team offense projection: ZIPS

in case you missed it, the much-anticipated cardinal ZIPS projections were posted on saturday. pretty lively discussion ensued in this thread, including some sharp exchanges about whether projections are or aren't a complete waste of time. i'd just as soon not rekindle that argument; it gets tiresome pretty quickly. if you think projections don't have any merit, then today's post isn't for you. just move on and come back another day. but fire away if you want to discuss the accuracy of particular projections for individual players. if you think that, say, chris duncan is more likely to slug .557 than the .457 to which ZIPS projects him, have at it --- and take a shot at explaining what factor(s) you think ZIPS is overlooking.

what i'm gonna do today is use the individual player projections as the basis for a guesstimate about where the cardinals stand as a team. pretty straightforward process: i'll make a few assumptions about roles and playing time, adjust the ZIPS figures accordingly, and crunch a few numbers. i did much the same thing last off-season using PECOTA projections and concluded that the cardinals would score about 780 runs (that's the title of the post); they actually scored 781. here's the full guesstimated line, vs the team's actual performance:

	AB	H	2B	3B	HR	BB	|	AVG	OBP	SLG
projected	5535	1480	288	26	171	548	|	.267	.340	.421
actual	5522	1484	292	27	184	531	|	.269	.337	.431

i may not be as lucky this time around, and ZIPS may not be as on-target this year as PECOTA was last year; that's all understood. it's also understood that many of the assumptions on which i based last year's forecast (viz. spivey, bigbie, and deivi cruz on the roster) were wildly off-base; no need to point all that out. even with all those caveats, this exercise helped us fumble toward a damn accurate forecast last year; might as well take another shot at it. in fact, i'll probably run the same test again when the PECOTA figures come out (they're due next week), and adjust assumptions as the roster changes and roles are clarified during spring training.

so let's get started. to begin with, we can safely assume (based on current information) that the following 13 players will constitute the core roster:

infielders: pujols, kennedy, eckstein, rolen, miles, spiezio
outfielders: duncan, edmonds, encarnacion, taguchi, j-rod
catchers: molina, bennett

if we take ZIPS' projections for those 13 guys, toss in 300 at-bats for the pitchers (using 2006 league averages for pitcher batting), and add it all up, here's what we get:

			AB	H	2B	3B	HR	BB	|	AVG	OBP	SLG
pujols			566	181	37	1	48	99	|	.320	.423	.643
edmonds			401	105	28	1	24	71	|	.262	.372	.516
rolen			481	138	37	2	22	59	|	.287	.371	.509
duncan			455	121	22	1	21	55	|	.266	.346	.457
encarnacion			442	117	22	2	12	28	|	.265	.315	.405
kennedy			440	120	23	3	4	34	|	.273	.332	.366
eckstein			514	145	20	2	3	39	|	.282	.346	.346
molina			423	98	21	0	8	24	|	.232	.273	.338
miles			426	117	17	3	3	26	|	.275	.319	.350
rodriguez			333	90	20	3	13	35	|	.270	.344	.465
taguchi			305	84	16	1	3	21	|	.275	.324	.364
spiezio			277	68	13	2	8	31	|	.245	.328	.394
bennett			174	39	8	0	2	16	|	.224	.292	.305
pitchers			300	45	6	0	2	10	|	.150	.177	.190
TOTAL			5537	1468	290	21	173	548	|	.265	.338	.419
2006			5552	1484	292	27	184	531	|	.269	.337	.431

per bill james' basic runs created formula (available at this link), this offense would be expected to score about 784 runs --- roughly the same number that last year's offense scored. (note: i added 60 team hbps to the formula, so if you're checking my math or playing around with these figures, include a "team hbp" line in your spreadsheet.) but how can this be? our projection calls for no change in the team's on-base pct, and a 12-point drop in slugging average; shouldn't the number of runs decrease? yes, it should --- but it doesn't because last year's cardinals undershot their runs-created target. the '06 team "should" (per runs created) have scored 804 runs but didn't, due either to random variance or to a deficiency in some offensive component that the runs-created formula doesn't measure. for whatever reason, the whole offense last season was 23 runs worse than the sum of its parts. so even though the 2007 offense does, in fact, project to be slightly less potent than the 2006 version, with a little more efficiency it might score just as many runs.

