February 2009 - Posts

• Runs Environment

  Posted Feb 23 2009, 06:56 AM by radio silence

   How much runs will be scored and allowed in the NL for 2008? This will greatly affect our estimates of how each team will fare, and greatly affect our estimates of wins and losses in the league, given that wins and losses for each team will depend upon the distribution of runs, as well as overcoming disparate runs distribution (i.e., stealing many 1 run victories, winning lots of close games with the bullpen, or having a completely awful bullpen, etc). Each of these factors represent a manner in which runs estimates are overcome; but in the first place, we need runs estimates.

  Here's the 10 Year Forecast:

  Runs Allowed

  2008: 11976 R / 2588 G
  

  2007: 12388 R / 2594 G
  

  2006: 12558 R / 2590 G
  

  2005: 11709 R / 2594 G
  

  2004: 12060 R / 2590 G
  

  2003: 11917 R / 2590 G
  

  2002: 11546 R / 2588 G
  

  2001: 12186 R / 2592 G
  

  2000: 13021 R / 2593 G
  

  1999: 12838 R / 2591 G

   

  Runs Scored

  2008: 11741 R / 2588 G
  

  2007: 12208 R / 2594 G
  

  2006: 12337 R / 2590 G
  

  2005: 11535 R / 2594 G
  

  2004: 12018 R / 2590 G
  

  2003: 11945 R / 2590 G
  

  2002: 11516 R / 2588 G
  

  2001: 12186 R / 2592 G
  

  2000: 12976 R / 2593 G
  

  1999: 12966 R / 2591 G
  

   

  After a peak in the heart of the steroids era, runs production fluctuated for several years before beginning a sharper decline from 2006 forward. Runs production levels have been lower than 2008's levels within the past decade, but at the moment, runs production levels are actually progressively lower. Given the effects of interleague play on National League pitching, it might be plausible to suggest that -- once again -- runs scored will slightly decrease, continuing the pattern, but runs allowed will be more likely to maintain a constant level, especially if the American League bounces back as an offensive league in 2009. 

  After a peak in 2000, runs allowed totals now stand at approximately 92% of that peak runs environment nine years ago. Runs scored totals stand at approximately 90% of peak runs environment nine years ago. Oddly enough, although both RA and RS declined in the past three years, decreases in RS have been slight from 2006-2007, and then sharp from 2008, whereas decreases in RA have simply been progressive:

  RA Decreases

  2006-->2007:  .986

  2007-->2008: .967

   

  RS Decreases

  2006-->2007: .990

  2007-->2008: .962

   

  Observations:

  (a) Both RS and RA were below the ten-year average in 2008.

  (b) RS seems to remain more constant for two-year periods, with declines or increases following stable two-year periods. RA fluctuates more, and exhibits more progressive trends.

  (c) The American League has traditionally been the stronger league within the last decade, which might suggest a rebound for the Junior Circuit in 2009.

   

  Expecting an American League rebound and National League runs scored to remain at its 2008 level once again, we might see Runs Allowed itself remain rather constant from its 2008 level, with the two factors taken together. The general gains and losses in starting pitching shifted the pitching balance to the American League, as well, with C.C. Sabathia's historical production in the second half of 2008 failing to receive a chance to be defended for another year; half or full year losses of Ben Sheets, Tim Hudson, and John Smoltz will also affect the league, as well as the retirement of Greg Maddux. The shift of Javier Vazquez might recover some of those runs allowed losses, as well as more NL starts by Yovani Gallardo, Rich Harden, Clayton Kershaw, Max Scherzer, and other young pitchers.

  Overall, the feeling that NL pitching will overcome both NL and AL hitting to overtake the league runs scored totals in 2009 seems to be misguided; as runs scored averages even out for another year, and AL hitting recovers, the outlook seems to favor hitting in the NL for 2009....

  NL 2009 RS:  11624 R

  NL 2009 RA: 11855 R
  

  With no comments

  Link this   E-mail This Link

• Quick Runs Created Correction

  Posted Feb 23 2009, 06:08 AM by radio silence

   There is a difference between the qualities of an aggregate and the qualities of a whole. An aggregate is pieced together by elements, the qualities of an aggregate are derived from the qualities of the elements. By contrast, a whole possesses its own qualities, which are not pieced together by elements.

