clock menu more-arrow no yes

Filed under:

team pitching projection: ZIPS

New, 99 comments

before we forge ahead with more ZIPS talk, some links:

so enough of all that. player projections: you know em, you love em, you can't live without em. yesterday, using the ZIPS forecasts and some common-sense assumptions about playing time, we pegged the cardinals' expected scoring output for 2007 at about 780 runs. today we'll run a similar exercise to see how the pitching staff shapes up.

before we do, let's revisit last off-season's pitching-staff forecast, based on PECOTA figures. while it wasn't quite as spot-on as the projection for the 2006 offense, it was accurate insofar as it warned, correctly, of a big increase in runs allowed. my accounting placed the bloat at 79 runs, which ended up being 50 runs low; but then, nobody foresaw the utter collapse that befell stl's pitching staff. on the contrary, most prognosticators had the cardinals remaining at or near the top of the league in era, where they'd finished in each of the previous two seasons. PECOTA at least raised a gigantic red flag vis-vis the cardinal pitchers --- and did it before a single pitch had been thrown:

IP H BB SO HR H9 BB9 SO9 | ERA WHIP RUNS
projected 1445 1462 473 959 156 9.1 2.9 6.0 | 4.08 1.339 712
actual 1430 1475 504 970 193 9.3 3.2 6.1 | 4.54 1.383 762

basically, last year's PECOTA exercise yielded an accurate forecast in every respect except one --- homers allowed, which it underestimated by 37. that alone explains the 50-run miscalculation on the bottom line.

at the end of that post, i plugged the PECOTA-derived runs-scored estimate and the PECOTA-derived runs-allowed estimate into the pythagorean won-loss formula to get a PECOTA-derived won-loss estimate of 90 wins, 72 losses --- a far more pessimistic, but ultimately more accurate, forecast than the pundit class was registering last spring. baseball prospectus' own PECOTA-generated projection was even more accurate: it put the cards at 87-75.

so much for 2006, and so much (for now) for PECOTA; let's look ahead to 2007, and look at ZIPS. we'll do that after the jump . . . .

projecting the pitching staff at this early date is complicated on a number of counts, beginning with the fact that the staff remains very much a work in progress. it's likely that the cardinals will acquire another starting pitcher before the season starts, either via free agency or trade. all we can do for now is go with the guys already on board. my best guess would be as follows:

rotation: carpenter, reyes, wainwright, wells, thompson
bullpen: isringhausen, springer, kinney, looper, hancock, flores, johnson

ZIPS projects wainwright and thompson as relievers in 2007, but dan szymborski (ZIPS' author) graciously provided separate projections for them as starting pitchers; did the same thing for looper, for that matter, although i'm not going to use that one. the wainwright projection is here, at comment #27; the thompson projection is here. (and looper is here, if you're curious.) many thanks to dan for the extra effort.

armed with those projections, and pro-rating all the starters for 33 starts (30 for thompson, the ostensible #5), we can construct a basic stat sheet for the 2007 staff:

GS IP H ER BB SO HR | ERA WHIP
carpenter 33 230 206 78 46 187 23 | 3.05 1.096
reyes 33 195 187 84 50 165 30 | 3.88 1.208
wainwright 33 181 191 77 58 112 24 | 3.83 1.378
wells 33 182 182 101 101 137 23 | 4.99 1.554
thompson 30 160 173 80 53 89 27 | 4.50 1.417
is'hausen 0 61 50 24 30 56 6 | 3.54 1.311
looper 0 74 75 29 23 43 5 | 3.53 1.324
springer 0 54 50 25 19 43 8 | 4.17 1.278
kinney 0 92 84 41 44 72 9 | 4.01 1.391
hancock 0 70 68 31 21 44 8 | 3.99 1.271
flores 0 47 48 23 20 39 5 | 4.40 1.447
t johnson 0 60 53 35 47 63 8 | 5.25 1.667
repl relievrs 0 35 41 20 18 25 7 | 5.14 1.686
TOTAL 162 1441 1407 648 530 1074 183 | 4.05 1.344
2006 161 1430 1475 721 504 970 193 | 4.54 1.383

tack on 0.35 unearned runs per game (the cards' average over the last three seasons), and the final tally comes to 704 runs allowed --- a 58-run improvement over 2006. now put that number into the pythagorean formula alongside the 781-run estimate for the offense, and we end up with a projected record of 89-73. that's only a starting point; we'll need to make some adjustments. but before we leave this table, a few of points of clarification:

