Table of Contents

Thursday, June 30, 2011

Who is an All-Star?

Tuesday, I had a piece run at Hardball Times that used projections to pick the participants in the All-Star Game.  The premise is to get away from just using 2-3 month samples to select our players, and instead use projections to pick the players who have the most talent.

One alternative to doing this, which came up in a round-table discussion between myself and Dave Cameron at FanGraphs on this subject, is to instead choose players based on past calendar-year performance.  The advantage here is that it's a bit more straightforward: assuming we can properly measure everything that happens on the field (that's a big if!), all a player needs to do is produce better than the other players in the league at his position to get the nod.  We don't have to worry about aging curves and regression and luck.  Just produce, and you'll be rewarded.

Here's a comparison for how the three approaches would differ (and be similar), comparing my projection-based roster to one according to FanGraphs' WAR Calendar-Year splits, as well as Vince Caramela's first-half All Star selections.

American League

PositionTrue Talent All StarPast-162 G All StarFirst-Half All Star
CJoe MauerVictor MartinezAlex Avila
1BMiguel CabreraAdrian GonzalezAdrian Gonzalez
2BBen ZobristIan KinslerHoward Kendrick
3BEvan LongoriaT-Adrian Beltre & Evan LongoriaAlex Rodriguez
SSYunel EscobarAlexei RamirezJ.J. Hardy
LFJosh HamiltonT-Brett Gardner & Josh HamiltonJosh Hamilton
CFCurtis GrandersonCurtis GrandersonCurtis Granderson
RFJose BautistaJose BautistaJose Bautista
SPFelix HernandezJustin VerlanderJustin Verlander
RPMariano RiveraJonathan PapelbonJonathan Papelbon

While you see different names in many of the positions, if you go back and look at my THT article, every single one of the FanGraphs' WAR picks are within the top tier of the projections.  The same is true if you look at the FanGraphs' WAR leaderboards.  Most of the time, even when there's disagreement, the true talent selection is very close.  Heck, at 3B and LF, they were in a tie with someone else!

On the other hand, we do see some putative small-sample selections in the first-half All Star picks.  Alex Avila's an interesting player and has had a great first half, but he ranks 7th by projections and is in a tie for second by the past-162 splits.  Howard Kendricks' even more extreme, ranking 11th by projections (behind even guys like Nishioka and Jason Kipnis) and 4th by past-162 game WAR.

National League

PositionTrue Talent All StarPast-162 G All StarFirst Hall All Star
CBrian McCannBrian McCannBrian McCann
1BAlbert PujolsJoey VottoPrince Fielder
2BChase UtleyRickie WeeksRickie Weeks
3BRyan ZimmermanChase HeadleyTy Wigginton
SSTroy TulowitzkiTroy TulowitzkiJose Reyes
LFMatt HollidayMatt HollidayMatt Holliday
CFAndres TorresMichael BournMatt Kemp
RFJayson WerthMike StantonLance Berkman
SPJosh JohnsonRoy HalladayRoy Halladay
RPMike AdamsCarlos MarmolCraig Kimbrel
There were two fairly major disagreements here between the projections and the past calendar year picks.  First, Chase Headley at 3B by FanGraphs' WAR, who ranked 10th in the projeciton rankings (Zimmerman ranked 2nd by FanGraphs' WAR).  The second was Michael Bourn in CF, who ranked 12th in the projection rankings (Torres ranked 6th in the WAR rankings).  Beyond that, again, there was good agreement pretty much across the board.

Among the first-half All Stars, the biggest outlier pick is Ty Wigginton, but Chase Headley would be my selection based on the extra playing time this season if nothing else.  You can also complain about Lance Berkman, although he was effectively in a tie with Jason Werth for the projection's choice...and a lot of his merit comes down to how good (or bad) you think his fielding is.

In the end, I have no argument with people who want to use past-162 game splits.  In most cases where they disagree with projections, they do so within the margin of error in either measurement or projection (or both).  Where they disagree, however, I side with the projections.  They seem more obviously to be selecting players with the best talent, and those are the players that I prefer to see make it to the All Star Game.

But insisting on using first-half statistics to justify an all-star game roster?  There, I just don't get it.