FanGraphs Baseball

Comments

RSS feed for comments on this post.

  1. Linear weights will never render simple back-of-the-envelope stats like OPS obsolete, because they’re much harder to calculate offhand. That said, I never liked the counterintuitive combination of OBP and SLG and have always preferred to list them separately.

    Comment by Joel — June 27, 2011 @ 3:21 pm

  2. wOBA will never obsolete OPS because people don’t understand what it means. It’s taken nearly a decade for something as simple as OPS to catch on.

    Comment by Steve — June 27, 2011 @ 3:27 pm

  3. Really? Either way, you need a pretty big envelope.

    [(H + BB +HBP) / (AB + BB + HBP + SF)] + [(1B+(2B*2)+(3B*3)+(HR*4)) / AB]

    vs

    (.70*BB + .73*HBP + .89*1B +1.27*2B +1.61*3B + 2.07*HR +.25*SB – .50* CS) / PA

    Comment by Yirmiyahu — June 27, 2011 @ 3:29 pm

  4. Right, because we all keep decimals memorized the same way we keep categories!

    Comment by Joel — June 27, 2011 @ 3:31 pm

  5. Any saberist worth his salt has a pocketful of napkins with a chart of wOBA coefficients for every year printed on it.

    You do the calculation on the back of said napkin.

    To impress your date.

    Comment by Yirmiyahu — June 27, 2011 @ 3:37 pm

  6. In general people are stupid (intellectually lazy at best), and stubborn, and scared of new things so I have to agree with Steve.

    OPS isn’t going anywhere and neither are it’s components or batting average. For player valuation it might get dropped but I don’t see your run of the mill fan giving it up. Or any broadcasters any time soon.

    Comment by David — June 27, 2011 @ 3:48 pm

  7. NFL fans don’t have the weakest hint of a clue what QB rating means, but it still gets used quite a bit (even though it’s a lousy stat).

    Comment by The Ancient Mariner — June 27, 2011 @ 3:50 pm

  8. ((OBP x 3) + SLG)/3

    OBP should be x2 right?

    Comment by mlstarr — June 27, 2011 @ 3:54 pm

  9. Good catch. I meant 1.7 (fixed), but 2 is pretty close.

    Comment by Matt Klaassen — June 27, 2011 @ 3:59 pm

  10. Does this mean we can start capitalizing the “w” now?

    Comment by Texas_Dawg — June 27, 2011 @ 4:05 pm

  11. Where do the coefficients on a wOBA come from? How do we know that, as Yirmiyahu says, that a home run has a little over twice the value of a single?

    Comment by corey — June 27, 2011 @ 4:12 pm

  12. Nolan Ryan answered a question of mine in a recent Twitter interview that the “sabermetrics” he uses most in evaluating pitchers are K/BB and H/IP.

    An MLB club president who supposedly has a lot of input on pitcher evaluations.

    I’m guessing wOBA is probably off his radar… which doesn’t give me much hope for the less-invested masses.

    Comment by Texas_Dawg — June 27, 2011 @ 4:14 pm

  13. I love wOBA. I hate whoa-bah.

    Comment by Telo — June 27, 2011 @ 4:16 pm

  14. That is correct. The ONLY thing you need for any stat to gain major acceptance is for the league to declare a stat OFFICIAL, and prepare leaderboards for it.

    Of course, you can’t push crap (like GWRBI).

    OPS has been around since at least The Hidden Game of Baseball, so that’s 30 years.

    As for back of envelope, I just memorize these numbers:
    0.7 BB, HB
    0.9 1B, Reach on error
    1.3 2B, 3B
    2.0 HR

    Not too hard to remember, is it? It’s all centered on “1″. The BB is as much away from 1 as the 2B is. HR is “2″. 1B is a bit less than 1.

    Comment by Tangotiger — June 27, 2011 @ 4:21 pm

  15. http://www.insidethebook.com/woba.shtml

    Comment by Tangotiger — June 27, 2011 @ 4:24 pm

  16. “With the increasing popularity of wOBA and other linear-weights-based offensive measures, OPS and its derivatives have become obsolete. That is as it should be.”

    I think the baseball circle you’re in (FanGraphs & the like) and your “enlightened” comrades has you overrating the increased popularity of wOBA by a large degree. wOBA certainly hasn’t gained the kind of traction to render OPS & the triple (three?) slash obsolete. It’s still very much used within a particular subset of the baseball community.

