In the last post in this series, I explained the reason why, when evaluating players, you need to take into account the context in which that player put up his numbers. After all, not all ice time is the same: some ice time is very favorable for scoring while other ice time is very unfavorable.*
However, simply looking at Time on Ice splits is not enough for us to truly understand the context in which a player puts up numbers, since NOT ALL 5 on 5 ice-time is the same. Some ice time is harder than others, due to the level of the opposition and where the shifts start. By contrast, Ice time can be made easier by a player having better supporting teammates (linemates and D-men) on the ice with him at any time.
In this post, we'll be looking at the first two factors I mentioned in my last post, which indicate how DIFFICULT a player's ice time really was.
*Note: This is NOT to suggest that any player can go from no PP time to PP time and be a successful power play player. Roles do matter. But what this means is that we need to keep certain numbers in perspective - if two players get the same amount of points, but one does it entirely on the PP while one does it at Even Strength, the 2nd player is more valuable as we expect that if that player was given power play minutes, he'd contribute more than he would at even strength and thus would surpass the other player, even if his power play scoring might not quite come at the rate of that other player. This concept of value will be explained further in Part 2.4.
Quality of Competition:
Of key importance in determining what a players' numbers mean is to figure out how tough his opponents were when he accumulated those numbers. If a player faces tougher opposition, odds are good certain numbers (especially numbers reflecting defense such as +/- and corsi) will be lower than they'd be against weaker opposition. And to truly understand what those numbers mean, we need to thus take into account the quality of competition a player has faced.
Fortunately, we have two key measures for measuring the quality of a player's opponents while that player is on the ice. The first metric is known generally as QualComp.
QualComp (short for Quality of Competition) is a measure of "the average Relative Plus-Minus of opposing players, weighted by head-to-head ice time" per 60. "Relative Plus-Minus," which I will talk about more in-depth in the next post, basically measures the +/- of a player as compared to his teammates. All you need to know about this metric for now is that higher Relative +/-s indicate in general that a player is good, while low (negative) Relative +/-s indicate that a player is poor.
So what QualComp does is it basically figures out the average Relative +/- of the opponents a player faced in his ice time, per 60 minutes. The higher the number, the harder the opponents one has faced. The lower the number, the weaker opponents one has faced.
The virtue of QualComp is best explained by examples:
The players who had the top 5 hardest opponents in the NHL in 2010-2011 (minimum 40 games) were:
1. Nicklas Lidstrom - 0.128
2. Dave Bolland - 0.121
3. Brent Seabrook - 0.117
4. Patrick Marleau - 0.115
5. Robyn Regehr - 0.114
These names, particularly #1, should be quite familiar to most of you: these are some of the highest reputation defensive players (2 forwards, 3 defensemen). All people you'd expect to be playing against top opponents.
Okay, so how about the Isles? Who faced the toughest competition last year (40 games played minimum, so Kyle Okposo doesn't qualify) on the team?
1. Travis Hamonic - 0.046
2. Andy Macdonald - 0.031
Huh, who would've guessed: the toughest opponents of the Islanders were faced by the team's #1 Defensive pair! Next up on the list was the Tavares line and then Nielsen and Grabner. Incidentally, this is slightly misleading...if you lower the minimum threshold to 20 games, Kyle Okposo pops up with the toughest competition on the team and since he played almost always as part of FN-GO, it's pretty clear that the Okposo-Nielsen-Grabner line faced the toughest competition of any forwards since Okposo came back.
Now QualComp has some issues: for one, it relies upon a +/- statistic (relative +/- to be specific). As such, it can be affected by statistical flukes - for example, Kyle Okposo had a poor +/- last year solely because of bad luck with Martin Biron, and thus players who faced him would have their QualComp lowered. This isn't really much of a problem over 82 games - the statistical flukes will tend to even out in that large sample size.
But in smaller sample sizes, lets say from 10-40 games (and possibly even for larger sample sizes), these errors can be meaningful and result in some odd context measurements. Fortunately, we have a way to deal with these:
The solution is the second Quality of Competition metric, what I call Relative Corsi Quality of Competition. This is listed on behindthenet.ca as Corsi Rel QoC (or Corsi Rel Quality of Competition), but it should be pretty obvious these are the same things. This metric is basically the exact same as QUALCOMP, but measures the average opposition with Relative Corsi rather than Relative +/-. I'm not going to explain Relative Corsi quite yet (that's for another post), but essentially corsi is a measure that measures shot differential as a way of approximating positional dominance (the team with a positive shot differential is spending more time in a more favorable position - in the offensive zone).
The advantage of Relative Corsi Quality of Competition, and corsi in general, is that because it uses shots rather than goals, it relies upon a much larger sample size. Whereas a single game will have only maybe 1-2 goals while a player is on ice, it will most certainly have at least 5+ shots (quite possibly far more) instead. Thus this metric will show the true quality of competition of a player in a lower amount of games.
The players who played the toughest competition in the NHL according to this statistic in 2010-2011 are:
1. Ryan Callahan +1.848
2. Nicklas Lidstrom +1.807
3. Marc Staal +1.602
4. Dan Girardi +1.583
5. Brad Stuart +1.501
Once again, we get a good five players who we might expect to see on such a list; Lidstrom once again shows up, while his D partner also shows up at #5. In addition, a trio of Rangers show up, including Defensive Forward Ryan Callahan and the Rangers' top D pair. You might note that 4 of the top 5 in this list aren't on the former list; however, all four of the players on the QUALCOMP list (for QualComp) show up on the Relative Corsi QualComp list if you go a little farther, and Staal, Girardi, Stuart, and Callahan all show up in the top 20 for QualComp as well.
