Time on ice competition plots for all 30 teams
Eric T.
August 17 2012 07:26AM
Duncan Keith had a negative Corsi Rel last year, but he's not weak competition
By Matt Boulton from Vancouver, Canada (Kane, Keith and Kopecky) [CC-BY-SA-2.0 (http://creativecommons.org/licenses/by-sa/2.0)], via Wikimedia Commons
Yesterday, we published an article about evaluating quality of competition by looking at the opponents' average time on ice. The best players typically see the most ice time, so a player whose opponents get a lot of minutes is probably facing tough competition. By this metric, top-line forwards appeared to face tougher competition than is suggested by other metrics.
The reason for this became clear when we separated out the quality of forwards and defensemen that a player faced. Even the offense-first forwards who are used against mediocre competition see their opponents' best defensemen. This opened up the interesting possibility of using quality of competition to evaluate not just the strength of the competition, but the type of situations -- facing good forwards and bad defensemen might be similar in difficulty to facing bad forwards and good defensemen, but it is a very different type of usage.
We showed a couple of examples, one from a team that matched their top forwards with the opposition's top forwards, and one from a team that had an offense-first scoring line and a defensive-minded shutdown line. A number of people inquired about their favorite team, so we decided to publish plots for each of the 30 teams for 2011-12.
Players who faced the strongest competition will be found in the top-right of a plot, while players who faced the weakest competition will be found in the bottom-left. Defensive specialists will be found in the bottom-right, and offensive specialists will be found in the top-left.
Click on any plot to get a larger image.
Recently by Eric T.
- A competition metric based on ice time
- Almost famous: Lessons from a near miss
- Projections for Rick Nash on the Rangers
- The importance of quality of competition
- Quantifying players' impact on teammate shooting percentage
- More on the advantages of puck possession over dump and chase
It seems there is a lot of defense matching happening but we can also clearly identify what teams don't seem to be using a line matching system (limited horizontal spread) and which teams are (significant horizontal spread).
Teams that clearly aren't line matching are: Islanders, Ottawa, Tampa, Toronto. To a lesser extent Philadelphia, Washington, Winnipeg.
What is interesting about that is they are all eastern conference teams. The only eastern conference teams that seem to line match in a significant way are Boston, Florida, Montreal, New Jersey, and the Rangers to some extent. Interestingly four of those 5 teams ranked among the top 5 eastern conference teams in save percentage and the other went to the Stanley Cup finals. Coincidence? Maybe. Maybe not. Line matching is more of a defensive tactic.
@David Johnson
I'd be cautious about the conclusion that limited horizontal spread means no line matching. Some other possible interpretations:
1) They shuffled lines through the year and most players ended up with some time in each role.
2) They line match aggressively, but the opponents aim for the opposite matchups, and the average of home and road games ends up neutral (this is why looking at home/road splits is on my to-do list).
3) Perhaps the western conference teams happen to have a larger TOI difference between their first and third lines than the eastern conference teams do? This is easily checked (and probably not the case, I'd guess), but would lead to the observed result if true.
@Eric T.
Yeah, it could be. I know at least in the Leafs case it is probably true (don't think Ron Wilson was a line matcher in general). Home/road splits would definitely be interesting.
@David Johnson
Actually, thinking about it more...
A team that ignored matchups probably would *not* end up with their guys in a vertical line. The opponents would get to control the matchups, and since most teams are going power vs power, your top line would end up seeing a lot of the opposing top lines.
I think the indication of line matching would not be having your guys in a vertical line, but having your guys deviate from whatever the average arrangement is. Since the opponent's desire to line match will pull you towards their average arrangement.
Fantastic work, Eric T.!
I think it's a very useful stat as it is, however, maybe using EV TOI would generate more of a horizontal spread for the forwards on some teams. No idea if it made a huge difference, though it would feel just a little more correct to me. Anyway, probably not a big difference in terms of ranking the players, but as said maybe in terms of the spread. Not sure if that made it a better stat, but anyway, I'm thinking of the Islanders here for example. Guys like Reasoner, Pandolfo, Martin and Bailey have been used on the PK and of course played against guys with high average TOI in those minutes (Kovalchuk, Malkin, Brad Richards, Giroux etc.). That's why I guess they're playing competition averaging around 14-15 minutes here and why they are relatively close to Tavares and Nielsen. However, my feel is that if you excluded PK time, their competition sample would not see much of Giroux etc. anymore at all and their QualCompTOI drop more significantly in relation to what excluding special teams would do with the QualCompTOI of Tavares or Nielsen.
I mean, I don't expect the effect to be huge because after all PK time is limited to a small part of the game. And on many teams the PK specialists will play decent or even strong competition at even strength, too, so it wouldn't change much for them. But, as said, it might help with the spread on some teams. I don't know some of other teams that well, but the Leafs seem to have used some PK specialists with minor roles at EV, too, right? Their QualCompTOI then might be too high in comparison to Grabovski or Kessel as a consequence here
@BenHasna
The x-axis is "what is the average total TOI of the forwards you played against at even strength".
So Reasoner's PK time against Kovalchuk won't count as time he spent playing against quality competition. But it will count in the assessment of how strong an opponent he is for other players.
In other words, if a guy played 8 minutes at even strength and 2 minutes of penalty kill, his qualcomp score is based on who he faced in those 8 minutes, and he counts as a 10-minute player when figuring out other players' qualcomp scores.
Can you weigh the charts for the players' own TOI? Like say in the same fashion that Vollman's bubble charts did for player usage?
@Eric T.
