Those of you paying attention in the off-season may recall the FederalBaseball.com W.A.R. Effort, an attempt to project out the Nationals' 2009 record using Wins Above Replacement (W.A.R.) calculations. We argued about lineups and playing time, looked at projections for hitters and pitchers, then plugged it into a big spreadsheet. The answer we got was that the Nats should manage around 75 wins in 2009, with a 99% chance of getting 61 or more wins. As we're now on a pace for 50ish wins, it seems that we were wrong. Quite wrong. Exceptionally, overwhelmingly, soul-ravagingly wrong. How did we miss so wildly?
After the jump, we'll take a look at what we got wrong (bullpen, fielding), what we got right (almost nothing), and what still doesn't make any sense (the Nats are worse than my math can calculate). Remember to show your work for partial credit!
But first, a quick aside...
Let me outline the idea behind W.A.R. The basic concept of "replacement level" is a player who's freely available on the waiver wire or from AAA--someone a team can pick up on the cheap to replace an injured starter. There's some debate in the stat community about exactly what replacement level means, but figure that a team made up of "replacement level" players is a AAAA team, and will win about 48 games in the NL (starting to look familiar?). Anyhow, the idea is that you compare a player's hitting/pitching stats and playing time (or their projected stats and playing time, if you're trying to look ahead) to "replacement level" to come up with a number of extra runs scored/saved. These runs turn into "wins above replacement" (or "below" replacement for some... I'm looking at you, Daniel Cabrera!). Add up the total W.A.R. for everyone on the team with the baseline wins for a replacement-level team, and you get the wins on the season. Of course, this is an average number, and the number of wins will probably be higher or lower by some amount.
Are we talking about the same team?
Back in the final W.A.R. update, I had a projection of 76.9 wins on average. Even I was a little suspicious over whether we could pull that off, but I figured we were a lock for at least 70. Adding up the projected playing time, plus projections for hitting, pitching, and fielding, I figured we were looking at a total of 28.4 Wins Above Replacement. Of those, 19.8 WAR came from the position players, almost all from hitting (I had us losing about 0.1 WAR from fielding, and gaining something like 0.05 WAR from baserunning). The other 8.6 WAR came from the pitchers, with 6.5 from the starters and 2.1 from the relievers.
We can check out our current WAR values at a stat site like fangraphs, which calculates WAR (it's a counting stat) as the season goes along, based on playing time, hitting, fielding, and pitching. As of July 13, our position players have generated a total of 8.7 WAR, 12.0 from batting and -3.3 from fielding. Compare to my projection of 10.6 WAR after 87 games, all from hitting: we're actually hitting almost 1.5 wins better than I'd projected, but our fielding is more than 3 wins worse! That puts us almost 2 wins off of the projected pace already. As for the pitching... Well, so far we only have 3.2 WAR from pitching, 4.0 from the starters and -0.7 from the bullpen. I was projecting 4.6 WAR from pitching at this point (split about 75-25 between rotation and pen), so there's another 1.5 wins gone! Even so, that puts us at 11.9 WAR (versus my prediction of 15.3 after 87 games). That's 38 wins after 87 games. The Nats have 26.
What's going on, here? We already know that the Nats' "Pythagorean" record has them at 33 wins, so they're already at least 7 games' worth of unlucky. WAR is also a calculation of average number of wins. You'd expect to be above or below that number most of the time. In fact, if our inputs to the WAR spreadsheet were right, we'd expect a record of between 66 and 81 wins 90% of the time (10% it would be either more than 81 or less than 66)--call it plus or minus 8 wins, which would be plus-or-minus 4 wins after 87 games. But even the Pythagorean 33 wins is just outside that 90% confidence interval with 38 wins from WAR. The Nats are either unlucky to a statistically-unlikely degree, or they stink in some way that Wins Above Replacement can't model. I don't know what the answer is, but it seems to me that both factors probably play a role (WAR calculations are an inexact science at best, and who's more snakebitten than the Nats?).
Did you get anything right?
Who, me? The guy who predicted 450-500 PA for Jesus Flores, Elijah Dukes and Lastings Milledge? I'm amazed I got as close as I did. Happily, several Nats did much better than I predicted:
- Even injured, Jesus Flores has been worth 1.0 WAR in limited playing time--I had him at 0.4 WAR over the season. Moral? Have a season-ending injury while your bat is hot!
- Zimm the Elder is worth his moneys. His hot bat and hotter glove are worth 3.2 WAR so far this season (ahead of my projection of 2.6 at this point).
- Willingham has been a pleasant surprise: his 2.2 WAR is second only to RZim thanks to a hot bat and league-average glove (I had him at +1.0 WAR on the season with limited playing time).
- NJ is about the only thing I've gotten right, with 1.5 WAR so far against my projected 1.6 (his fielding slightly worse this season than I projected).
- Zimm the Younger's +1.7 WAR in a half a season is as good or better than I was projecting for any starter over the whole season. Stammen's +1.0 ain't bad, either. Lannan's +0.7 is off the pace (+0.9) I had for him.
Sadly (but predictably), many Nats did worse:
- Dunn is not like a box of chocolates: +2.9 WAR with the bat, -2.2 WAR with the glove...
- How did I ever believe Cabrera would be worth 1.7 WAR over the season? Instead, he's cost the Nats -0.4 in limited playing time.
- Elijah Dukes was another disappointment: -0.5 WAR in limited playing time, when I was quite rationally calling for +2.5 over the season.
- Beimel is best in the bullpen at +0.1 WAR... everyone else is replacement-level or worse. But you knew that.
What have we learned?
Aside from, "don't go to Vegas with Doghouse"? I think we may be running into limitations of the WAR approach: it may work best for a team this isn't too bad in any one area. You can assign a win value to each part of the club and add 'em up. But if you have a team with, say, really bad defense and a really bad bullpen, it may have a disproportionately bad effect on the number of wins. You can't just add up the bullpen, rotation, and hitters anymore--some awful synergy takes over and applies the '62 Mets Factor to all of your results. In short, it seems that the badness of the 2009 Nats defies mathematical analysis!
I'll revisit this at the end of the season when we have full-year performance and playing time, and see if the WAR calculations hold up any better then.