Saturday, January 16, 2010

Quality of Teammates

A lot of people who read here are familiar with the notion that the context of a player's ice time has a huge impact on their counting stats and how good they look on the ice. By context, largely we're talking about who they are playing with, and the quality of opponent that they are generally playing against. Also, where their shifts are starting more, be it by faceoff location or on the fly. And lastly, whether or not they are being run out against opponents with tired legs. Those are the principle elements, anyways. And they aren't always easy to pin down with numbers.

Nobody has done more to popularize these ideas than Gabe Desjardins at his stats site. The faceoff zone data can't be argued with, and the quality of competition information always seemed reasonable, given enough games. The quality of teammate data, however, always seemed veryy dubious to me.

That's changed, Gabe has a new quality of teammate stat that is based on a player's underlying stats, specifically his Corsi number. The results seem completely sensible. They are presented in the chart below. Click to enlarge.

The old QUALTEAM numbers seemed madass to me, but I can't argue with these, they mesh with what I've seen in the games. Your mileage may vary.

6 Comments:

Blogger Vic Ferrari said...

By the by. If anyone has:

A.) a list of EV strengh scoring chance % for the Oilers, and

B.) a table of on-ice times together (like the bottom table generated by http://timeonice.com/H2H0910.html?GameNumber=20568&submit=Go), except summed up for all the games that we have scoring chances for.

Drop both into a spreadsheet and multiply them using the array mulitply function on your spreadsheet, that's =MMULT() in Excel.

Voila ... you'll get a list that's strikingly similar to new Qualteam Stat from behindthenet. This because the world is round, and hockey has a whole helluva lot less magic in it than we've been lead to believe.

You might be able to get "B" from David Johnson or Gabe Desjardins, if you have the list of NHL game numbers, and you ask nicley.

1/16/2010 1:55 pm  
Blogger R O said...

Related: the new QOC numbers appear completely FUBAR for the Flames. The orderings between GF/GA QOC and Corsi QOC are off by a lot, and (I hate to say it) the GF/GA QOC jives better with what I've seen. Sutter's line matching strategy changes with the wind but by my eye he's put up Iginla against the best (nonwithstanding circumstances) in about half the games.

So I don't think there's any way in hell he's 6th on the Flames forward depth chart in that regard, and even if I'm wearing the Flames goggles re: Iginla, it doesn't jive with his faceoff starting location where he separates himself from his linemates.

Of course for the Canucks (whom I follow a bit) the ordering is sane.

This kind of annoys me since I had originally asked Gabe to do this (although Sunny asking was the final catalyst), thinking it would be more reliable.

In any case I have abandoned my previous total reliance on these QoC/QoT metrics, I prefer just to let them guide my thinking anyway. ZoneStart is more compelling, less of a moving target.

1/16/2010 5:48 pm  
Blogger Vic Ferrari said...

I like the original Q of Comp as well. If guys are a mile apart, it always seems to show up there. If six guys are fairly close, though ... using rank seems a little wild.

A mix of the two probably is fairest.

Playing against the Sedins ... both systems catch it.

Playing against Thornton/Heatley ... only the original qualcomp catches it.

Bear in mind it's taking into account the soft minutes too. I doubt that many would argue that Pronger played tougher minutes than Jason Smith when they were both Oilers. You could go through shift charts and, more often than not, you'd see that was the case. Pronger also played a lot more against the other team's worst players too (if the Oilers were trailing and needed a goal, or if the 4th line on the other team iced the puck ... that was Pronger time too.) Smith never got that icetime. And that's prime icetime for point getting and +/- boosting.

So Pronger played more against the other team's best, but Smith probably played just as tough icetime on the whole ... if we consider the effect we would expect on his results, which AFAIK is what Gabe is trying to give an indication of.

1/16/2010 7:48 pm  
Blogger Olivier said...

R O and Vic: I've wondered over the last few months what a QualC/QualT based on ice time would look like. Obviously, you need to differentiate between forwards and Defensemen icetime, but still.

It's always the same challenge tough: how to build the darn database :)

1/17/2010 12:33 pm  
Blogger Kent W. said...

Acutally, RO, I think the new QoC jives with my impressions of the team. I think Sutter vacillates between tough comp and soft comp for Iginla, sometimes game to game but often shift-to-shift: for example, Jarome is usually the first guy over the boards if the opposition ices the puck with their slubs on the ice. Which is a sensible strategy. He's also the first guy on after the penalty kill.

Langkow et al, however, almost never have it easy. Either they're taking on the tough competition head-to-head or they're faced against whatever secondary threats the opposition possesses.

1/18/2010 2:31 pm  
Blogger Sunny Mehta said...

I think they're both useful for exactly what they're supposed to be useful for. If you wanna look at QoC based on goal scoring, you know there's gonna be a bit more randomness/goaltending soaked in there than with Corsi. If you look at QoC based on Corsi, you know there's gonna be a little bit of injustice done for the Sh% outliers as well as the guys who got shitty or good assignments based on matchups/faceoffs.

1/19/2010 12:32 pm  

Post a Comment

<< Home