Navigation: Jump to content areas:


Pro Quality. Fan Perspective.
Login-facebook
Around SBN: Jeremy Lin's Game-Winner Was Incredible, Worth Remembering

The 2010 Astros according to the clack/DQ projection system (and the return of the community projection project)

With the release of every other major projection system's standings finally completed, it's high time that we finally unveiled our joint attempt at projecting the 2010 Astros.  HLP revealed his first go around with projecting the 2010 Astros recently, but we took it upon ourselves to haggle with each other's corrections/additions/allocations etc., in order to form a more perfect projection.  We also corrected what we perceived as a flaw in the tool we were using, as it didn't include hitting for pitchers.  The result is a decrease in total projected wins.  As you'll come to find, our drop in projected wins doesn't necessarily mean the previous win total of 84 is impossible. It's just less likely than it was. Conversely, we also don't mean to imply that 79 wins (our projected total) is what we believe the Astros will finish with.

Star-divide

Below, you'll find our projections for the hitters.  Our input to the graph included the player, PA, wOBA, BR (base running), and Fld (fielding).  Both BR and Fld are on a scale of wins.  Thus, because ten runs constitute a win, you can read .25 as 2.5 runs.  BR, Fld, and the first WAR columns are the scale of 700PA, or a full season of player.  The last column, the second WAR column, is the player's scaled WAR.

Hitter Pos PA wOBA Hit BR Pos Fld Rep WAR FA $ WAR
Humberto Quintero CA 175 .290 -2.74 -.20 1.25 .25 2.00 .56 $1.1 .1
J.R. Towles CA 334 .315 -1.22 .00 1.25 .00 2.00 2.03 $5.1 1.0
Jason Castro CA 150 .299 -2.19 .00 1.25 .25 2.00 1.31 $1.7 .3
Lance Berkman 1B 579 .390 3.35 -.10 -1.25 .90 2.00 4.90 $19.8 4.1
Geoff Blum 1B 150 .294 -2.50 -.10 -1.25 .00 2.00 -1.85 -$1.5 -.4
Chris Johnson 1B   .295   .00   .00        
Kaz Matsui 2B 400 .315 -1.22 .50 .25 .00 2.00 1.53 $4.6 .9
Jeff Keppinger 2B 160 .329 -.37 -.10 .25 .00 2.00 1.78 $2.4 .4
Edwin Maysonet 2B 35 .300 -2.10 .00 .25 .50 2.00 .65 $0.6 .0
Tommy Manzella SS 475 .290 -2.74 .00 .75 .90 2.00 .91 $3.4 .6
Jeff Keppinger SS 70 .329 -.37 -.10 .75 -.50 2.00 1.78 $1.3 .2
Geoff Blum SS 60 .294 -2.50 -.10 .75 -.35 2.00 -.20 $0.3 .0
Pedro Feliz 3B 500 .308 -1.64 -.10 .25 1.50 2.00 2.01 $7.3 1.4
Jeff Keppinger 3B 60 .329 -.37 -.10 .25 -.15 2.00 1.63 $1.1 .1
Chris Johnson 3B 20 .295 -2.43 .00 .25 -.20 2.00 -.38 $0.3 .0
Carlos Lee LF 581 .365 1.83 -.50 -.75 -.70 2.00 1.88 $7.9 1.6
Cory Sullivan LF 70 .300 -2.13 .10 -.75 .05 2.00 -.73 $0.0 -.1
Jason Michaels LF 10 .308 -1.64 .10 -.75 .10 2.00 -.19 $0.4 .0
Michael Bourn CF 465 .327 -.49 .50 .25 2.00 2.00 4.26 $14.0 2.8
Jason Michaels CF 70 .308 -1.64 .00 .25 -.20 2.00 .41 $0.6 .0
Corey Sullivan CF 70 .306 -1.77   .25   2.00 .48 $0.6 .0
Hunter Pence RF 592 .358 1.40 -.25 -.75 1.50 2.00 3.90 $16.2 3.3
Jason Michaels RF 65 .320 -.91 .10 -.75 .10 2.00 .54 $0.6 .0
Cory Sullivan RF 10 .300 -2.13 .10 -.75 .10 2.00 -.68 $0.4 .0
Carlos Lee DH 60 .365 1.83   -2.00   2.00 1.83 $1.2 .2
Jason Michaels DH 30 .308 -1.64   -2.00   2.00 -1.64 $0.1 -.1
Towles/Quintero DH 20 .310 -.91   -2.00   2.00 -.91 $0.3 .0
Jeff Keppinger DH 15 .329 -.37   -2.00   2.00 -.37 $0.4 .0
Pitchers DH 330 .140  (11.87)   .00   2.00 -9.9 -$22 -4.7
Total   5556 .319 -.96 -.03   .56   1.49 $68.1 11.9