  One of the issues with utilizing individual runs created is that should individual RC totals be added up for an MLB team, the RC estimate probably does not equal the actual team's runs scored. The reason for this is simple; a team is more than an aggregate of individual production from different players; a team features its own situations and interaction throughout the line up that produces runs. Runs Created can certainly create a close estimate, but it cannot perfectly recreate or capture the interaction of a line up throughout game situations.

  Which is why Runs Created can and should be corrected by team  runs scored. (a) Add up all the RC estimates for every player on a team; (b) Dvide actual team RS into the aggregate RC estimates (Team RS) / (RC aggregate) = (RS / RC differential); (c) Multiply individual RC estimates by the % difference between RS and RC (RS / RC differential)*(Individual RC estimate). For example, the 2008 Brewers' RC estimate -- added from each of the players -- was 813 RC, which is approximately 7% off of the team's actual 750 RS (Team RS / RC aggregate = .92251). Multiplying each individual RC estimate by .92251, then, we have a correction of that player's RC estimate. For example, Ryan Braun's RC estimate in 2008 was approximately 111 RC; with the correction, his runs production would be around 102 runs.

  This is rather simple, but time consuming. Which is why I suggest another quick correction, which might take you in the direction of a lower estimate all together, but that's not necessarily a bad thing.

  We have a record of R and RBI, which provide us with the outlines for runs stats. R is the truest runs-stat there is, and RBI is a contrived runs stat that is meant to distribute team runs on an individual level. Throwing spaghetti against the wall, I suggest taking the harmonic mean of R and RBI to produce a quick and easy correction of RC:

  (2*R*RBI) / (R+RBI). We'll call this "RRBI." The Brewers' 2008 RRBI was (2*750*722) / (750 + 722) = 1083000 / 1472 = 735.7 = approx. 736 RRBI. This number is lower than the Brewers' actual runs scored, but it is rather close to the Brewers' actual runs scored total, coming with 98% of the actual runs scored.

  It's not a perfect method, but if you're looking for a faster correction for individual RC estimates, it might not be a bad one. It's even better in conjunction with a full, team-context correction:

   

  Player		RC	Correction	RRBI
  Kendall		59	54.43	47.5
  Fielder		114	105.17	93.3
  Weeks		72	66.42	60.6
  Hall		47	43.36	52
  Hardy		90	83.02	75.9
  Braun		111	102.4	98.5
  Cameron		75	69.19	69.5
  Hart		81	74.72	82.8
  Starters		649	598.71	580.1

  
  As you can see, the correction using the team's RS / RC % and the correction using RRBI are rather close. While the team RS / RC % estimate will always be more accurate because it doesn't use RBI, and specifically uses runs alone, RRBI should be a quick and simple correction, otherwise.
  
  With no comments

  Link this   E-mail This Link

• What do Manny Parra and Johan Santana have in common?

  Posted Feb 22 2009, 10:56 PM by radio silence

   They're both basically average facing left-handed hitters, and at that, they posted very similar platoon splits against lefties in '08 (.247/.289/.389 for Santana, .233/.313/.367 for Parra). The difference in their performances can be attributed almost solely to their respective performances against right-handed batters.

  That's right, it's left-handed platoon time. I've been itching to look at lefty/lefty platoon splits for a while, to see what kind of lefty pitchers actually provide a platoon advantage.

  This one was simple: I listed the NL 2008 average lefty vs. lefty platoon split, and then listed all left-handed pitchers on NL 2009 rosters that pitched in the MLB in 2008 (no matter the IP. You can raise a sample size issue if you like, but I say that that's the whole point! I mean, we are doing platoon splits after all; that's basically writing 'SAMPLE SIZE' on your forehead in black marker).

  So, here we go...

