...............

=== SSI's System ===

Is, as you know, The World's Most Dangerous Saber System.  It is, of course, a near-exact analogue to chess pattern recognition and position evaluation.  Give or take a few weak squares.

So, Igor, before I throw the switch and electrocute myself, get your notes straight.  It works this way:

Intro.  Be aware of anything that others have said about a player's potential comps.

1.  Use data-nourished human judgment to isolate a player's truly defining characteristics.

1b.  If possible, use a think-tank mechanism to refine this attribute set.  In other words, if G-Money can convince you that Matt Tuiasosopo's age-and-level is part of this set, then use it.  If Taro can convince you that Trayvon's K > 225 is part of his defining characteristic, then use that.

2.  Gather a list of other players who match these defining characteristics.  Set a prototype, the best player in the group.

3.  Identify the ways in which your comp player differs from his comp group.  Maybe Ichiro is Pete Rose, plus speed, minus walks.  Maybe Ichiro is at the 90-95% talent level of the prototype Pete Rose, or maybe the 105% level.  Bill James characterized this as "looking for something out of alignment in the statistics."

Whatever position you've got, it's been played before.  Figure out how that little h3-pawn changes the normal way of winning the position, make a plan, and launch your attack.

4.  After all that, be wrong and lose anyway.  Life's tough, then you die.

If Bill James' thought process, ca. 2011-12, differs from the above, Dr. D is not aware of it.

..............

Let's take Michael Pineda.  In SSI's view, his defining attributes were [extreme velocity] intersected with [extreme command].

Somebody else says his defining attribute is the lack of three pitches.  Another person says his defining attribute is his lack of AAA experience.  And so on.

IF AND WHEN WE HAVE AGREEMENT ON THE ATTRIBUTES, we can proceed to an analysis that will be helpful to all of us.  SSI agreed with itself :- ) on Pineda's attributes, so his comp pool was obvious.  Curt Schilling, Bartolo Colon, Smokey Joe Wood, Bob Feller ... Walter Johnson used a located fastball over 90% of the time.  The young Roger Clemens fanned 20 Mariners with located fastballs.

There were guys who had adjustments, of course.  Bartolo Colon's rookie season was tough...

.

=== Dr's Diagnosis, Dept. ===

Using the Pitcher + Hitter Families paradigm isn't a guarantee.  It's tough to define the attributes with total confidence, and even if we did, an individual player can underperform or overperform.  Is Ozzie Canseco a comp to Jose Canseco?  They're twin brothers.  They didn't exactly follow the same career path.

But if we're going to use Pitcher + Hitter Families, I suggest that we do two things.  One, if we're going to do it by computer, be aware that PECOTA, and B-Ref.com / James, are already doing about as good a job of it as can be done by computer.

Two, if we're going to use human judgment, pay excruciating attention to the first part, the player's defining attributes.

BABVA,

Dr D