I've been struggling to come up with a way to value individual players in an analytical environment. Scratch that; "struggling" isn't strong enough of a word. I've been bashing my face against the wall to try and find a way to determine individual production, and the results have been mixed at best. It's just hard to see through all the data and try and formulate a methodology that actually makes sense. Luckily, I no longer think that I'll need to take ambulance rides due to unfortunate encouters between my forehead and the sides of my apartment.
Dr. Larry Feldman, the mind behind LaxPower, appears to have re-launched his individual player ratings and they look like they make a lot of sense. He's calling the measure "Player Impact Rating" and it's just about as complete as you can get:
The Player Impact Rating (PIR) is based on events which take place on the field. Each event influences the outcome of the game in either a positive or negative way. The player, his/her position, the event and the significance of the event all impacts the rating. Only those events which are statistically listed on a team's web page are used.
Now, the ratings aren't perfect in my mind, but they are as close as you can get without putting together additional advanced concepts into the formula:
- What I Like: Comprehensive to a standard box score, all adjusted for opponents played. It's the adjustment for strength of schedule that really make the ratings sing.
- What Concerns Me: The adjustment for strength of schedule is based on per-game metrics rather than tempo-free measures. Without getting too technical, the "noise" concomitant in per-game metrics has a hand in all this, potentially helping or hurting players that play for or play against teams that play lots or few possessions per contest. Also, face-off specialists are given heavy consideration arguably out of balance to their overall value.
In totem, though, the approach and implementation is sublime. Dr. Feldman did a terrific job conceptualizing the formula -- from delineating between positions and function to valuing each underlying metric -- and making it all work. The results prove out this fact: Peter Baum was rated as the most impactful attack in 2012; C.J. Costabile was the most notable midfielder; Brian Megill was the top defenseman in terms of overall value; and John Kemp was rated as the top keeper. It's hard to argue deeply with these ratings (although, you can start to get yappy as you move down through the list of each position).
So, with an eye toward these ratings, what would constitute a preseason All-America list for 2013 built exclusively from the Player Impact Ratings from 2012? I think it'd look something like this:
ATTACK: Peter Baum (Colgate); Will Manny (Massachusetts); Marcus Holman (North Carolina)*
MIDFIELD (OFFENSIVE): Jeremy Noble (Denver); Tom Schreiber (Princeton); Wes Berg (Denver)^
MIDFIELD (LONG-STICK AND SHORT-STICK DEFENSIVE): Scott Ratliff (Loyola); Chris LaPierre (Virginia)
MIDFIELD (FACE-OFF SPECIALIST): RG Keenan (North Carolina)
DEFENSE: Brian Megill (Syracuse); Michael McCormack (Yale); Bobby Lawrence (Colgate)
GOALIE: John Kemp (Notre Dame)
* Holman is oddly listed as a defensive midfielder/defensive player in the ratings. My assumption is that Holman would finish above Kyle Smith (Massachusetts) if the attack formula was applied to his production.
^ There are lots of face-off specialists listed ahead of Wes Berg, and while they do contribute to offensive and defensive postures, I've washed them out for the purposes of this piece.
- Statistics never tell the entire story, but given where things are coming out, this is a very workable measure from Dr. Feldman. As always, statistics should work in concert with an eyeballs analysis. There are no silver bullets, after all.
- I like the rating of Bobby Lawrence. The defensemen has been somewhat overlooked entering 2013, but he is arguably the strongest piece of Colgate's defense going into next year. Entering his junior year with the Raiders, he could get himself into a position as a first- or second-team All-American should his improvement continue.
- My personal first-team preseason All-America list is pretty close to what the Player Impact Ratings are showing. The obvious difference of including Rob Pannell aside, I didn't consider Wes Berg, Lawrence, or McCormack, and only considered LaPierre for the short-stick defensive midfield position (ultimately going with Josh Hawkins). Understanding and valuing defensive players is a cruel mistress.
- The biggest ratings gaps between the top-rated and second-rated players at the various positions come with Kemp in goal (+12.84) and Baum at attack (+6.62). Not only were these guys incredibly valuable to their teams in 2012, they were significantly ahead of their peers in terms of production at their various positions. Good show.