Monday, March 26, 2007

Unobjectively Statistical

In preparation for the highly anticipated Haiku Season Preview (where last year, you learned that the Tigers would lose 92 games), I've been hard at work at CP labs, crunching numbers for you, our loyal reader.

Although I don't want to jump to the conclusion, ruining the 14,000 words between now and the end, I was sort of surprised by what the numbers told me. Not only won't this team be historically bad, they might not, well... let's look at the numbers.

Keep in mind that these are just thumbnail estimates, mostly pulled out of my backside and shaped a bit to make them look normal. If you think I've gone wrong anywhere, just pop in a comment.

First, about the stats. Runs Created (RC)is a stat created by Bill James. It aims to use basic stats, assembling them in such a way so as to estimate what a player actually contributed to the team's bottom line. No, it's not the best system, but it's a simple system, especially for the purposes of what I'm doing. If you want more info on it, there's a ton here.

One of the beautiful things about RC is that it can be as complex or as simple as you need. I'm keeping things simple and using the hacksaw version. If you multiply a player's On-base times his Slugging times his ABs, you're pretty close to the more elaborate forms of RC.

To estimate the player's OBP and SLG, I'm bringing back CRAP (Chris' Random-Ass Projections). Basically, I eyeball the numbers, looking for trends, expectations, etc. My guesses are all in line with some of the other various projection systems out there. (You can check most of them out at fangraphs.com)

To estimate ABs, I started out by creating a depth chart at each position. At catcher, for example, I expect Brian Schneider to get 70% of the starts. Jesus Flores gets 25% and Robert Fick plays 5%. You can finagle around with the percentages for the backup, but their limited PT isn't going to affect the bottom line more than a run or two here or there.

To get the total of ABs for each position, I assigned each roster slot a lineup spot and divvied up the ABs according to how many that lineup slot typically gets in a season. So second base, by virtue of having leadoff hitter Felipe Lopez, gets more ABs than Catcher, which is slotted at the bottom of the lineup.

The chart should be self-explanatory, even if it's not the most readable. Next to the position total, I've included the '06 position total for rough comparison though the AB totals won't match because the fielding positions don't necessarily have the same batting order position.



They're due for big downgrades at first, center and left, giving the Nats lineup an RC total of roughly 697 runs. They scored 746 last year, so they're roughly 50 runs worse. With the loss of Soriano, Johnson and the additions of Guzman and Logan to the lineup, that's a number that passes the smell test.

What about the other side of the ball? This is where I'm having problems. Projecting pitchers is a fool's errand. It's even more difficult to judge playing time because of injury, and ERA is dependent on a lot of factors outside the pitcher's control. So, I'm reallllly pulling numbers out of my butt, and the conclusion I've reached? Scary stuff...


I played it conservative with Patterson's and Hill's IP because of their injury histories. And I tacked on a bunch of 6.00 (ie: replacement level) innings on the back of the pen and rotation to account for the filler that will inevitably trickle through the roster at one point or another. I'm expecting about 50 more IP out of the starting pitching and the team total IP would've been middle of the pack last year.

The total earned runs allowed is 790. That's 13 fewer than last year, and that's assuming a so-so injury-filled season from Patterson. Lost in all the noise about this year is the fact that the Nationals had ONE starter with an ERA below 5 last year, Mike O'Connor's 4.80 -- and that was in just 20 starts. The pitching was already on the verge of being historically bad. None of the new guys brought in this year are likely to beat out what Ortiz and Armas did last year, but it's going to be just as hard for them to be worse.

If they get anything at all from Patterson, the pitching is going to be vastly better. (Plug Patterson in at 180 IP of 3.50, for example, and that's another 17 runs off the board)

If we want to be even more conservative, let's plug in a 7.00 ERA for those starting pitcher replacements -- the team is only about 10 earned runs worse than last year's version.

The lesson of the day: When you're at rock bottom, it's hard to go lower. That projected 4.90 ERA is .13 runs better than last year, but it would still be good enough for 16th -- last! -- in the NL.

You've noticed that to this point that I've only been discussing earned runs. We need to make an adjustment to those totals to figure out the total runs allowed.

I'm going to fudge it a bit, but the average NL team allowed .27 unearned runs per game last year. The Nationals were quite a bit worse (.35) last year. I think the defense will be less error-prone this year, so we'll split the difference. 162 games times .31 unearned per game gives us an additional 50 unearned runs.

We'll add that back into the earned total, giving us 840 runs allowed. Last year's team gave up 872. The pitching is thirty runs better?!?

It's possible to go even lower if you make subjective adjustments to account for what should be a better defensive team. Just to pull more numbers out of my butt...

Catcher is probably around 5 runs better with more attention to base stealers. Take away 10 runs for Dmitri Young's stone hands and immobility. Add 10 back in for Felipe Lopez' range. Zimmerman's a wash. Shortstop should see a marginal improvement, so we'll plus it up by five. Left is likely a wash. Right should be an improvement because Guillen was mediocre, so let's say another five. Then at center, despite Logan's problems, he's probably worth 15 or more runs over Church and Byrd from last year. Add it up and you're at a +30 runs on defense over last year.

So what're we left with?
Runs Scored: 697
Runs Allowed: 840
Runs Allowed with Defensive Adjustment: 810

You can estimate win totals with runs scored and allowed, thanks to our good friend Pythagoras.

Winning % = (Runs Scored ^2) / ((Runs Scored ^ 2) + (Runs Allowed ^ 2))

Plug the numbers in, and you get a 66-96 record, 5 wins worse than last year.

If you prefer the subjective defense method, you get 69-93

Bad, for sure, but nowhere near historically awful. I'll take it!

  • If you want to do your own tinkering, in general, every 10 runs saved (or scored!) adds another win to the bottom line. So if Patterson does put up that 180 IP of 3.50 ERA, it's another two wins.

    The question many of you'll ask first is what happens if you plug Church into center. The offense projections assume that Logan gets 80% of the PT in center, Church 15%. If those flip, it's a 22-run improvement on offense. When you adjust for defense, most -- if not all -- of that gap closes. That's also assuming that Logan can get his OBP and SLG up to .310 and .345 respectively

  • 15 Comments:

    Post a Comment

    << Home