David Johnson, whose stats.hockeyanalysis.com site is, along with BTN, indispensable (particularly as fears for the future of timeonice.com mount), disagrees with me about the impact of ZoneStarts on the numbers for defencemen. From the previous thread:
Let’s take your third chart where you take average corsi% by zone start group. That chart seems to make it pretty clear that zone starts greatly impact corsi% but the problem is I suspect that QoT would show a very similar pattern. Bad offensive players don’t start often in the offensive zone (and thus generally start more often in the defensive zone) and if you are a bad offensive player it is difficult to get a good corsi%…So, is that defenseman’s resulting bad corsi% the result of his defensive zone starts or his weaker quality of teammates? I will argue the weaker quality of teammates is the far more significant factor.
The same problem arises in your final chart. If a defenseman gets a significantly different zone start profile from one season to the next it probably means he also has a somewhat different set of team mates he shares the ice with so again, was the change in corsi% due to the zone start or due to the quality of team mates.
Yeah, the problem is you are taking players who are playing with different players and against different players and with different zone starts and concluding that the linear variation in corsi% is due to zone starts and not those other factors.
David provided some links to work that he and Gabe Desjardins have done that supports the idea that, after ten seconds, a lot of the advantage/disadvantage of an offensive/defensive zone draw has washed out. Their findings are broadly consistent with my own small sample look at it a few years back.
Maybe I am biased but I am more inclined to accept my 10 second analysis than that in this post which makes the assumption that all of the observed correlation is due to zone starts and not other factors.
I’ve copied Gabe’s fine graph (hopefully he won’t sue me), which I’ve reproduced above. It illustrates the effect that we’re talking about (the volume of shots seems off to me, which I spoke with Gabe and he agreed with, although the effect seems to be correct). The green line is shots per 60 minutes X seconds after an offensive zone faceoff win. The brown line is shots per 60 minutes x seconds after an offensive zone faceoff loss. You can flip this around to determine how many shots a team that wins or loses a defensive zone faceoff allows per 60 minutes X seconds after a faceoff loss.
You’ll note that there’s some information missing, in a manner of speaking, from this graph. It doesn’t include the impact on SA/60 of winning or losing an offensive zone faceoff (and, by necessary implication, of winning or losing a defensive zone faceoff). David Johnson touched on that issue in his own post on the matter, which is linked above.
David tried to isolate the ZoneStart effect by comparing a player’s Fenwick/20 with his Fenwick/20 with 10, 20 and 30 seconds after faceoffs removed. He found that after 10 seconds, there isn’t much of a difference if you remove 20 and then 30 seconds. His conclusion is that to the effect that the ZoneStart effect bears on Corsi, it bears on it in the first ten seconds after a puck is dropped and then it’s washed out.
OK. So that’s the case to meet. In my previous post, I mentioned that I had created pairs of D-seasons for every defenceman who was on the ice for 100 faceoffs in consecutive seasons and that, as a rule, when ZoneStart goes down, Corsi goes down. I’m going to quote another thing from David here:
So, take a look at how things change for a few players with extreme zone starts (2 year 2010-12 data).
H. Sedin 5v5 FF%: 55.2
H. Sedin 5v5 ZS Adj. FF%: 53.0
Malhotra 5v5 FF%: 42.5
Malhotra 5v5 ZS Adj. FF%: 43.9
P. Kane 5v5 FF%: 54.7%
P. Kane 5v5 ZS Adj. FF%: 54.9%
Bolland 5v5 FF%: 49.0
Bolland 5v5 ZS Adj. FF%: 50.8
As mentioned, David’s ZS Adj. numbers ignore the first ten seconds after an offensive or defensive zone faceoff. What David is saying is that if most of the value (or cost) of an offensive/defensive zone faceoff is found in the first ten seconds, then if they’re a significant factor, we should be seeing a big impact in the numbers of guys who get extreme ZoneStarts. The difference for Sedin, Malhotra, Kane and Bolland seems small; therefore the impact of ZoneStarts is negligible.
Now, leave aside the possibility that things are different for defencemen (which is what we’re talking about) than they are for forwards. I don’t think it’s disputed that guys see big Corsi drops when their ZoneStart gets significantly more difficult. I’ve identified a group of 48 defencemen who saw their ZoneStart drop at least 10 percentage points from year to the next; the mean change in their Corsi% was about 4.4 percentage points. My working theory is that this is ZoneStart related; David’s theory is that this is related a bit to ZoneStart but more to the change in the quality of their teammates.
