Statistics: Odd-man rushes/breakaways NHL

[Gormo]

Someone has to be keeping track of this somewhere in terms of numbers of opportunities and scoring percentage right?

I expected this would be easy to find, but I need some guidance.

Thanks very much.

Click to read more...

 

[eklunds source]

I don't know that anyone has kept track of these events. A lot of people have done work tracking scoring chances, shot locations versus expected goals, shot quality... and all of them come back with the same result: they correlate very strongly with shot totals, and they aren't reproducible. Thus, spending time tracking them is not a worthwhile investment of time.

How Something as Important as Shot Quality is not that Important
Shot Quality Revisted
Luck vs Shot Quality
Stop looking for shot quality wizards

On that basis, even if you found a resource that tracked these, I don't think it would provide any "new" information..

 

[hatterson]

I don't know that anyone has kept track of these events. A lot of people have done work tracking scoring chances, shot locations versus expected goals, shot quality... and all of them come back with the same result: they correlate very strongly with shot totals, and they aren't reproducible. Thus, spending time tracking them is not a worthwhile investment of time.

http://www.sportsnet.ca/hockey/nhl/introducing-the-shot-quality-project/

That seems to imply otherwise

I haven't seen him post any study of the sustainability of the results but I certainly don't get the indication that his results suggest it's not worth spending time on.

I'd say it implied quite the opposite. Either shot quality isn't reproducible and thus can help you better identify which teams are expected to regress/improve in terms of shooting or save percentage, or it is reproducible and must be factored in to analysis.

Pittsburgh has been top 5 in 5v5 shooting percentage every year from 07-08 until now with the exception of 10-11 where Crosby and Malkin both missed half the year and Tyler Kennedy was their #4 scorer. Since 10-11 the Leafs have been top 7 in shooting percentage every year.

The Pens have frequently outperformed their corsi numbers. The Leafs (since stabilizing their goaltending) have also outperformed their corsi numbers. Several teams have frequently underperformed theirs over a long stretch.

 

[eklunds source]

Either shot quality isn't reproducible and thus can help you better identify which teams are expected to regress/improve in terms of shooting or save percentage, or it is reproducible and must be factored in to analysis.

I'm not definitively saying it is or is not reproducible, but that the effect is so small it's not really worth considering. At the team level, after 82 games, 2/3rds of 5v5 sh% is chance.

Even Strength Shooting Percentage
Team Even Strength Shooting Talent

Pittsburgh has been top 5 in 5v5 shooting percentage every year from 07-08 until now with the exception of 10-11 where Crosby and Malkin both missed half the year and Tyler Kennedy was their #4 scorer. Since 10-11 the Leafs have been top 7 in shooting percentage every year.

I'm not surprised Pittsburgh has outperformed 5v5 sh%, seeing as 5v5 sh% is relatively persistent at the individual level (over very large samples), and Pittsburgh has 2 of the best over the past 8 seasons in Crosby/Malkin.

The largest gap Toronto has had between actual sh% and league average sh%, since 2010, is 0.706% over average. That gap is really quite small - 61 teams have deviated from league average more than that out of 180 team-seasons.

The Pens have frequently outperformed their corsi numbers. The Leafs (since stabilizing their goaltending) have also outperformed their corsi numbers. Several teams have frequently underperformed theirs over a long stretch.

Not sure what Corsi has to do with it. AFAIK there's no correlation between shot differential and shooting percentage.

 

[hatterson]

I'm not definitively saying it is or is not reproducible, but that the effect is so small it's not really worth considering. At the team level, after 82 games, 2/3rds of 5v5 sh% is chance.

Even Strength Shooting Percentage
Team Even Strength Shooting Talent

I'd say that 1/3rd of the variance of something is significant enough to not just sweep under the rug

I'm not surprised Pittsburgh has outperformed 5v5 sh%, seeing as 5v5 sh% is relatively persistent at the individual level (over very large samples), and Pittsburgh has 2 of the best over the past 8 seasons in Crosby/Malkin.

The largest gap Toronto has had between actual sh% and league average sh%, since 2010, is 0.706% over average. That gap is really quite small - 61 teams have deviated from league average more than that out of 180 team-seasons.

The gap for Toronto hasn't been that large, but it has been consistent, which I think hints that it's worth looking into deeper.

Not sure what Corsi has to do with it. AFAIK there's no correlation between shot differential and shooting percentage.

I brought up Corsi because if shot quality is irrelevant and random then corsi is the best long term predictor (or fenwick). However if shot quality exists (either on a team system level, or an individual talent level) then we must consider more than raw shot +/-

And to me, shot quality means more than just where a shot came from. It also (ideally) includes data regarding what was happening (like Chris Boyle compiled) but also data on who was doing the shooting. Stamkos taking a shot is a lot more dangerous than John Scott taking the shot. Even a .7% expected difference in shooting percentage (either from team system or talent level) means a difference of 15-20 goals scored per year which can take a team from average to well above average. That's equivalent to 5-7 expected points over the course of the year