A total projection of 11.9 WAR from the bats is not bad-for this crop of players.  Compared to the Astros 2009 batting WAR of 12.5, the drop of in hitting doesn't appear to be as extreme as we've considered it (although this is a measure of both hitting an fielding, and the fielding is projected to be better than 2009). The caveat we should apply here is that the Pitchers line may be generous.  The line was arrived at by looking at our pitchers wOBA's from recent years and roughly averaging them with approximations of Felipe Paulino and Bud Norris' wOBA included.  There is probably room for a drop there, too, but nothing earth-shattering.

The two inputs that were most debated were Pedro Feliz and Tommy Manzella's defense. Feliz was a +18 to +25 defender in his recent years in San Francisco, but his defensive production declined abruptly to +5 to +7, when he moved to Philadelphia.  Is this an age-related drop or does this reflect a change related to joining a new team?  Possible explanations include the change in ballpark, team decisions with respect to positioning and handling bunts, and the difference between playing beside Jimmy Rollins instead of Omar Vizquel.  The defensive plunge, which coincided with the move to Philly, seemed too abrupt to be solely age-related decline, and we assumed that Feliz's defensive performance will fall somewhere between his defensive production with the Giants and Phillies, pegging him at +15.     

Manzella is more enigmatic, so how do we account for that? We could assume a replacement level year, or we could be somewhere on his assumed scale of ability for -5 to +5.  Given the above points about the interactions between the left side of the infield, the fact that Kaz Matsui can turn a double play, and Lance Berkman's ability to scoop, it doesn't seem unreasonable to assume a plus year for Manzella with his glove.

Onward to the projections for the men of the mound.  Below is the table for the Astros starting rotation and bullpen.  In all honesty, this is tough.  DQ stared blankly at the pitcher's table for an hour and left more uncertain than he started.  These projections relied on what the projections systems have spat out, coupled with what we know about each pitcher through their peripheral stats.  Allocating IP for this staff isn't easy as the injury bug could strike a majority of our starters, which is why Brian Moehler gets so much love. As far as relief pitchers go, there are number of inputs we have that someone could look at and scratch their head.  The inputs we had control over were IP, ERA, and LEV (leverage).        

Pitcher S/R IP ERA LEV FA $ WAR
Roy Oswalt S 189 3.95 1.0 $15.5 3.1
Wandy Rodriguez S 180 3.89 1.0 $15.5 3.1
Bud Norris S 145 4.40 1.0 $8.2 1.6
Brian Moehler S 120 5.00 1.0 $3.1 0.6
Felipe Paulino S 145 4.60 1.0 $6.6 1.3
Brett Myers S 130 4.60 1.0 $5.9 1.1
Matt Lindstrom R 60 3.67 1.8 $5.7 1.1
Brandon Lyon R 64 3.66 1.3 $4.6 0.9
Alberto Arias R 65 3.60 1.0 $3.9 0.7
Jeff Fulchino R 80 3.84 1.0 $3.5 0.6
Samuel Gervacio R 65 3.73 0.7 $2.5 0.4
Chris Sampson R 70 3.77 1.1 $3.7 0.7
Tim Byrdak R 59 4.42 1.3 $0.7 0.1
Wesley Wright R 40 4.85 0.5 $0.0 -0.1
Brian Moehler R 35 4.85 0.5 $0.1 -0.1
Starters   909 4.35   $52.7 10.8
Relievers   538 3.95 1.06 $21.5 4.4
Total   1447 4.20   $73.8 15.2

