is systems and structure. Coach's essentially decide statistics (especially for goaltenders) today. If a coach puts a player in a situation to defend a lot (whatever metrics we have to measure that), that's the guy with the best defensive attributes.

Re: hits. Careful. Check home/road splits.

 

[Verviticus]

im pretty sure road hits/blocked shots (the only fair way to measure) are both negatively correlated with goals against. both of those are things you do to get the puck - therefore, if you always have the puck, you can't do them

can you incorporate possession stats and toughness of minutes in some way from http://stats.hockeyanalysis.com/players.php ??

also, absolutely include zone starts

 

[EVBetting Site]

Usually a combination of stats taken with context and a dash of judgement can paint a general picture of a players game.

Take a player like Paul Gaustad this year.

Starts 21% of his ES shifts in the defensive zone (as lopsided as there has ever been in the NHL since the stat was recorded). He plays against average competition with around the same level of teammates. Despite his zone starts, he manages a Relative Corsi of just -2.0, which is very respectable.

His GA/60 is merely 1.45, which is absurdly low given his zone starts. He also chips in with a GF/60 of 1.45, meaning he breaks even (give or take a lot of variance due to a small sample size) with his matchup over the long run. Add in that he is used on PK as well, usually first unit (highest TOI/60 among forwards) and manages a respectable 6.34 GA/60

Given how Trotz uses him, being able to break even given a 21% zone start and similar competition, Gaustad is essentially a get-out-of-jail-free card for his coaches.

He is a major reason Buffalo is struggling badly this year, especially defensively (3.2 GA/game compared to ~2.8 GA/game last year and ~2.8 GA/game the year before that as well). His injury is also costing Nashville this year, as their GA/game without him this year is 3.5GA/game, compared to ~2.0 GA/game with him in the lineup. Quite the difference (sample size is a factor)

Nashville gave up a first round pick to get him. Buffalo wouldnt let him go for less. Nashville then signed him to their 4th highest cap hit, longest term for a forward extension. Shows you what kind of value a player like this has. Almost every fan, though, considers Gaustad to be a 30-point player or an OK third-fourth line tweener when, in reality, this is a "franchise" caliber bottom 6 player.

From the stats, you can determine that this is a very valuable player, and when you watch him to figure out the how and why, you see it's because he's got size and he's a decent skater and he's just a very defensively aware player.

 

[Mantha Poodoo]

First off, stats like hits, shots blocked, and takeaways are useless on their own. These are indicative of style but not overall effectiveness. Some players rack up massive hit totals, often on the forecheck, but are lackluster to bad in the neutral/d-zone. Shots blocked can be just as indicative of bad performance leading up to lots of plays where shot blocking is necessity, or poor skating/other defensive skills with blocking as a necessity to make up for such. Takeaways are a little more directly valuable, as removing the puck from the opponent's possession is always a good thing defensively, but situation and resulting play are important to look at (If you take the puck away, but then immediately turn it over to the opposition because you were in a bad position to perform a takeaway where you should have went with a more passive defense, and this results in a goal against, was it a defensively good takeaway?).

+/- is worthless on it's own. +/- adjusted for save percentage is a bit better, but still doesn't take into account the quality of teammates or opposition. I'm more impressed by a guy with an adjusted +/- of 2 that plays with 3rd/4th lines vs top line opposition than the sheltered 3rd pairing d-man that has an adjusted +/- of 10 playing largely against the other team's checking lines.

Fenwick and Corsi both have a lot more to offer in terms of analyzing the impact a player has on the ice, but you still have the savvy to be able to understand just what all of the substats together are saying and most importantly be able to correlate what you see in the numbers with what you see from watching the performance.

Usually, Fenwick and Corsi are what I use when I need to get a rough idea of how a player performs with whom, against whom, and in specific situations.

Some descriptions of Corsi stats:

http://www.broadstreethockey.com/pages/bsh-advanced-stats-glossary

http://www.behindthenet.ca/stats_faq.php

And some Corsi stats you can look at:

http://www.behindthenet.ca/nhl_statistics.php

 

[Micklebot]

Sportsnet posted some Defensive zone stats that focus on touches and zone exits the other day. The accompanying article can be found here

The full list can be found here

I don't think it's a great metric for defensive accumen, but the data is nonetheless an interesting look into how defensmen perform at some specific aspects of DZ play, for example, Karlsson is the top at successfull exit %. It's also interesting to see who ices the puck more often (lovejoy ices it a ton). It also has turnover %, Ranger, Ballard, Harrisson and M Weber top that lovely metric.

