THE BOOK--Playing The Percentages In Baseball

Yeah, I don’t know how much I “care” about the All Star game.  But it’s always a good opportunity to take a moment and reflect on the current best players in baseball.  So on that level, I like it.

I should have caught the Zimmerman fielding thing, but missed it.  Caught it for Kelly Johnson.  I’m ultimately just not a huge fan of the Oliver/CHONE fielding stats, but they are probably close more often than they completely miss it.  Something’s better than nothing.  Maybe I should pester Steve for his UZR projections.
-j

I would love to just abolish the all star game altogether. maybe set up a non-HR Derby skills competition, but I’m not sure what all you’d do.

why not just take their weighted average WAR/game season to date, and WAR/game from previous 3 seasons (or 2). With the weight on this season being G/162, and 3 previous years as (162-G)/162. That way you can capture UZR, and base it on historical performance, rather than the projection systems that I think would include things like aging and other factors that are not based on what they actually did.

Aging is part of how we predict future performance.  My goal here is to pick the best players.  Past performance is part of it.  But it needs to be interpreted, regressed, etc.  That’s what using a rest of season projection does for us.

Re: UZR--I’m going to re-do this with Steve’s uzr projections for NL tonight, and will use them for AL projections too. 
-j

Maybe you could argue that projecting one day (or one month right now) into the future requires a different aging curve than projecting an entire season’s worth of games into the future.  It’s probably not going to matter much, though.

There seem to be some really odd choices in there. Andres Torres is a better player than Andrew McCutchen? Is Adam Dunn really superior to Adrian Gonzalez (who doesn’t even rate in the top 5) at 1b?

I wonder if any of these projections are not park adjusted (or projected neutral than translated back into park-influenced versions to more closely match what will actually be recorded)?

Sky/7,

All of the projections are supposed to match the park environment of the hitter.  That’s why I park-adjusted the offense RAA numbers.

BrianK/6,

Torres’ line is all fielding.  Depending on how real you think it is, it can push his value up or down a lot. 

As for Gonzalez, that’s a screw-up.  I don’t know how it happened.  Complete user error.  He is at +41 runs in my spreadsheet, right behind Fielder, 3rd among NL 1B’s.  I must have accidentally pushed him into the AL group somehow--maybe in anticipation of a trade?
-j

Post updated.  Gonzalez now included, and Steve’s UZR projections are now given 50% weight in the fielding calculations.

Cool beans with Gonzalez. Still though, I think you are weighting defense a bit too much for players who don’t have a enough data to really get a feel for what kind of defensive player they are (Torres, McCutchen). I’m not questioning the results. I would have just chosen a different methodology. Fun exercise though. Thanks.

Oliver and CHONE’s fielding projections include minor league data, so I’m not sure that this is true. 

Also, while it’s spread over a bunch of seasons, we have almost a full season’s of MLB innings on Torres, and his career UZR/150 across all outfield positions is +22 runs (+20 in CF, where a bit more than half of his innings have been).  McCutchen has a similar MLB innings totals, but his career UZR/150 is -6 runs (admittedly all in CF). 

If I were picking players, I’d certainly take McCutchen over Torres.  But Torres might be a poor man’s Franklin Gutierrez.  Or maybe not so poor man’s--both rate as +14 run players vs. other CF’s.
-j

"Still though, I think you are weighting defense a bit too much for players who don’t have a enough data to really get a feel for what kind of defensive player they are (Torres, McCutchen).”

The defensive projections are regressed so that takes into consideration the fact that you might not have much fielding data on a player.  So as long as the regressions are in the proper ballpark, weighting them 50% for every player is fine.  IOW, if a player is a non-regressed +10 (per 150) in fielding but it is only based on 1/3 of a season, it is going to get regressed 75% so the number used for the projection is only +2.5.

I love a list like this.  As far as the All-Star Game, many people express the opinion that the participants should be the best players and not those who are having the best first halves.  Of course in reality it is sort of a combination, as a guy with a great reputation (i.e. he is a great true talent player) who is having a poor first half often makes the team, while an unknown (often a player who is not a great or even a good true talent player) who is having a great first half almost always makes the team.

I think the idea that only truly good or great players, regardless of their first half, would make the team is a ridiculous notion from the standpoint of the average fan for obvious reasons.  They could tolerate a popular and historically great player making the team even though he is having a poor first half, but they are not going to tolerate a “sabermetric darling” (say, Polanco from a few years ago) who is having a poor first half making the team. Plus, their (the fans) notion of a great player is nowhere near the sabermetric notion of a great player.

The ASG is for the majority of the fans, not the .1% of the fans who are hard-core sabermetricians or sabermetric aficionados.  Plus, even the fans recognize that they are voting for players who are having great years.  And the fact that there is voting by the fans in the first place automatically means that it is a popularity contest.  As far as I know, it does not say on the ballot, “Vote for the players whom you think are the best (in true talent) or have the best projection, regardless of how they are doing the first half of the season.”

Regardless, the majority of the time, the fans and the sabermetrics will pick the same players as #1 at their position. If/When the fans don’t, they generally pick the 2nd or 3rd ranked guy ... And all 3 are really good players.

Occassionally, there’ll be a Ben Zobrist that gets hosed. But still, both methods generally pick the same dudes.

It’s not as if the sabermetric method comes up with a completely different set of names from the fan vote.