With batters, it's a little easier to project a WAR because the only things we're actually projecting are directly involved in their ability to create outs.  For pitchers, and especially relievers, the process is obscured somewhat by assigning an unpredictable, unreliable stat that doesn't necessarily portend to the creation of outs and then attempting to scale that stat to the situation in which they are creating outs.  Describing that feels like reading the Allegory of the Cave. We think everything we have here helps everything come out in the wash.  However, you are free to disagree vehemently if wish you and free us from our shackles and turn our vantage point to reality.

As we've stated, this is just our attempt at providing our reasonable approximation of the performance level of each individual player and then determining a projection for the team based on the sum of its parts.  It is by no means definitive.  When we say that we've projected the Astros to win 79 games in 2010, we're really saying that we believe the true talent level of the Astros, as a team, is 79 wins.  As such, what this next table tells you is the relative probability of a 79-win true-talent team winning various numbers of games over the course of a 162 game season.

Wins Prob
55 99.99932391
60 99.98077003
65 99.70726218
70 97.53406615
75 88.07713448
76 84.6533771
77 80.63582323
78 76.03604352
79 70.89749776
80 65.29632027
81 59.33888723
82 53.1561422
83 46.89513457
84 40.70866275
85 34.74422216
90 11.91280116
95 2.444342897
100 0.285028545

As you can tell, there is no guarantee that a true-talent 79-win team would win 79 games; there's really only a 71% chance.  Having all of our inputs tells us this fact is useful, but not the most useful aspect of this entire exercise.  What we have is a very well informed foundation from which sensitivity analysis can be performed.  Some of you probably read this and thought, "for goodness' sakes, Bud Norris is 3.97 SIERA pitcher, he's going to post better than a 4.40 ERA." Or conversely, "Bud Norris is a ticking time bomb; his arm falls off after 70 innings."  The examples could go on and on.  The point is that you can figure out just what each of those scenarios might mean for the 2010 Astros.

DQ has uploaded our final version of the spreadsheet to Google Docs in a folder entitled Astros WAR 2010.  What we'd like to do is have you play around with the numbers.  Correct where you think we went wrong, both in playing time in production, and then upload the results.  We'll give it a week or two and pull the various true talents that are produced, average them, and see what we end up with.  Heck, we could even do the same for every player's input, and go the same route.  The ultimate in community projections, eh?

If you fill a spreadsheet out, and already have a Google account, upload it and place the link in the comment section.  If you don't, state as much in the comments with the relevant WAR for hitters, pitchers, and the total WAR, along with your email address formatted as such: whatever at whatever dot com.  DQ will send you an email that you can reply to and send him your spreadsheet (you can also be proactive an look at his SBN profile and send it to him that way).

Comment 4 comments  |  0 recs  | 

Do you like this story?

Comments

Display:

One of the things I mentioned to DQ when we looked at this spreadsheet is that the range of reasonable results around a prediction like this are quite large. It’s really impossible to say that 79 wins is a better prediction than, say, 82 wins or 76 wins. So that’s why I think one of the useful aspects of these kinds of spreadsheets is to perform sensitivity analysis; you can use it to provide information on questions like, “What happens if Felipe Paulino has break out year?” or “what happens if Carlos Lee gets injured and 100 of his plate appearances are replaced by Michaels?”

