OT: New Stat? Opinions please

[Neiler]

Hey guys, looking for some input on this.

I've been pondering a measurement for how players impact the team but wanted one that would include everyone - stars, grinders, call-ups, goalies, etc - and this is what I came up with.

Perhaps this has been done before, perhaps not which is one reason why your input would be appreciated. I'd also like to know how you feel about the stat itself... is it a fair measuring tool, too obscure, etc.

What I've done is taken the last 40GP for Ottawa and compared it against only the games that players have played in. The statistic I chose to look at was goal differential as I feel it's one that every team member has an impact on.

Essentially it comes down to: [GD only for games player has played in] - [GD for the team, last 40 GP] = Team impact (TI)

For example, Ottawa's GD for last 40 is +0.250 per game. The team GD for the 20 of those 40 games Karlsson played in is actually +0.350. Therefore, Karlsson's TI is +0.100.

Attached is a screenshot of what I put together ranked from highest to lowest. Let me know if it makes sense or not... sometimes I think out in left field somewhere. If this works I would use a rolling 40GP as a measurement so the stat would change daily for most players. Also of note - players who play all games come out even which I think is great... they are the team if they're playing every game.

Click to read more...

 

[Benjamin]

GD?.

 

[Neiler]

goal differential... GF - GA

 

[GimmeMyJetpack]

GD?.

Goal Differential.


I'm not sure how this makes sense? Turris Greening Neil all have 0? Are you sure you did this properly?

In my mind shot differential would probably provide a better analysis. Did you do this only at 5v5?

 

[Neiler]

no, it's by games played. not broken down further.

The reason those guys have zero is because they've played every game in the last 40 so there's no measure of what happens when they're separated. Essentially, their stats are equal to the team's.

I'm not sure I'm looking at it the right way though.

 

[Benjamin]

http://www.behindthenet.ca/nhl_stat...5=OTT&c=0+1+3+5+4+6+7+21+22+23+24+25+26+27+28

 

[moz]

no, it's by games played. not broken down further.

The reason those guys have zero is because they've played every game in the last 40 so there's no measure of what happens when they're separated. Essentially, their stats are equal to the team's.

I'm not sure I'm looking at it the right way though.

I'm far from a statistician, but if Turris and Greening are regarded as equal by this metric, is it a useful metric?

 

[Neiler]

I'm far from a statistician, but if Turris and Greening are regarded as equal by this metric, is it a useful metric?

Excellent point.

 

[Skrymir]

The problem with this metric is if a player plays all 40 games they have by default 0.000 TI. Not very useful IMO.

 

[CanadianHockey]

As others have said, you can only quantify a contribution if the player is newly acquired, was scratched, a call-up, or has been recently injured.

 

[moz]

As others have said, you can only quantify a contribution if the player is newly acquired, was scratched, a call-up, or has been recently injured.

And then the problem becomes that the sample size is so small it hardly holds value.

I like what you are trying to do and encourage you to keep working on a metric, but this one as it stands doesn't appear to be the one.

 

[SenzZen]

I'm far from a statistician, but if Turris and Greening are regarded as equal by this metric, is it a useful metric?

Yes.

That's Da Costa and Corvo's fault.

 

[Laoghaire]

Two things are missing. A time on ice variable and I believe shots (or like Corsi does, shots directed at the net, though I'm not sure how I feel about that), would be a better metric than goals.

 

[CanadianHockey]

Number of ES shifts might be a better metric than GP, since you've got substantially more data to work with.

 

[Neiler]

Thanks for feedback everyone... sometimes you stare at something yourself and it makes sense but more eyes reveal the flaws.

 

[Skrymir]

Thanks for feedback everyone... sometimes you stare at something yourself and it makes sense but more eyes reveal the flaws.

I'm all for new metrics... I'm not sold on corsi either. I think you need to normalize the data with something other than GP, TOI or shifts should do the trick. The GF/g vs. GA/g is good for a macro comparison but may diminish good players playing on bad teams. You may want to look at other stats to compare.

 

[Countdown0]

The biggest problem I see with this kind of statistic is that it wont work with players who have played in all of the last 40 games. Players like Spezza, who get hurt frequently, can be shown to help or hinder the team when they are in the line-up. But someone who is always playing, and never nursing injuries or scratched, is basically immune to this kind of stat.

Its a good idea, but it only applies to players who actually miss some games.