  Above Average (26 P; 19 RP)

  Daniel Ray Herrera, Cin (RP).            .000/.200/.000 (.200)               7.3 IP

  Jeff Ridgway, Atl (RP).                     .133/.133/.200 (.333)               9.7 IP

  Michael O’Connor, Was (RP).            .167/.167/.167 (.333)               9.0 IP

  J.C. Romero, Phi (RP).                       .102/.193/.153 (.346)               59.0 IP

  Arthur Rhodes, Cin (RP).                   .157/.253/.200 (.453)               35.3 IP

  Tim Byrdak, Hou (RP).                      .135/.222/.247 (.469)               55.3 IP

  Sean Burnett, Pit (RP).                       .171/.238/.276 (.514)               56.7 IP

  Scott Schoeneweis, Ari (RP).             .178/.243/.277 (.520)               56.7 IP

  Oliver Perez, NYM (SP).                    .158/.250/.271 (.521)               194.0 IP

  Billy Wagner, NYM (RP).                  .220/.283/.244 (.527)               47.0 IP

  Scott Olsen, Was (SP).                       .187/.262/.276 (.538)               201.7 IP

  Mike Hinckley, Was (RP).                  .222/.263/.278 (.541)               13.7 IP

  Hong-Chih Kuo, LAD (RP).              .202/.216/.340 (.557)               80.0

  Mitch Stetter, Mil (RP).                      .158/.304/.263 (.568)               25.3 IP

  Pedro Feliciano, NYM (RP).              .210/.280/.295 (.575)               53.3 IP

  Franklin Morales, Col (SP).                .200/.182/.400 (.582)               25.3 IP

  Justin Hampson, SD (RP).                  .250/.391/.292 (.583)               30.7 IP

  Phil Dumatrait, Pit (SP).                     .206/.351/.238 (.589)               78.7 IP

  Alex Hinshaw, SF (RP).                     .205/.318/.274 (.592)               39.7 IP

  Barry Zito, SF (SP).                            .213/.316/.287 (.603)               180.0 IP

  Trever Miller, StL (RP).                      .209/.305/.308 (.612)               43.3 IP

  John Grabow, Pit (RP).                       .239/.321/.296 (.617)               76.0 IP

  Wesley Wright, Hou (RP).                  .207/.295/.326 (.621)               55.7 IP

  J.A. Happ, Phi (SP).                            .209/.261/.395 (.656)               31.7 IP

  Greg Smith, Col (SP).                         .232/.270/.393 (.663)               190.3 IP

  Scott Eyre, Phi (RP).                          .220/.264/.400 (.664)               25.7 IP

   

  Within 3% of Average (9 P; 5 RP)

  R.J. Swindle, Mil (RP).                       .333/.333/.333 (.667)

  Charlie Manning, Stl (RP).                 .203/.284/.392 (.676)

  Johan Santana, NYM (SP).                 .247/.289/.389 (.678)

  Manny Parra, Mil (SP).                       .233/.313/.367 (.680)

  Doug Slaten, Ari (RP).                       .232/.317/.375 (.692)

  Jonathan Sanchez, SF (SP).                .235/.287/.424 (.711)

  Jaime Garcia, Stl (RP).                        .250/.400/.313 (.713)

  Alan Embree, Col (RP).                      .232/.304/.415 (.719)

  Jamie Moyer, Phi (SP).                       .240/.321/.400 (.721)

   

  Below Average (35 P; 15 RP)

  Jo-Jo Reyes, Atl (SP).                         .255/.314/.415 (.729)

  Jack Taschner, SF (RP).                      .279/.339/.394 (.733)

  Royce Ring, StL (RP).                        .264/.339/.396 (.735)

  Bill Bray, Cin (RP).                            .260/.360/.384 (.744)

  Andrew Miller, Fla (SP).                    .226/.378/.366 (.744)

  Jeremy Affeldt, SF (RP).                    .269/.301/.444 (.745)

  Pat Misch, SF (RP).                            .281/.309/.438 (.746)

  Zach Duke, Pit (SP).                           .279/.344/.405 (.750)

  Tom Gorzelanny, Pit (SP).                  .261/.373/.391 (.765)

  Jeff Francis, Col (SP).                         .248/.312/.460 (.772)

  John Lannan, Was (SP).                      .259/.317/.460 (.777)

  Cole Hamels, Phi (SP).                       .262/.308/.471 (.779)

  Jorge de la Rosa, Col (SP).                 .289/.353/.455 (.807)

  Wandy Rodriguez, Hou (SP).             .282/.311/.500 (.811)

  Clayton Kershaw, LAD (SP).             .250/.337/.475 (.812)

  Dan Meyer, Fla (RP).                          .314/.385/.429 (.813)

  Randy Wolf, LAD (SP).                     .283/.368.447 (.815)