I saw a comment by one of the BaseballThinkFactory regular posters, Walt Davis, which struck me as fairly insightful on the uncertainty issue. I’ll quote a passage from that comment:

An underlying problem is that effect sizes in baseball are small. 20 points of OBP is an important difference but that’s just reaching base 2% more/less often. Think about your standard political poll which is usually 1,000 people which produces a standard error of about 3%. A “poll” of 600 PA would produce a standard error of about 4% — we’re trying to detect effects the equivalent of 1/2 of a seasonal standard deviation. (which is why you need about three years of data to begin to be able to say anything.) If you want to compare two players, your standard error is 1.4 times larger, assuming they are independent.

And given we’re in the projection time of year, what annoys me is that so little attention is paid to the uncertainty of these estimates — despite the best efforts of ZiPS and PECOTA (and maybe others) to present that variance in their output. Take the recent Carlos Lee projection. He’s got an 8% chance of being “poor”, putting the standard error on an OPS+ projection around 15-20 points. Hunter Pence has a 19% chance of being poor which means that his projection probably isn’t significantly different than replacement level.

We can really say very little with confidence. Pujols is definitely a significantly better hitter than Hunter Pence; I’m not sure that we can say the same for Berkman vs. Pence and maybe not even for Pujols vs. Berkman.

Regardless whether you use a model like ZIPS or CHONE to predict players or adjust those for subjective opinion, in putting together a W/L forecast the error bands are significant. So, it’s fun to do, but I wouldn’t bet money on it.

by clack on Feb 16, 2010 1:07 PM CST reply actions  

I feel like Ben Stein in Ferris Beuller

But:

Anyone?

The Crawfishboxes
A good friend of mine used to say, "This is a very simple game. You throw the ball, you catch the ball, you hit the ball. Sometimes you win, sometimes you lose, sometimes it rains." Think about that for a while.

by Stephen Higdon on Feb 17, 2010 8:26 AM CST reply actions  

I love projections but...............

………the game is decided on the field and not in some computer data base. Computers and rankings cannot account for injuries and luck. A team that wins it all must a) stay healthy and b) be at least a little lucky. I do not in any way dispute what the computer has said about the Astros but there are a lot of intangibles not taken into consideration in that ranking.

by AstrosBill on Feb 17, 2010 11:13 AM CST reply actions  

All valid points

But the point of this exercise is not to say they will/should win X number of games. The goal is to get an approximation of this group of player’s true talent level based on what we would predict the ability is. From there, we can get various probabilities of what kind of luck would be needed to win X number of games, as well as test what the impact of health/under or over performance has on those probabilities.

The Crawfishboxes
A good friend of mine used to say, "This is a very simple game. You throw the ball, you catch the ball, you hit the ball. Sometimes you win, sometimes you lose, sometimes it rains." Think about that for a while.

by Stephen Higdon on Feb 17, 2010 12:27 PM CST via mobile up reply actions  

Comments For This Post Are Closed


User Tools

Welcome to the Crawfishboxes, the SBNation blog for the Houston Astros.

FanPosts

TCB Tweets!

    

Current Series

3 game series vs Rockies @ Minute Maid Park

Sat 04/07 6:05 PM CDT
Sun 04/08 1:05 PM CDT

NL Central Standings

W L PCT GB STRK
Chicago 0 0 .000 0 Lost 0
Cincinnati 0 0 .000 0 Lost 0
Houston 0 0 .000 0 Lost 0
Milwaukee 0 0 .000 0 Lost 0
Pittsburgh 0 0 .000 0 Lost 0
St. Louis 0 0 .000 0 Lost 0

(updated 2.15.2012 at 6:05 PM CST)


Managing Editors

Tcb_icon_small Timothy De Block

Old_school_dome_logo_small David Coleman

Editors

Nsapcs13_large_small clack

H_astros_small Subber10

Astrobritrs2_small AstroB

Small conroestro

Small native_astro

Small CRPerry13

Authors

Astros_logo121009_small OremLK

This-is-my-boom-stick_small allphilla

Th_houston-astros-logo-3_small Will Bonn