Wednesday, February 27, 2013

Talent Flow, Part 3

Prior to Season 20, I played a little game to see if I could come up with a simple number to summarize preseason talent movement among the 8 divisions (See "Talent Flow Update" way back on Aug 12).

I just assigned a value of 1 to Type B FA's, 2 to Type A's, and 4 to "Super A's" (just a term I made up to describe the highest-quality free agents); and then just applied the values to each Type A and B free agent move (if a team lost a Type A free agent, that was -2 for that division, etc.).  Then I applied the values to trades involving Type A and B-caliber players.  Finally, I just totaled up the scores of the divisions.

Then I wondered if this simple metric had any predictive value...could it predict which divisions would win more games than in the previous year (knowing, of course, that this simple number didn't capture most of the free agent moves, none of the prospect promotions, etc.).

For Season 20, the Talent Flow predicted the direction of win-loss changes for 6 of 8 divisions.  75%.  It didn't do nearly as well predicting the magnitude of W-L changes, so for Season 20, I'd have to say the predictive value of my number was mixed at best.

How'd it do last year

It only predicted the direction of W-L changes for 4 divisions correctly...if my dog could throw darts she'd likely do that well.  Only 2 divisions had pretty big talent moves (+ or - 5 or more) - AL East and NL South lost a lot of talent, but both divisions had big win improvements.  So Talent Flow predicted neither direction or magnitude right in those cases.

The verdict is in:  my Talent Flow Number sucks as a predictive tool.  Maybe all the other talent moves unaccounted for by my number overwhelm preseason trades and A and B free agent moves.  Maybe the usual performance variations from year to year render it meaningless.  Maybe (??), because my number is mostly based on the bigger-name FA moves, we're overpaying for those free agents.

No comments: