Sunday, August 22, 2010

How Good is Alexander Semin?

This upcoming season, George McPhee has a big decision to make regarding the future of Alexander Semin--to either keep him and sign him to an extension, keep him and let him go, or trade him. To sign Semin means to invest a lot of money in a guy who at best is the second best winger on his team, a guy probably "doomed" to be a second liner in Washington (admittedly not too bad). To keep him and let him walk on July 1 means to hope that Alexander Semin takes a hot streak into the playoffs and stays healthy, immediately taking the Capitals from dangerous to very dangerous, and to trade him means to take a good return to help build for the future and/or be more set at center or defense (or wing, but that wouldn't make much sense) for the playoffs. Obviously, there are pros and cons to each situation. It will be best if McPhee knows exactly how much Semin is worth and extracts that much value from him, if not more.

One way to look at how good a player is is by looking at Corsi: specifically, how well a player does with and without his regular linemates, to try and estimate how much success can be attributed to the player in question. It's called with-you-without-you analysis (WOWY for short). Some examples are these on Alex Ovechkin, Shawn Horcoff, Frans Nielsen, Ilya Kovalchuk, Paul Martin, Travis Zajac, Ales Hemsky, Jason Strudwick, and Ryan Callahan and Brandon Dubinsky.

The way to do this is to find who a player played with during the season, then use Vic Ferrari's TimeonIce scripts (links to instructions on the right sidebar) to find the Corsi numbers you're looking for.

By default, Ferrari's scripts are even-strength only, not including empty-net situations.

First, Bruce Boudreau really shuffled Semin around with different linemates. I count Alex Ovechkin, Nicklas Backstrom, Tomas Fleischmann, Brooks Laich, Brendan Morrison, Eric Belanger, Jason Chimera, Mathieu Perreault, David Steckel, Mike Knuble, Quintin Laing, Boyd Gordon, and Eric Fehr (and I'm sure I missed people too). The ones with significant ES TOI with Semin (#28) though are Ovechkin (#8), Backstrom (#19), Fleischmann (#14), Laich (#21), and Morrison (#9) (I won't count Belanger here). So we'll look at those five with and without Semin.

Here's the script I'm using:

(The first part looks kind of ugly, but when you scroll down you see a beautiful chart)

I'll format the table like in the Ovechkin WOWY, but switching horizontal and vertical labels to make it thinner and thus better to Google Spreadsheet-embed in Blogger.

"Corsi+" is blocked shots against + goals for + saved shots for + missed shots for (tally the first columns under each section under the Player "On Ice" Stats. Similarly, "Corsi-" is the sum of the second columns. "Total" is" Corsi+" + "Corsi-." "Corsi%" is listed as a bolded column. To find "Semin without X", I went to the "28" row to get Semin's totals and subtracted X's totals with Semin from Semin's overall totals. To find "Player without Semin" I substituted the player's number at the end of the script instead of Semin's 28 (so for Ovechkin, script ends in shawn=8), went to the player's number to find the player's overall totals, and subtracted the totals with Semin from these overall totals.

If you do this yourself, having a calculator handy is crucial unless you run formulas in Google Spreadsheets, which I don't know how to do. I find the computer calculator is slow to use, even with NumLock.

Here is what I got. Delta1 is how much better the player's Corsi% is with Semin than without Semin. Delta2 is how much better Semin's Corsi% is with the player than without the player. Better players like Ovechkin have higher Delta1s and lower Delta2s across the board in general, since better players themselves drive Corsi to a great extent.

I decided not to do defensemen at this point since that would double the work I just did, but it's a safe bet that Semin's Corsi% would improve with Green, Schultz, and Carlson--they were the players aside from Corvo would have had strong Corsi% without Ovechkin (follow the Ovechkin WOWY link for more).

A lot of it goes as we would think--Semin is better with Ovechkin and Backstrom, and worse with Fleischmann and Laich. It's interesting to see that Brendan Morrison made Semin a lot better. It appears that they had some chemistry together while Morrison was playing well earlier in the season.

Semin isn't the Corsi engine that Ovechkin is, but he looks like more of one than Backstrom, who doesn't break even without Ovechkin (though it should be noted the relatively small sample of Backstrom without Ovechkin separated). Semin slightly more than breaks even on his own. If McPhee can find a center like Morrison (hey, he's still unsigned, right?) with whom Semin has chemistry than that line could potentially be as good in terms of percent of play in the offensive zone as the Ovechkin-Backstrom-Knuble line. That's a deadly one-two punch.


  1. "If McPhee can find a center like Morrison (hey, he's still unsigned, right?) with whom Semin has chemistry than that line could potentially be as good in terms of percent of play in the offensive zone as the Ovechkin-Backstrom-Knuble line. That's a deadly one-two punch."

    Dead nuts on - would hav e loved to see what the Belanger numbers looked like that said it's clear that "Flash" isn't the answer as a pivot for a second line between Semin and Laich when you look at these numbers so I wait with baited breath for some sort of move in that area... or I guess some sort of move to see who takes over Poti's spot after this season is over on Defense....

  2. I didn't add Belanger because it seemed like BB was shuffling him to a different spot every night--there just wasn't any consistency in the forward lines aside from the top line, and I'm also uncertain if the stats will include his time in Minnesota or not. If you're interested, here they are. He wore #18.

    Belanger total: .529 Corsi%, 7.4% shooting
    Belanger with Semin: .553 Corsi% (but their sh% was only 5%. No wonder he didn't look good)

    Only 17 games though. His sample size is much smaller than BMo's, though it's still decent.