Why 5v5?

[theranfordflop]

I've noticed that in corsi analysis, people tend to use even strength numbers to evaluate performance. I think this makes sense if your dataset includes players that don't get a decently equivalent amount of special teams time. However, if you're only looking at players that all have an equivalent special teams role (eg. first PP unit), is it best to include those situations in your data? Or is it considered poor practice in all cases?

I ask because I rarely see analysis that uses all situations, even when I think they should.

Thanks!

Click to read more...

 

[doncherrysgrill]

Just because it's the standard/most prominent game type. Including someone's special teams numbers can heavily skew the stats in both directions. There's obviously value to special teams stats but they are counted separate for that reason.

 

[gorangers0525]

You're not going to find many players with a good corsi on the pk and bad on the powerplay lol. Special team's hockey is just a completely different game with a lot less sample size.

 

[theranfordflop]

You're not going to find many players with a good corsi on the pk and bad on the powerplay lol. Special team's hockey is just a completely different game with a lot less sample size.

Yes, but two players with equivalent PP and SH usage would both be effected by that skew correct? I understand if I'm looking at a big dataset of players of all roles, that can be an issue.

But, if I'm looking at say 5 top line players, doesn't it make sense to include their PP time in the evaluation?

 

[Doctor No]

Ultimately, it depends upon what you're trying to do (and depending on the goal, I've seen both things you've described as well as others).

So what's your goal? Estimate a player's value on another team? Compare the values of players A and B? Estimate a player's value in a different situation? Something else?

 

[wgknestrick]

Special team results vary highly (or much more than 5v5) from year to year.

5v5 /60min is the equal situation that normalizes as much as possible and tries to remove coaching bias between players and their ice time.

All players are a sum of their

5v5 generally 80-90% of their icetime and
4v5 10-20% or
5v4 10-20%

contributions. Good 5v5 players are generally good special team players too. Once again, 5v5 is the most important aspect IMO in evaluating players. I think it good practice to listen to the results players get in 80% of their ice time as a primary indicator, and use special teams to break ties, or add context.

There are only so many special team spots to go around, and you have to ensure this new player will even get a significant opportunity on special teams to even utilize their results. Penalty occurrence also randomizes their opportunity for special team contributions too.