I can see what you are saying, but I am not convinced it would always work out that way. On the road your opponents dictate the matchups. Some of them might try to match up strength vs strength, some of them might try to match up defensive 3rd line vs offensive first line, while others might not try to match lines at all. So, in theory your road ice time QoC might not vary a ton across players and under that scenario the players ice time QoC would be driven by their home matchup scenarios. If the coach doesn't match lines there won't be much horizontal spread.
Carolina looks like a team that used the defensive 3rd line vs 1st line model as the two players with the toughest forward ice time QoC are Sutter and Dwyer. This matches up with my work that indicated that Sutter and Dwyer had the second and third highest opposition goals for rates last year. Skinner and Jokinen played the easier minutes vs opposition forwards. Calgary clearly used the power vs power model.
That said, it may be the case that in the western conference the majority of the teams do employ a power vs power line matching system and that is why most western conference teams have a wide horizontal spread as you suggest. There does, for one reason or another, seem to be a difference between the eastern conference and the western conference in these charts. Interestingly Vancouver doesn't have a huge horizontal spread but that is because they don't employ a line match system bur rather they employ a zone start system instead so in theory the opponents could still get the line match ups they wanted.
Look forward to your home/road split work to further test these theories.
So cool.
Neat how you can pick out teams that don't have tremendous differentiation amongst their top-six (Boston, Calgary, Colorado, Chicago) versus teams with one big line that was seeing most tough comp (Florida, Vancouver, Minny, Philly)
Also, boy do the goons stand out. Parros, Hordichuk, Jackman. Hah-ha!
@Eric T. Oh, ok, that's how it works - that makes sense. Thanks!
@"East vs. West" I guess the reason for the different patterns is indeed that power-vs-power is more common in the Western conference - just because the majority of the outstanding two-way players play in the Western conference (Datsyuk, Toews, Kopitar etc.). That also creates a dynamic where you better do match power-vs-power indeed, or you could get in trouble. In the East, on the other hand, you have some of the offensive stars who excel when sheltered to some extent and that again creates a dynamic where power-vs-power simply isn't what the teams would want to do.
I'm guessing this system is going to visually underestimate the difference in forward strength of opposition. The time on ice of top nine forwards is pretty tightly clustered together with a few outliers compared to the very hierarchical structure of defenseman ice-time.
But the gap in quality of play from top pairing to third pairing defenseman probably similar to the gap between top forwards and 3rd liners.
@Stephan Cooper
Yeah, I think that's true. It's partly because I had to make the charts go all the way down to 12.5 minutes to accommodate the goons.
I stretched them a little horizontally to help compensate, but just doing plots with just the top 9 from each team would probably be a little more visually accurate.
Rick Nash is the most interesting name I picked out there in terms of the QOC degree and the separation from anyone on CLB.
Minor correction:
The way I collected the data, when a team had four forwards on the ice for whatever reason, one of them was counted as a defenseman. That artificially lowered all of the D TOI measures a little bit -- the ranks don't change appreciably, so the plots are fine, but just add a couple tenths of a minute to everyone's total.
It's interesting how some teams (e.g. Vancouver) seem to have one line that faces much tougher D than any other forwards, while a lot of teams have two lines (or more) in the top-right. I wonder if this effect is a result of how teams play their lines, or could it be more the result of how other teams counter them? e.g. Perhaps this is a sign of a team that has one highly offensive line that other teams try to shut-down?
Not really true in the case of the Isles....here you show clear two types of line matching....You have the Tavares-Moulson line (third player often Parenteau), matched up with the top opponents of all types as they were the beset players, and then you had the Nielsen-Grabner line, which was the clear defensive forward line (and thus not matched up against top defensemen).
After that you had no matching whatsoever, which is not surprising given that the 3rd and 4th lines had no clear role for the Isles.
Have you thought about adding Zonestarts to TOI Qualcomp charts on a Z-axis (black dots for players with >50% offensive zonestarts and red dots for
players with less than 50% offensive zonestarts or blue for those right at 50/50). It could provide a better level of insight.
How about defencemen TOI Qualcomps? It might show the value of Erik Karlsson type defencemen, even if it doesn't show much for a majority of the other D-men.
@DavidJohnson
I can speak for how Winnipeg did their matching:
1st line (Ladd, Little, Wheeler) faced second hardest offensive minutes with slightly higher OZ starts. 2nd line (Kane, +2of Burmi, Antro, Welly) were sheltered as much as possible in line matching and zone starts. 3rd line (Glass, Slater, Thorburn) was used as typical shutdown with facing top forwards and lots of DZ starts 4th line (Stapleton, revolving door of AHL player,1of Burmi, Antro, Welly) were hidden as if they could lose the game if they faced anything difficult
I'm late to the party, but this is amazing
@David Johnson
I've looked at matchups for the Blackhawks since Coach Q is so match-uppity. Teams at home don't get 100% of the matchups they want. They probably on average get a 60/40% split.
A Coach dedicated to matchups, however, can get maybe a 70/30 split at home and almost a 50/50 one on the road. This gives you about 10% of the data that is entirely from your coaches moves.
Along with that, your coach accounts for 50% of all the data and the other team's coach only gets 1/82 for every game they play.
So even if 40% of the data is indeed noise there is still a significant amount of data that comes from your coaches use of his players. And that data is the biggest difference in these graphs.
I've looked at 4 years data of the Blackhawks top 4 forwards; Toews, Kane, Hossa and Sharp. And the results have been surprising. Sharp is the least two way player, as these graphs show, of those four forwards. And that is surprising considering his two way rep compared to Kane's. So whether you use TOI QoC or one of the already existing methodologies, you get very similar results.
And by the way, the moment I read TOI Qoc, my initial response was, darn, why didn't I think of that. I really like this data and in particular splitting off the D-men from the forwards. I think that makes a huge difference.