 

[Caeldan]

Hmm.
I have an inkling of an idea for a zone start/zone exit ratio... but I can also see some problems with it too.

 

[wgknestrick]

Sportsnet posted some Defensive zone stats that focus on touches and zone exits the other day. The accompanying article can be found here

The full list can be found here

I don't think it's a great metric for defensive accumen, but the data is nonetheless an interesting look into how defensmen perform at some specific aspects of DZ play, for example, Karlsson is the top at successfull exit %. It's also interesting to see who ices the puck more often (lovejoy ices it a ton). It also has turnover %, Ranger, Ballard, Harrisson and M Weber top that lovely metric.

That data is fantastic. Don't fool yourself. If this stuff was standard stat fair, we'd wouldn't be on such a quest to "quantify defensive play". We need multiple seasons of it though..

Exiting the zone (with possession) is probably the most important defensive stat IMO. Low % Icings are very valuable too. The great "idea" for most to learn is that generally, the "offensive" minded defenseman are the best at "defense" (5v5 GA/60). The reason is that they rarely play "defense" in their own zone. Your "shot blockers", and "crease clearers" generally have little value without some skill to handle/move the puck.

All that needs to be done is run some correlation tests on that data to GA/60, GF%, CF%, etc. I am sure something will be found.

 

[Micklebot]

That data is fantastic. Don't fool yourself. If this stuff was standard stat fair, we'd wouldn't be on such a quest to "quantify defensive play". We need multiple seasons of it though..

Exiting the zone (with possession) is probably the most important defensive stat IMO. Low % Icings are very valuable too. The great "idea" for most to learn is that generally, the "offensive" minded defenseman are the best at "defense" (5v5 GA/60). The reason is that they rarely play "defense" in their own zone. Your "shot blockers", and "crease clearers" generally have little value without some skill to handle/move the puck.

All that needs to be done is run some correlation tests on that data to GA/60, GF%, CF%, etc. I am sure something will be found.

All the data is coming from Corey Sznajder's ambitious Zone entry/Exit project, and there have been a number of blogs popping up with entries crediting him for the data. Another one just showed up on the score that focuses specifically on Karlsson.

Really hoping that Sportvu will automate the process as it really is labour intensive right now. There is also a Sens Blogger that tracked all zone entries for every Sens game.

I really do believe zone exit efficiency, and prevention of zone entry are two areas that are often overlooked when evaluating defensive play.

 

[ChiGuySez]

Blackhawks 6 dmen, lowest is Leddy at 42.

 

[hairylikebear]

Corsi is a team statistic. 80% of this data is completely out of the players' control. How can individual aptitude possibly be measured this way?

 

[Micklebot]

Corsi is a team statistic. 80% of this data is completely out of the players' control. How can individual aptitude possibly be measured this way?

Yeah, not really impressed by the touches/CA stat, however, I do find the breakout%, Turnover% and icing% very interesting. I'd also be interested to see a touches/60mins rather than just a raw touches (sounds dirty) stat.

 

[wgknestrick]

Corsi is a team statistic. 80% of this data is completely out of the players' control. How can individual aptitude possibly be measured this way?

Does a car handle differently if 1 of the 4 people sitting in it is extremely overweight?
Can you tell if I put double the salt in a recipe of 5 ingredients?

You can still spot trends from individuals in the group. This is how we develop. Just because something may not be perfect, does not mean it has no value. This is probably the best data there is on defensive play.

 

[hatterson]

Corsi is a team statistic. 80% of this data is completely out of the players' control. How can individual aptitude possibly be measured this way?

That's generally why you look at corsi rel. Although that's limited by how much the teams lines move around. If you have a dman who always played with the same partner or a forward who was always on the same line with 2 other guys, it's hard to tease out their contributions from the players they're frequently with.

Clearly if you play with an elite defensive player your defensive numbers (whatever ones you choose to use) will be improved. The hope is that you also spend enough time away from that player that you can see the difference in numbers between when you play with him and when you play without him (and the same for him)