  • the "repl relievers" line is a catch-all parcel of innings that will be thrown by rincon, cavazos, cate, falkenborg, kiesler, and/or whoever else spends a few random weeks on the big-league roster filling in for an injured pitcher. i pulled the projected stats more or less outta me bum; we can safely say that's a replacement-level line.
  • pro-rating each pitcher's games started causes some of the other values (including runs allowed and inning pitched) to end up as fractions. the rounding might lead to a few minor discrepancies on the totals line; just ignore them. for the same reason, the pitchers' individual era figures vary by a point or two from their ZIPS-projected figures; ignore that, too.
  • szymborski's supplemental "as-starter" projections for wainwright and thompson include the information that really matters --- ie, runs allowed --- but not figures for component stats like hits, walks, strikeouts, etc. those are not particularly important, so i simply made up totals to make the sheet look pretty. in wainwright's case, i based my ad hoc figures on two things: his 2005 major-league equivalent, and his 2006 ZIPS projection (which forecasts wainwright as a starting pitcher). in thompson's case, i constructed a template based largely on jeff suppan's pre-stl rate stats.
now let's start refining our premises, making them more realistic. our initial table assumes that the cardinals will only use five starting pitchers in 2007; hah. they have had some remarkably stable rotations under la russa and duncan, but 2007 isn't shaping up as one of those years. let's just say the top 5 guys miss a total of 20 starts; just a give-or-take figure. we'll hand those 20 starts to replacement-level pitchers already in the system --- narveson, kiesler, hawksworth, josh hancock, what difference does it make. we'll lump `em all into one line called "replacement 1" and assign to it replacement-level stats --- a 5.75 era and a 1.667 whip. our table now reads like so:
GS IP H ER BB SO HR | ERA WHIP
carpenter 30 209 187 71 42 170 21 | 3.06 1.096
reyes 28 166 158 71 42 140 26 | 3.85 1.208
wainwright 28 165 174 71 53 102 22 | 3.87 1.378
wells 28 154 154 85 85 116 23 | 4.97 1.554
thompson 28 155 172 77 52 86 26 | 4.47 1.417
repl starter 20 100 117 64 50 56 20 | 5.76 1.667
is'hausen 0 61 50 24 30 56 6 | 3.54 1.311
looper 0 74 75 29 23 43 5 | 3.53 1.324
springer 0 54 50 25 19 43 8 | 4.17 1.278
kinney 0 92 84 41 44 72 9 | 4.01 1.391
hancock 0 70 68 31 21 44 8 | 3.99 1.271
flores 0 47 48 23 20 39 5 | 4.40 1.447
t johnson 0 60 53 35 47 63 8 | 5.25 1.667
repl relievrs 0 35 41 20 18 25 7 | 5.14 1.686
TOTAL 162 1441 1431 667 546 1054 190 | 4.17 1.371
2006 161 1430 1475 721 504 970 193 | 4.54 1.383

the overall runs total under this set of assumptions rises to 723; the pythagorean won-loss estimate drops to 87-75. this strikes me as a more credible projection; the 20 replacement-level starts represent a reasonable counterweight to ZIPS' buoyant expectations for the young pitchers.

and suppose those expectations are too buoyant? a lot of people think they already are. so let's bloody the young pithchers' noses a bit and see where that leaves us. i'm gonna leave reyes' projection alone, for a couple of reasons. first, the odds are that at least one of these young pitchers will have a good year; it would be just as unrealistic to assume the worst about all three as it would be to assume the best-case scenario for all. reyes has proved himself capable of shutting down good-hitting teams, and ZIPS' optimistic projection for him is echoed by bill james and CHONE; the consensus is that he will have a good year. besides, he's got cool socks and a flat brim. the 3.85 era stays.

