Thursday's Numbers: Adjusting Individual Fenwick for score-tied situations

We count quality shots, too, but we hardly need to. There's a similarity between individual player shot differential and scoring chance differential the more games you play, I think these numbers tend to even out, and if a player ends up with a lower scoring chance percentage than Corsi or Fenwick percentage, that isn't necessarily a bad thing.

Consider the case of Alex Ovechkin: He's taking fewer shots this year and the year before than he had in years past, but he's taking them from closer locations. (If you eyeball Ovechkin's shooting data from 2010 compared to his 2011 data, you see well more shots to the outside) Yet he went from scoring .69 goals per game in '10 to .41 goals per game last season, on the back of his taking .4 fewer overall shots per game.

For a Canuck example, I'd bring up Ryan Kesler. While the term "perimeter player" has been oft thrown around with a negative connotation in Vancouver since the days of Markus Naslund captaining the team, I think it perfectly describes Kesler. He takes a lot of shots from the outside, but that opens the door to rebounds, lucky bounces, lapses in goaltending, and overall scoring-chance creation.

I look at two players, Henrik and Daniel Sedin, who are content to work their way to finding quality shots in the zone. This is a talent that doesn't pay off at even strength so much and the two are often hemmed in their own end. The thing with the Sedins, however, is that they have an unusual talent reserved for a few key players in the league to actually be able to improve the quality of their shots. There was a great post on the playmaker-effect from our friends at Broad Street Hockey about this effect last month—Henrik Sedin was found to have the highest positive effect on teammate shooting percentages, followed by Joe Thornton among the players studied.

So, contrary to popular belief, there are only a few players who can truly control their PDO, which is the addition of on-ice shooting percentage and on-ice save percentage. Much of goal-scoring is luck, and there was a larger link when I looked in December between scoring chances (quality shots) and Fenwick numbers than there was between scoring chances and overall goal rates.

This is why I love the Fenwick number. Fenwick is a variant on Corsi, which can be looked at as either a percentage or an overall number that measures overall shot attempt differential. It's a great predictor of future team success, especially when totaling up score-tied situations. Here's how the regular Canucks have fared in this department this season:

You can tell a few players at the top and bottom of the list, we don't really have a good sample for. I just kept Mike Duco in there because I really like Mike Duco for some weird reason. Point being, however, this isn't a good indicator of overall team talent. We all know about how many extra offensive zone starts Henrik and Daniel Sedin get compared to Manny Malhotra, and that gives them an unfair advantage.

Normally, we adjust Fenwick numbers by assuming that every extra offensive zone start is worth 0.6 of a Fenwick event (a Fenwick event is a goal, shot or missed shot), but we can't do that here because we don't have score-tied faceoff data available at timeonice.com (a brilliant site which supplies me with most of my data). However, over at Behind The Net, we do have overall faceoff zone start statistics, which I used to create a proxy for the number of offensive or defensive zone starts a Canuck player had with the score tied at even strength.

I had to adjust for two things:

A) Since Corsi and Zone Starts sort of correlate, I had to assume that the Canucks, who have a 52.3% Corsi rate but 53.9% with the score tied, would probably have slightly more offensive zone faceoffs with the score tied.

B) The overall number of events. The amount of Corsi events (goals, shots, missed shots and blocked shots) that took place in Canuck games thus far is 3766, compared to just 1147 with the score tied.

Using the Behind The Net data and factoring in these two things, I came up with an "Adjusted score-tied Fenwick" percentage that I think is more indicative of each Canuck's play at even strength. I only took players who have been in the lineup for at least 10 games this season:

You can see that the Sedin twins take a very sharp drop in the adjustment, and Manny Malhotra earns a sharp increase. He's been noticeably better, and, gee, the Canucks will be lucky to get David Booth back out in the lineup, who is probably the team's strongest possession player.

I'd suggest this is a relatively good proxy for Canuck play, as the team's overall Fenwick rate rests right around the median here (between Alexander Edler and Dan Hamhuis). Anything over 50% is considered good, anything over 55% is great, and anything over 60% is considered elite. Now, are Booth and Malhotra the two best players on the Canucks? Probably not, considering Booth has played just 18 games, but I still consider Malhotra extremely important to the team's success. That's why I like the fact that he used to play more minutes than Cody Hodgson, because he can handle to Canucks tough shifts much better. Not many young players can control possession.

Finally, after a quick start, it looks like Mason Raymond has cooled off, and I think that he may be on his way out. From looking at this, you need to assume the Canucks need another top-four defenseman to make up for the fact that Keith Ballard and Alex Sulzer are pretty terrible (although somehow Andrew Alberts looks good under this stat, but he's been pretty good game-to-game since the loss in Carolina).

Also, Aaron Rome is an extremely underrated defenseman in Vancouver. He's very low-event, although unspectacular, doesn't allow a lot of shots. Let's, however, not confuse possession statistics with overall importance, because this doesn't consider the quality of opposition faced or the quality of a player's teammates. It does, however, give us a good idea of where the puck is likely to be with a player on the ice.

Any comments on the list? Also, to any visiting statgeeks who actually know how to do math, want to pick apart my method?