View Single Post
Old
04-29-2013, 10:35 AM
  #744
Freudian
Clearly deranged
 
Freudian's Avatar
 
Join Date: Jul 2003
Country: Sweden
Posts: 39,781
vCash: 50
Quote:
Originally Posted by ABasin View Post
I believe that trying to use statistics to define an individual player's defensive play is very extremely difficult. Actually, there are so many real-time variables, I pretty much believe it's close to impossible. Thus, my picking which stats are relevant in our discussions. Team defensive play is another matter, re: using stats.

However, I do find it amusing that you give me the "Jesus Christ", then do exactly the same thing I did in regards to these advanced stats. Attend:
Shots is one of the best indicators of how well a team plays because it's more or less not dependent on luck. A team that outshoots the opposition plays better than a team that does get outshot. Not over a single game of course, but the larger the sample is the more reliable it is as a representation of how a team plays.

Of course EJ isn't the only explanation that Avs shoot more and allow less shots when he is playing. But he does draw the toughest opposition and outshoots the opposition. This means other defenders get easier matchups and they can then in turn outshoot the opposition/avoid being outshot. Landeskog and O'Reilly are guys that also helps the team outshooting the opposition so they also deserves credit like EJ does.

Quote:
Not so very different than what I did, is it? You just cited different stats than I did. We just believe different stats have different relevance.

To be clear, while I've given you a bit of crap for leaning so heavily on advanced stats, in truth I don't have a problem with all of them. Mostly just the ones that lead people to sweeping conclusions about individual defensive play.
Completely different to what you did. I explained why stats without a lot of randomness (shots) are better than stats with a lot of randomness to describe what is happening on the ice.

The teams record with or without players isn't representative of a players contribution.

With EJ: 9-19-3
Without EJ: 7-6-4
With Landeskog: 11-20-6
Without Landeskog: 5-5-1
With O'Reilly: 8-17-4
Without O'Reilly: 8-8-3

These with/without streaks obviously overlap so we have a situation where the team does well without these players and much worse with these players. It makes even less sense when it's already been shown that Avs got massive outshot during the winning streak and outshot the opposition during the losing streak.

Which brings us to goaltending. Goaltenders playing well and getting saves they shouldn't gives you wins when outplayed and letting in goals they shouldn't give you losses when you outplay the opposition.

Varly had a save percentage of 91.4% first half of the season (when EJ, Landeskog and O'Reilly didn't play) and a save percentage of 88.4% the second half of the season (when EJ, Landeskog and O'Reilly played). Suddenly it's not that strange that the team loses when outshooting the opposition and winning when outshot, is it? The players in the first half get good GA/60. They get good on ice save percentage. The players in the second half get much worse on ice save percentage and GA/60. Most of it is due to how well the goaltender plays and not that the skaters suddenly allow much worse scoring chances for the other team.

Shots is one of the least random stats in the game. You don't shoot out of luck. Goals and saves have a bigger random effects. McGinn hitting 8 posts in his first ten games was him getting unlucky and he wouldn't have been more skilled in any significant way should five of them gone in. At the same time the opposing goalie wouldn't have played significantly worse had those gone in or the defenders in front of the goalie played worse. There really is such a thing as shooters luck and goalies luck that have a huge impact on results in the short run and that's why they aren't suited for stat analysis.

I realize that when showered in advanced stats it's easy to just grab one and think it has great explanatory power. Most of them doesn't. It's either a stat being affected by luck and randomness or it's an event that happens so rarely that it's not reliable over a season. Most of the stats on behindthenet are more or less useless for providing any insight.

There are a few stats that we get a critical mass of that is large and aren't all that random. They are very useful to help us describe what is happening.

Zone starts is a fantastic stat. It tells us how coaches use a player, especially in combination with Quality of Competition (which isn't a a great stat in itself, only when used with other stats to add context).

Zone finishes is a fantastic start. It's an event that happens so often (around 1500 times a season) that signifies how a player is doing possession wise. Most of the random stuff that have a big impact in the short run isn't a factor in the long run. If you finish in the offensive zone more often than the defensive zone over a season, that's a sign that you're doing well. For example all of us intuitively know that Barrie has been a great contributor for Avs when it comes to transition, possession and creating offense. It's no surprise that his offensive zone finishes is much higher than his offensive zone starts. He affects the flow of the play in a positive direction for Avs.

Shots is the same. We get such a critical mass over over a season that it does have correlation with how well a team is playing. Avs outshooting the opposition is much better for Avs than the opposition outshooting Avs.

Those three stats are pretty much all I care about from the advanced stats. Actually none of them is an advanced stat. They're very straight forward. They have a higher reliability and explanatory power than many of the others because they are much less affected by randomness and luck. It's so easy to dive in and stat trawl, looking for stats that confirm what you want to be confirmed. But if you don't know which are junk stats and which ones are useful stats you end up wrong.


Last edited by Freudian: 04-29-2013 at 10:45 AM.
Freudian is offline