Wednesday, September 17, 2014

A New 500 Resims (355 Anyway)

Well, I got to 355 and had enough.  DMB says 100 simulations is needed  to understand a players true baseline so we are well above the minimum.

Before our upcoming ATB 17 (!) season just a few months away, I wanted to tackle another resim baseline.  Recall, in the past two years we have tweaked the era and added players, and I figure every few years it will be valuable to re-baseline the league.  I purposely did not wait until the MLB season so that each year there is some mystery to how the newest players will perform.

The full report is here and the player cards are updated as well.  Below are some highlights for the newest players:

Batters worth a Look
.271 / .366 / .451, 31 HR - Miguel Cabrera (2013)
.270 / .382 / .395, 47 2B - Joey Votto (2012)
.243 / .345 / .422, 38 SB - Mike Trout (2013)

Batters to Pass On
.255 / .312 / .381, 19 HR - Andrew McCutchen (2012)
.228 / .299 / .358, 18 HR - Paul Goldschmidt (2013)

Pitchers worth a Look
1.85 ERA, 0.86 WHIP, 1.92 Adj ERA - Koji Uehara (2013)
2.27 ERA, 1.03 WHIP, 2.49 Adj ERA - Fernando Rodney (2012)
2.69 ERA, 1.08 WHIP, 2.63 Adj ERA - Jacob McGee (2012)
3.51 ERA, 1.25 WHIP, 3.62 Adj ERA - Clayton Kershaw (2013)
3.35 ERA, 1.24 WHIP, 3.66 Adj ERA - Max Scherzer (2013)

Pitchers to Pass On
3.59 ERA, 1.25 WHIP, 3.91 Adj ERA - Glen Perkins (2013)
4.07 ERA, 1.34 WHIP, 4.53 Adj ERA - Mark Melancon (2013)

Adjusted ERA is my new stat for Resim pitcher data.  It is the pitchers ERA adjusted to reflect the defense behind him.  The adjustment is calculated using the defensive metric values from studies I've shared over the years.  Each position can add or subtract a few percentages points of ERA.  Add up the positions for a while team, weighting them for the importance of that position (ex. SS Range more important than C range).

A few examples:

Sandy Koufax (1966) looks pretty good by previous resim measures.  4.08 ERA, 1.31 WHIP and 189 hits in 198 innings.  That is definitely worth a draft pick.  Digging deeper though, his defense was among the best in the resims.  Eddie Collins, Ron Santo, Bill Dickey, and Omar Vizquel held down the infield (along with Av/31 Harmon Killebrew at 1st) while  the Roy Thomas, Ed Delahanty, and Paul Waner roamed the outfield.  That is quite good and as a result many runs were saved.  Accounting for this gives Koufax a gaudy 4.63 ERA had the defense been average.

At the opposite end of the spectrum is Randy Johnson (2001).  He simmed well with a 3.31 ERA and 1.28 WHIP but his defense was suspect in several places.   Adjusting for it brings his ERA down to 3.06.

Cherry picking some interesting cases:
3.37 adjusted to 3.16 - Roger Clemens (1986)
3.44 adjusted to 3.18 - Johnny Podres (1957)
3.30 adjusted to 3.67 - Jerry Reuss (1980)
4.21 adjusted to 4.62 - Bill Foster (1932)
3.34 adjusted to 3.60 - Jason Schmidt (2003)

The vast majority of adjustments are upward.  Since the auto draft was used, it is important to remember that the DMB AI relies heavily on defense and the the coputer fields a much better fielding team than ATB managers normally would.  As a result, only 15% of the pitchers saw their ERA improve.

Finally, it is interesting how close some of the players results are to the 500 sims from a few years ago.  Randomly picking some players to compare:

Adrian Gonzalez 2009
355 - .223 / .306 / .407
500 - .224 / .305 / .405

Benny Kauff 1915
355 - .287 / .375 / .435
500 - .284 / .373 / .428

Edgar Wesley 1925
355 - .266 / .328 / .465
500 - .269 / .332 / .465

Honus Wagner 1908
355 - .296 / .346 / .455
500 - .302 / .345 / .464

The vast majority of players remain unchanged, but there are a few curiosities:

Ted Williams 1941
355 - .321 / .451 / .560
500 - .327 / .475 / .570

Oscar Charleston 1924
355 - .324 / .407 / .604
500 - .338 / .442 / .625

Lou Gehrig 1927
355 - .299 / .407 / .622
500 - .303 / .422 / .632

Those are the three biggest changes and it is odd that they are all big names and odd they all suffered instead of improved.  I don't have an explanation.  Among full time players, 95.0 % of the time the difference OPS between the 500 resims of two years ago and this years 355 resims in were within 2.5% of each other.  But a handful of players suffered relatively wild swings.