  Sean Marshall, ChC (RP).                   .269/.354/.463 (.817)

  Tom Glavine, Atl (SP).                       .290/.355/.464 (.819)

  Doug Davis, Ari (SP).                         .321/.384/.438 (.822)

  Boone Logan, Atl (RP).                      ,291/.324/.505 (.829)

  Eric Stults, LAD (SP).                        .314/.314/.543 (.857)

  Neal Cotts, ChC (RP).                                    .269/.338/.522 (.860)

  Renyel Pinto, Fla (RP).                       .264/.371/.529 (.900)

  Ted Lilly, ChC (SP).                           .307/.386/.542 (.928)

  Randy Johnson, SF (SP).                    .303/.382/.545 (.927)

  Mike Hampton, Hou (SP).                  .339/.369/.559 (.929)

  Taylor Tankersley, Fla (RP).               .360/.467/.480 (.947)

  Scott Elbert, LAD (RP).                     .417/.500/.500 (1.000)

  Jonathon Niese, NYM (SP).               .353/.353/.647 (1.000)

  Matt Chico, Was (SP).                        .351/.377/.632 (1.009)

  Wade LeBlanc, SD (SP).                    .318/.423/.636 (1.059)

  Mike Gonzalez, Atl (RP).                   .259/.250/.815 (1.065)

  Joe Thatcher, SD (RP).                       .414/.424/.862 (1.286)

  Eric O’Flaherty, Atl (RP).                  .500/.579/.875 (1.454)

  So, the best platoon advantage for left-handed relievers in 2008 -- among prospective NL 2009 relievers -- belong to J.C. Romero, Arthur Rhodes, Tim Byrdak, Sean Burnett, and Scott Schoeneweis. Notably below average lefty v. lefty platoon splits belong to Jeremy Affeldt, Mike Gonzalez, Renyel Pinto, Tayler Tankersley, Neal Cotts, Sean Marshall, and Bill Bray. Managers facing those relievers might think twice about leaving their lefty-bat in the game.

  What I found especially interesting is that starting left-handed pitchers fared worse against lefty bats than relievers. Off the top of my head, I'd guess that that occurs simply due to how relievers are used, and also how those starting pitchers approach the game. Some of the more prominent pitchers on the "below average" list throw a notable change up (see Lilly and Hamels), and others rely on either a slider or cutter (ever seen Doug Davis and Randy Johnson in a sentence together other than "sharing the 2008 Diamondbacks' rotation..."?). Both cutter/slider and change up pitchers utilize a major pitch meant for neutralization, which might explain why those pitchers pitch relatively poorly against lefties.

  So, be careful with your left-handed relievers, and be careful pinch hitting against them. Platoon advantages can indeed work, if you manipulate them properly, but there are situations in which managers might prefer to leave left-handed bats in against left-handed pitchers.

  A pitch-type by pitch-type batting order might help a team to produce more runs, rather than a strict platoon based on handedness (i.e., perhaps batting orders should be constructed to face either curveballers, sinkerballers, change-up artists, or slider pitchers, etc.)
  

  With no comments

  Link this   E-mail This Link

• Batting Luck

  Posted Feb 20 2009, 06:43 AM by radio silence

   After working on a way to measure the percentage of balls put in play by a particular player, I think that we can figure out ways to measure luck beyond BABIP, or Batting Average on Balls in Play. Without context, BABIP is a difficult stat to use because it says nothing about the frequency of balls in play, which frames the value of having a batting average for that type of batting event.

  Using BABIP and BIP% together, we can measure (a) how frequently the batter put balls in play, and (b) how frequently those balls in play dropped for hits. Together, between various years, we can eventually explain increases and decreases in OBP, AVG, and maybe even SLG, depending upon how BIP% and BABIP work together. We will also be able to analyze a player's context and a player's luck for his AVG, OBP, and SLG. If a player is relatively lucky, we might expect his production level to remain reasonably consistant throughout the years. However, if a player is extremely lucky, we might not expect the player to repeat past production in the coming season.

  Now, an extremely simple survey.

  2007 NL BABIP: .301
  

  2008 NL BABIP: .298

   

  2007 NL BIP%: .7057
  

  2008 NL BIP%: .6967

   

  Now...