in wainwright's case, there's perhaps more justificable reason for pessimism. according to nate silver of baseball prospectus, the standard era conversion rate from relieving to starting is +25 percent; if a guy has a 2.80 era as a relief pitcher, project him to a 3.50 era as a starter. ZIPS projects adam wainwright to post a 3.45 era in relief this season; if we pad that era by 25 percent, it comes out to 4.31. it so happens that last spring PECOTA and ZIPS both projected wainwright to an era --- as a starter --- of 4.50 or thereabouts; 4.31 is right in the same ballpark, so it's got the ringa truth about it. so we'll use that figure for wainwright, rather than the sunnier 3.85 projection supplied by szymborski.

for thompson, we'll just gratuitously inflate his stats. ZIPS projects him to 3.60 as a relief pitcher; using the standard conversion rate, that becomes 4.50 as a starting pitcher, or exactly what szymborski projected (4.47). but let's agree that thompson's previous stats (the basis for the projection) have benefitted from the careful management of la russa and duncan, who shield him from adverse matchups and thus make him appear to be more effective than he really is. instead of a 25 percent inflation, let's charge thompson with twice that --- 50 percent. that pushes his expected era as a starter to 5.40.

but now let's make one final assumption: if brad thompson's era is 5.40, he's not going to stay in the rotation. they'll go out and acquire somebody in-season, be it jon lieber or kris benson or their equivalents from some other team. maybe it'll be dontrelle willis; i dunno. let's just say this mystery acquisition will be a league-average pitcher, ie with a 4.50 era or thereabouts; we'll call this guy replacement #2 and postulate that he comes aboard at around the all-star break. that would leave us with:

GS IP H ER BB SO HR | ERA WHIP
carpenter 30 209 187 71 42 170 21 | 3.06 1.096
reyes 28 166 158 71 42 140 26 | 3.85 1.208
wainwright 28 165 180 79 53 102 22 | 4.31 1.433
wells 28 154 154 85 85 116 23 | 4.97 1.554
thompson 17 90 105 54 52 86 26 | 5.40 1.544
repl start #1 17 85 99 54 43 47 17 | 5.72 1.667
repl start #2 14 80 84 40 27 46 13 | 4.50 1.389
is'hausen 0 61 50 24 30 56 6 | 3.54 1.311
looper 0 74 75 29 23 43 5 | 3.53 1.324
springer 0 54 50 25 19 43 8 | 4.17 1.278
kinney 0 92 84 41 44 72 9 | 4.01 1.391
hancock 0 70 68 31 21 44 8 | 3.99 1.271
flores 0 47 48 23 20 39 5 | 4.40 1.447
t johnson 0 60 53 35 47 63 8 | 5.25 1.667
repl relievrs 0 35 41 20 18 25 7 | 5.14 1.686
TOTAL 162 1441 1440 683 548 1051 189 | 4.26 1.379
2006 161 1430 1475 721 504 970 193 | 4.54 1.383

even under this bleak (some would say "fair" or "realistic" or "probable") set of conditions, the cardinal pitching staff would improve over last season's --- but not by much. now the runs-allowed estimate is at 739, and the projected pythagorean record is 85-77. in this telling the rotation resembles last season's: two good pitchers and a pack of bums. without question, there's a chance that things will go this way in 2007; not a very good chance, in my view, but it's clearly a possibility. if that's how it breaks, then the cardinals won't make the playoffs.

but we know it won't break exactly that way; it'll break some way that nobody, and no projection system, can anticipate. having said that, i trust the parameters we've sketched out over the last couple of days using ZIPS as our guide: the cardinals, as presently constituted, look like an 85- to 89-win team. their main areas of need appear to be the same as they've been since june 2006: in the rotation and the outfield. rotation help will surely be forthcoming, one way or the other; when and in what guise (and to what effect) remains to be seen. as for the outfield, we might as well steel ourselves for 850 at-bats from encarnacion and taguchi.

the PECOTA projections are due out next week. when they appear i'll re-run this whole exercise and report out the bottom-line estimates, but i won't bore you with all the chapter-and-verse exposition. likewise, if the cardinals make a roster move that alters their projection by a game or more in the standings, i'll report that out too.