OK. What if we actually look at one of the guys who had a big drop in his ZoneStart and look at how he did when on the ice with given teammates from one year to the next? Phoenix’s Zbynek Michalek (I know how to pick guys from large markets to write about to draw hits) had a 55.3 ZoneStart in 2007-08, with a 51.1% Corsi%. In 2008-09, he had a 35.5 ZoneStart and a 43.4% Corsi%. Was it teammate related? I doubt it:
Michalek’s main defensive partner switched from being Ed Jovanovski to Kurt Sauer from 2007-08 to 2008-09 but Michalek and Jovanovski still took a big hit when they were together. He got hammered with all sorts of guys with whom he’d put up big Corsi numbers the year before. I took a look at the ZoneStart adjusted information on David’s site and it doesn’t really show much of a difference between it and the raw Corsi numbers.
Want another? Mark Giordano’s ZoneStart fell from 63.8 to 52.7 from 2008-09 to 2009-10. His Corsi% fell from 59.4% to 54.4%. Here’s his data with different teammates in 2008-09 and 2009-10.
Again: same deal. The fall is across all of his teammates and it’s awfully uniform looking. If David’s theory was right, the data wouldn’t look this way – we’d only see tiny drops in the raw data when comparing, for example, Gio/Iggy in Yr1 with Gio/Iggy in Yr2 (subject to there being massive changes elsewhere in the team). I don’t think that that’s what this data is showing.
I mentioned above that my group of defencemen who sustained drops in their ZoneStart of 10 points or more has 48 people in it. Only four of those people saw an increase in their Corsi% in the year in which they sustained the drop in ZoneStart. By eye, a lot of them sustained pretty dramatic changes in their teams/circumstances. The four are Radek Martinek (who went from the Isles to the Blue Jackets and was only on the ice for 100 faceoffs in year 2), Jason Garrison (whose Florida team was gutted between the 2009-10 and 2010-11 seasons), Mark Fraser (who was only on the ice for 206 draws in Yr2) and Jay Harrison (who played for the disappointing 2009-10 Hurricanes and then saw his ZS get tougher the next year, albeit on a much different team).
I haven’t cherrypicked the players who I looked at. I’ve posted the list here and people are free to go and look up different players than I picked; maybe they’ll find something different. I doubt it but fill your boots.
Johnson’s got another post up that I haven’t looked at in detail, although he seems to acknowledge a greater possible spread than he originally did. I see that he suggests that some of the difference might be due to me using BTN data while he uses his site’s data, which has 6v5 situations stripped from it. I’m surprised that he’s suggested this – it’s easy enough to look and see that the differences between the two sites’ data are negligible.
What do I think is going on here? One potential answer that leaps out at me is the impact of faceoff wins/losses. Take Michalek. In 2007-08, the Coyotes won 53.1% of the offensive draws when he was on the ice and 49.7% of the defensive zone draws. In 2008-09, those numbers were 43% and 46.4%, respectively. Given that we know that winning/losing faceoffs leads to Corsi events, I’d guess that a swing like that has some impact. I doubt that it’s a major thing in most cases, but Michalek suffered one of the worst swings when his ZoneStart declined (and likely had one of the biggest swings in the faceoff winning percentage in the offensive and defensive zones). It’s probably worth looking into how the faceoff W/L stats changed and considering whether some correction needs to be applied for that.
I’m also inclined to think that what may be true at a team level as far as the faceoff effect on Corsi isn’t necessarily true on an individual level. For example, maybe guys who are shifted into heavier ZoneStart roles have a tendency to change when the puck leaves the defensive zone. What might be true on the team level then, in terms of how long the faceoff effect lasts, would not be true on an individual level. Say the Yotes were running Michalek as almost entirely a defensive guy. He would generally shift off when the puck went up ice. If the Yotes lost a draw and the play stayed in their end, he’s stuck out there. If the Yotes win the draw or obtain possession quickly, maybe he changes.
I’m spitballing here but the data we can see simply doesn’t seem to support Johnson’s hypothesis. This is a question that can be answered (although not without a lot of tedious work, for me, anyway) by simply examining the shifts of players. Eventually, someone will get around to doing that.Email Tyler Dellow at firstname.lastname@example.org