onward after the jump . . . .

our projected roster makes some patently unlikely assumptions. for starters, we've got a full season's complement of at-bats, but only 13 position players doing the batting; that's bogus. last year 21 (or so) different position players took at-bats for the cardinals, and in most years the number's even higher. there will be injuries, trades, callups; gotta account for that. so we'll add a generic "replacement player" line, which will represent the at-bats that will be taken by this year's iterations of timo perez, skip schumaker, and jose vizcaino. if you're feeling bitter toward the ownership, call it the "DFA line." i'll set the following wild-ass-guess parameters for this parcel of plate appearances: rate stats of .260 / .315 / .370, with one double per 25 at-bats, one triple per 200 at-bats, one homer per 50 at-bats, and one walk per 12 at-bats.

we also have to adjust some of the individual projections to reflect each player's likely role and playing-time allotment. ZIPS might think john rodriguez merits 333 at-bats, and you or i might give him that many, but he's not going to play that much. likewise, juan encarnacion's estimate of 442 at-bats is probably too low; he's been well above that in each of the last five seasons. eckstein will probably surpass his ZIPS-projected playing time, so we'll bump it up; correspondingly, miles probably won't get 425 at-bats.

with these adjustments, the projected stat sheet comes out like so:

			AB	H	2B	3B	HR	BB	|	AVG	OBP	SLG	RC/27
pujols			566	181	37	1	48	99	|	.320	.423	.643	10.7
edmonds			401	105	28	1	24	71	|	.262	.372	.516	7.2
rolen			520	149	40	2	24	64	|	.287	.371	.509	7.0
duncan			455	121	22	1	21	55	|	.266	.346	.457	5.8
encarnacion			525	139	26	2	14	33	|	.265	.315	.405	4.6
kennedy			440	120	23	3	4	34	|	.273	.332	.366	4.4
eckstein			560	158	22	2	3	42	|	.282	.346	.346	4.3
molina			423	98	21	0	8	24	|	.232	.273	.338	3.2
miles			300	82	12	2	2	18	|	.275	.319	.350	4.1
rodriguez			150	41	9	1	6	16	|	.270	.344	.465	5.8
taguchi			305	84	16	1	3	21	|	.275	.324	.364	4.4
spiezio			277	68	13	2	8	31	|	.245	.328	.394	4.5
bennett			174	39	8	0	2	16	|	.224	.292	.305	3.1
DFAs			144	37	6	1	3	12	|	.260	.315	.370	4.2
pitchers			300	45	6	0	2	10	|	.150	.177	.190	1.1
TOTAL			5540	1468	289	20	172	546	|	.265	.338	.419
2006			5552	1484	292	27	184	531	|	.269	.337	.431

after all that tinkering, our bottom line has hardly changed; this permuatation of the roster projects to 781 runs created, or 3 runs fewer than the unadjusted estimate. you'll note i added a column to this table, runs created per game (RC/27); it suggests that the cards have four good offensive players, one decent hitter on the bench, two terrible-hitting catchers, and about half a dozen interchangeable parts.

to check our bottom-line conclusion, let's feed some numbers into david pinto's lineup toy. it's a much simpler tool, only measuring the production of the 9 starting players. according to it, this lineup

1. eckstein
2. duncan
3. pujols
4. rolen
5. edmonds
6. encarnacion
7. kennedy
8. molina
9. pitcher

would average 4.86 runs a game, or 787 over a full season if they played all 162 games. of course, that lineup won't play all 162; lesser hitters (e.g., taguchi and spiezio) will be taking several hundred at-bats in relief of rolen edmonds and duncan, and so we'd revise our estimate downward by 15 runs or so. but at a rough-eyeball level, the lineup toy gibes with our ZIPS guesstimate. so, too, does the Diamond Mind simulation exercise i cited last week, which employed data from another projection system (CHONE); in that set of trials, the sim-cardinals averaged 769 runs per season.

tomorrow i'll run this same exercise for the pitching staff; it's far more complicated, because there are so many more variables involved.