The difference between an intelligent GM... and Sabean

Er, what about him? He’s a very, very good fielder?

within the margin to not get the boxers in a twist

For me it’s about the collective group of signing – it’s all the same thing. Old, washed up veteran who is probably worth his contract if he’s playing over replacement player, but he’s actually playing instead of an above replacement player (Ishikawa, for instance), or even a legitimate starter (Lewis), and on the collective whole we ended up spending an amount of money that could have gotten us a real upgrade, instead of a bunch of marginal ones that add up to much less than said “real upgrade”…

So that’s why my boxers are in a twist about. ;)

So you usually shouldn’t try. One WAR costs $4-5M in free agency, so $3M normally won’t buy an average player. For $3M you’re either getting a gritty veteran who’s barely over replacement (if you’re lucky; someone like Rich Aurilia the last two seasons), or a guy who might be very good but who has serious question marks, probably because he’s coming off a significant injury (somebody like Carlos Delgado this year, for example). A good GM should almost never sign the first type of player unless no plausible in-house option is at hand. The second type is high-risk/high-reward and might sometimes be justified.

In short, Aubrey Huff is exactly the sort of player you’d expect to get for $3M. The problem is that the $3M should not even be expected to produce much better than slightly above replacement-level performance, so it’s essentially wasted. The charitable view would be that Huff is a lottery ticket, probably worthless but holding some chance of a big payoff if he recaptures that 2008 magic.

I like judging the process, and the process sucks. Looking at results in hindsight is a poor way to judge how “the process” will perform in the future.

He may bounce back. Maybe he was injured, or maybe he’ll get new contacts or lasik or something. But 4 walks and 3 extra-base hits (just 19 total hits) in 125 PAs is not a fluke. That’s a guy who legitimately sucks.

There are two aspects to small sample size in baseball, and they tend to get conflated for the precise reason that it really is difficult to distinguish them, at least for an outside observer. One is actual luck; for the same reason if I try long enough I can surely toss a coin and get heads five times in a row, a .300 hitter can go 0-10 over some period strictly by chance. But there’s a second aspect: a player’s ability — his “true” chance of success — can also fluctuate because of injury or a newborn child or whatever. Winn’s 2009 as a RHH can not be explained by the first one; it wasn’t bad luck (or, to be precise, it’s extremely unlikely that it was random). It could be the second one, however; he might have sucked for some reason that will be corrected.

For his sake I hope he does bounce back. It would be pretty sad if he finally played for a playoff team but then got DFA’d before the end of the year.

Players don’t do this much if at all. If a team isn’t willing to give him the arbitration raise he thinks he has earned why would he re-sign with that team for less when he could re-sign with a different team that didn’t “screw him” instead?

High 700 OPS looks good for Garko for his career, but if 75% of the time he’s a .750-ish OPS playing 1B, then that’s not so good for your offense.

It’s also worth noting that 65% of the time, Garko is an .000 OPS player! If that’s the Garko that shows up, you’re in real trouble.

Most splits are not predictive. It’s sample size. You’re ignoring better data and choosing to use worse data in it’s place, because you’re looking to confirm your beliefs. L/R splits are fine, if you have a big enough sample. They make sense. H/R splits are almost always meaningless. Seeing how a guy hits in a different part of the lineup is meaningless. Seeing how a guy hits when playing a different position is mostly meaningless (has some merit if you’re talking about C, but that’s it). Use the full data set. It’s a much, much, much better estimate of a guys talent level, and these splits you’re looking for (other than L/R) don’t change talent levels, which are what we’re trying to estimate to predict future performance. They’re just confirming biases. Use the full data set – it’s the only objective way to do it, unless you have a really, really, really good reason why a split is meaningful AND the sample size is very large. Seriously, read The Book blog (or The Book itself), read Fangraphs (I know you do), these points come up. Splits are meaningless, and if you use them to make decisions for the future, you’re doing it wrong.

88 wins was not entirely luck. We got the pitching to back that up and the pitching and offense should be even better in 2010. An average offense with the defense we had last year should net us 90+ wins.

I mean, no, it wasn’t entirely luck, we were seemingly more of an 82-84 win team, a definite improvement on years past, but still not 88 or a playoff team or anything meaningful when it comes to winning the World Series, which should be our goal. Our pitching is great, but who’s to say it’s as good as it was last year? All it takes is a key injury and then what are we gonna do? What if Wilson regresses, or Lincecum (it’s more likely he does worse than better)? And who’s to say the defense is the same? Old guys get worse. Ishikawa to Huff is a clear downgrade. Winn is gone. And we’re a far cry from an average offense right now.

Look, there’s a reason people do projections. You can’t just say “if our offense is average we’ll win 90 games because we allowed X number of runs last year”. The objective, correct way to do this is to estimate talent levels and project each individual and add it up at the team level. We’re not a 90 win team if you do that. Not all that close, either.

By “not that much below average” I was sort of thinking “significantly better than Adam Dunn.” I was envisioning he could be in the -5 range, but you’re right, even that might be generous.