  2007 to 2008 Brewers Batting Luck:

  Ryan Braun 2007: .6300 BIP%, .361 BABIP

  Ryan Braun 2008: .6772 BIP%, .305 BABIP

  Braun's exceptional rookie campaign can be explained by the fact that Braun did not put a high percentage of balls in play, and the balls that were put in play dropped for hits at an extremely high rate. His 2008 regression can be easily explained by the fact that his batting luck returned to average, basically.

  2009 will help to unluck whether or not Braun was indeed exceptional in 2007, and is capable of repeating that performance -- or, was he exceptionally lucky?

  Mike Cameron 2007: .6068 BIP%, .298 BABIP
  

  Mike Cameron 2008: .5531 BIP%, .296 BABIP
  

  Cameron's career % might help answer questions about how he might play in a full season: .303 BABIP, .6014 BIP%. Cameron is a consistently low-in-play batter, and his game consistently revolves around K, BB, and HR.

  Prince Fielder 2007: .5962 BIP%, .283 BABIP

  Prince Fielder 2008: .6196 BIP%, .298 BABIP

  Fielder, despite his 2008 regression, might be the most consistent young Brewer from 2007 to 2008 in terms of batting luck.

  Bill Hall 2007: .6322 BIP%, .319 BABIP

  Bill Hall 2007: .6004 BIP%, .284 BABIP

  Yikes! I'm not sure what to say -- I suppose if you lowered your number of balls in play by 3% and subsequently lowered the batting average on those balls in play by 3.5%, you'd regress, too!
  

  J.J. Hardy 2007: .7806 BIP%,.279 BABIP
  

  J.J. Hardy 2008: .7218 BIP%, .305 BABIP

  Due to Hardy's increase in batting luck in 2008, I believe he is a strong candidate for a regression in 2009, given the fact that his BABIP increased while he placed a lower percentage of balls in play. Hardy's 2007 BABIP was much closer to his career norm thus far.

  Corey Hart 2007: .7802 BIP%, .321 BABIP

  Corey Hart 2008: .7529 BIP%, .293 BABIP

  Although we might have expected Hart's 2007 luck to correct itself, perhaps his 2008 was a bit of an uncharacteristic regression. We might expect Hart to rebound, even slightly, in 2009 due to this fact.

   Jason Kendall 2007: .8346 BIP%, .259 BABIP

  Jason Kendall 2008: .8126 BIP%, .265 BABIP

  Sweet, sweet consistency. Sort of. I think? Expect a regression out of Kendall once his BABIP corrects itself. Although, in case you're wondering, his career BABIP is .310.

  Rickie Weeks 2007: .5573 BIP%, .287 BABIP

  Rickie Weeks 2008: .6268 BIP%, .277 BABIP

  Now that's a regression in luck. The kid puts 7% more balls in play in 2008, but his AVG on those BIP drops by 1%. Weeks' career BABIP -- not to mention his 2007 BABIP -- is steadily close to average defensive efficiency, so a correction in luck is probably in order for the 2B.

   ***

  Really, we shouldn't be surprised the team regressed by 50 runs in 2008; look at Braun, look at Hall, look at Hart, and look at Weeks -- some corrections in luck, and some simply unlucky results on balls in play. Looking at four young players -- none over the age of 29 for 2009's season -- there must be something that helps to explain the regression from 2007 to 2008, and also allows us to analyze the likelihood of returning to a higher offensive output in 2009....

  I am certain that aspects of hitting approach, plate discipline, patience, etc., plague Braun, Hall, and Hart (Weeks is actually one of the best disciplined and most patient of the Brewers' young core, so I wouldn't include him in the discipline and patience list); but beyond that, I believe there are certain corrections in luck that are in order for Hart, and maybe Hall for 2009. I'm not sure what to make of Braun, given that even his 2008 BABIP was above average (but much closer to average).

  Is yet another batting luck correction in store for Braun? How will that affect the 2009 offense? 

  The 2009 Brewers will probably kick back into gear with some of the young players recovering, but I'd still note concern for Braun and Hardy's recent hitting records, in terms of luck.

  How far can Braun and Hardy regress in the luck department, and still remain competitive threats in the Brewers' middle order? 

  I don't know about you, but the 2009 season cannot come fast enough.
  