In other words, after 82 games, the measures don't have the exact same numbers, but they're generally in agreement about who has played tough or super tough competition, while who has not. We can see this again with the Relative Corsi Quality of Competition for the Islanders:
1. Andy Macdonald +.739
2. Travis Hamonic +.729
NOTE: There is a third version of Quality of Competition, Corsi QoC, on behindthenet.ca. I'd highly recommend not using it - it essentially uses straight Corsi rather than Relative Corsi for it's QoC calculation. The result is that players on teams in certain divisions always jump up to the top: thus a prevalence of Anaheim Ducks and Dallas Stars on the list. There's some merit toward using it to compare players on different teams...but not really as much as you'd think.
Now, Quality of Competition is one way in which we can measure the DIFFICULTY (not ease, difficulty) of certain ice time. But it doesn't capture the whole picture. See, a hockey coach sends out his lines, when possible, to match players of certain skill against certain opponents. Quality of Competition essentially measures this effect; it shows what players the coach feels should face stronger opponents and what players should be sheltered immensely.
But Coaches send out certain players onto the ice for another reason other than matching the opponents on the ice; coaches take into account the position of the puck when they decide who they want on the ice.
What do I mean by this? Well, consider when play is stopped. Play restarts with a faceoff. But that faceoff can be in various locations: it can be in the neutral zone, but quite often it's either in a team's offensive zone or its defensive zone. And a coach takes this into account when deciding what players to throw out onto the ice....if the puck is in the defensive zone, the coach will likely send out his defensive players (if available), even if the opponent's top scorers are not on the ice. If it's in the offensive zone, a coach will likely send out his scorers to take greater advantage. In addition, there are various odds and ends such as faceoff-specialists, who very likely will be out for defensive faceoffs a lot, even if they're not good defensive players.
As you might suspect, this can have a decent sized impact on a player's numbers: If a player is sent out for a ton of defensive faceoffs, but doesn't get many offensive faceoffs, that player is going to have "worse" numbers all around, because the opponents will get more shots while he's on the ice, his team will get less shots, and he himself will get less opportunities to score. In some cases this can be a small effect, but for certain players it's not, and we need to account for this. Fortunately, we can.
BehindTheNet.ca includes statistics calculating the number of faceoff wins and losses of each type (defensive zone, neutral zone, offensive zone) while a person is on the ice. Included in these numbers is a statistic labeled "OZone%." This number is known more commonly as the Zone-Start %. What it measures is the percentage of offensive zone faceoffs out of total non-neutral zone faceoffs is for a player, so that this measure is calculated as follows for a player:
Total Number of Offensive Faceoffs Taken While a Player is on the Ice
The Total Number of Offensive AND Defensive Faceoffs Taken while a player is on the ice
(Neutral Zone Faceoffs are ignored in this calculation since these are naturally neutral - not inherently favorable or unfavorable).
Players with high zone-start %s face easier ice time - as they're out for more offensive draws than defensive draws, they start shifts more likely to be able to get shots on net than those players with lower zone-start numbers.
Once again, this is simply a measure of context: it doesn't say that a player is good or bad. After all, a team's best scorers (its top line) will nearly always have the highest zone-start numbers because a team will want them on the ice when the team has the best opportunity to score. We see that with the Islanders as well: the top zone-starts on the team, minimum 40 games, are as follows:
1. P.A. Parenteau - 56%
2. John Tavares - 55.6%
3. Matt Moulson - 54.2%
Well, this makes sense, the Isles top line is on for more offensive zone faceoffs than defensive faceoffs, so they can score more easily.
How about the lowest Zone-Starts on the team?
1. Zenon Konopka 30.2%
2. Frans Nielsen 41.4%
Once again, this makes sense: you have your faceoff specialist, who by design is meant to lower the damage of defensive zone faceoffs, and your top defensive forward, whom you want also in the toughest situations.
Once again, like with ToI and Quality of Competition above, we use zone-start % to establish our baseline for how players should perform. Players who have high zone-start %s are expected to score more and to have better possession (corsi) and +/- statistics, while the opposite are expected to score less. Thus, for example, while Daniel Sedin leads the NHL in points, we need to discount the fact that he had extremely favorable zone starts - with a zone-start % of 74.5%, the HIGHEST IN THE LEAGUE - when we compare him to other top point getters. In Sedin's case, that's a big extreme number, which means that his point total looks a bit less impressive as a decent part of it may simply be the result of him being given a great deal more opportunities than nearly any other player in the league. (Daniel Sedin is still great, don't get me wrong. But we expect a higher performance from such a zone-start %, so we need to reduce his numbers somewhat in our head.)
Now players on worse teams will on average have lower zone-starts than players on better teams, as the better teams are able to get more offensive zone faceoffs. But for comparing individual players, this is irrelevant; to avoid analyzing players without context, we need to know where they started their shifts (on faceoffs at least), and thus their zone-start %. It is not an excuse for the individual player that their team doesn't get as many defensive faceoffs, so we shouldn't discount some of his numbers due to having easier ice time.
NOTE: BehindtheNet also includes a second statistic labeled "Fin OZone%." Ignore that statistic for now, it has nothing to do with Zone Starts.
Please hit me up in the comments if some of this is confusing, I could quite well have botched the explanation. But the big thing to take away here is this: We CAN measure how tough ice time is based upon players facing tougher competition and where they start their ice time (on faceoffs). We CANNOT make judgments in a vacuum; we need to take these things into account. And we can, thanks to these numbers.
Before you ask: We can take into account quality of teammates as well! But that's a discussion for the next post, which will come after I'm done with finals (so in a week+).