   

  With 1 comment(s)

  Link this   E-mail This Link

• Staying on base

  Posted Feb 19 2009, 07:16 AM by radio silence

   Recently I've become enamored with Runs Created, a formula that weights and combines individual stats in a manner that attempts to reflect (a) a player's ability to get on base, (b) a player's ability to stay on base and keep others on base, and (c) a player's ability to advance runners.

  While investigating and working with the calculation, a thought occurred to me: the very first portion of the calculation reflects a player's ability to get on base and stay there; in the context of the equation, this is then multiplied by the player's ability to advance runners, and that number is divided by opportunities (PA).

  The technical stat goes something like ((H+HBP+BB-CS-GDP)*(TB*((.26(BB-IBB+HBP))+(.52(SH+SF+SB))))) / (PA).

  It's the underlined part that I'm currently interested in: should we isolate that sequence, and divide it itself by PA, we end up with a number that reflects the player's OBP along with that player's other out-making tendencies. Consider it counting the second out of the GDP (the first out is already covered in PA and basic OBP -- i.e., whether or not the player makes a basic out in the first place) and the basepath out, CS. We could conceivably add in other baserunning outs, but let's keep this simple for the moment.

  Per 600 PA, this front equation lets us know that 6 GDP + 6 CS (or 10 GDP + 2 CS, etc.) lowers a player's OBP by 2% -- not necessarily a small number whatsoever. For example, if a player sports a .340 OBP in 600 PA, those 204 times on base drop for each GDP and CS -- an example of 12 GDP + CS effectively lowers that player's times on base to 192, which in turn lowers the OBP from .340 to .320.

  The point is, even though we see and debate explicit OBP numbers from players, there are other elements that are external to basic OBP that nevertheless affect the player's on-base capabilities. CS and GDP are two such factors that basically serve to lower the team's ability to produce runs because they are outs that take a baserunner off of the basepaths.

  You might protest that a player does not deserve to have baserunners lost in GDP counted against his record, but I'd protest that that player should simply quit hitting into so many double plays.

  Here's an example of the stat at work: the 2008 Brewers and their OBP capabilities, including CS and GDP (w/ basic OBP listed in parentheses):

  Adjusted OBP (I like the term "Secret OBP")

  Prince Fielder (.372 OBP): .352 adjusted
  

  Rickie Weeks (.342 OBP): .323 adjusted

  Mike Cameron (.331 OBP): .313 adjusted

  J.J. Hardy (.343 OBP): .310 adjusted

  Jason Kendall (.327 OBP): .310 adjusted
  

  Ryan Braun (.335 OBP): .309 adjusted

  Bill Hall (.293 OBP): .270 adjusted
  

  Corey Hart (.300 OBP): .262 adjusted
  

   

  2008 NL Average (.331 OBP*): .302 adjusted

  *basic calculation, not park-adjusted. When evaluating individuals, you should use park-adjusted factors, available on Baseball Reference player pages.

   

  The average NL batter, in 2008, trimmed nearly 3% of his trips on base due to other batting outs -- the second out of a GIDP, a caught stealing. According to that average, Fielder, Weeks, Cameron, Kendall, Braun, and Hall were rather stingy with their extra outs; this is accounted for by Cam and Weeks' smart baserunning, and Kendall and Hall's lack of double plays. I was rather surprised that Hall only lost 2.3% of his outs, given his sloppy baserunning. I thought for sure that would affect him more.

   Hardy and Hart were absolutely hurt by rather high double-play totals. That really knocks some effectiveness out of Hardy's OBP, and lowers whatever OBP Hart did have to a terribly low level.
  

  Here's the basic point: not making outs is a huge deal. People sometimes argue that stating "not making outs" as a goal is too abstract to be a hitting strategy; but using this exercise, we can see that the value of not making outs is not abstract whatsoever. Not making outs is manifest in not hitting into double plays, running on the base paths in a manner that is effective and smart, and overall, adopting a plate approach that is a balance between patience and discipline, waiting for your pitch, seeing your pitch, swinging at your pitch, and hitting your pitch.

  Not making outs is a pretty fantastic hitting strategy, and those hitters who take a considered approach at the plate and approach the basepaths with care and attention, and simply don't hit a lot of ground balls will be more effective in the not-making-outs department. 
  

  With no comments

  Link this   E-mail This Link