    BTW, it was a bit of an odd setup to start with the line about the triple slash components being obsolete, but then outlining ways in which it is still useful today. It set kind of an unnecessarily negative tone for the piece.

    There’s never going to be one super-stat, they all have their uses. It doesn’t really have to be an either-or thing. Same with stats v. scouting.

    Comment by pjs24 — June 27, 2011 @ 4:27 pm

  17. Sure, but OBP and SLG are pretty common, and if you have those all you need to calculate OPS is the rectangle where the stamp goes.

    Comment by aladou — June 27, 2011 @ 4:35 pm

  18. “There’s never going to be one super-stat”

    No one Stat to rule them all?

    Comment by Yirmiyahu — June 27, 2011 @ 4:36 pm

  19. Nope, no King stat… ‘cept maybe Schilling’s new one. :)

    Comment by pjs24 — June 27, 2011 @ 4:39 pm

  20. Thanks for the comment.

    I may be overrating it. I do think that “hardcore” fans on the ‘net are using it increasingly: a few years ago, OPS was the main thing, now the places I frequent are using wOBA. But you’re right, my limited circulation may exaggerate how popular I think it is. I do think I’m right to say that it is _more_ popular.

    As for OPS and the three-slash: I see them as two different things. Since OPS itself doesn’t break down into _components_, it’s only use is straight value comparison, and wOBA clearly beats it there.

    I see the three slash is different from OPS (which results from adding two of the three components), it sticks around because it is “descriptive” in a way that wOBA by itself is not.

    I would agree that I spent way too much time on OPS. I take it as a given that OPS is obsolete — the main point was supposed to be the valuation/description distinction as a reason to retain the three-slash.

    Comment by Matt Klaassen — June 27, 2011 @ 4:48 pm

  21. I’ve seen WAR get used in various MLB broadcasts … but I’ve yet to hear a broadcaster explain it well.

    Comment by AdamM — June 27, 2011 @ 4:49 pm

  22. Can we get normal baseball fans, announcers and sports radio people to use and understand OPS first (which is pretty freakin simple)?

    wOBA is of course better, but geez, these guys still use batting average.

    Comment by bill — June 27, 2011 @ 4:58 pm

  23. I gotta admit, I read a rather convincing argument –possibly in the comments section of a BP article — proposing that the value distribution between OBP and SLG be altered to accurately reflect *current* value in a lower-scoring run enviornment.

    The premise being- in a run-starved enviornment, the ability to instantly produce runs via SLG in actually more valuable in the current state of MLB then the historical balance of “don’t make outs” presupposes based on more productive run-scoring balance of yr’s past.

    Basically, if MLB is trending more towards 2 or 3 run output games…shouldn’t the ability to “out-slug” an opponent w/ instant Offense gain more value as opposed to means of one-base (singles, walks, etc) advancement?

    Not calling for a complete change of guard, simply saying that in a radically reduced enviornment of run production, both the variables and weighted vaules of performance are subject to review…if not minor revision.

    Comment by Romodonkulous — June 27, 2011 @ 5:09 pm

  24. Oh yes, Matt, I’m 100% w/you on *more* popular and especially when it comes to hardcore fans. I may have mistakenly read you to mean in the initial line that it’s more popular across the *entire* baseball landscape. We’re on the same page now. Ty for responding & clarifying.

    Comment by pjs24 — June 27, 2011 @ 5:20 pm

  25. OPS might be obsolete, but the other two clearly are not and I doubt they ever will be. I like to think of wOBA (and WAR for that matter) as “integrated” stats. They are designed to compare the overall value of a player. They are useful for that purpose, but will ultimately be subject to constant reevaluation (in WAR’s case, different systems of measuring defense, in wOBA’s perhaps in finding different new ways to compare the value of offensive categories).

    OPB, slugging percentage, batting average, etc. are “reductive” statistics. They tell you exactly how good a player is doing at one thing. No one category gives you that much information, and alone they cannot tell you how to evaluate a player, but together, they tell you more than one integrated statistic can.

    To believe otherwise is to suggest that there is no reason for someone to evaluate how good a player is at walking, or how good they are at hitting home runs, or how good they are at playing defense. If reductive stats are obsolete and we measure value only through things like wOBA or WAR, you are suggesting there is not any point to knowing which individual skillsets a player is good at.

    Comment by Dan — June 27, 2011 @ 5:23 pm

  26. that’s a really interesting point, Romo. it can also vary on a lineup-to-lineup basis. a low AVG/high OBP guy like Swisher is probably more valuable to the Yanks’ lineup than he would be to the Mariners’.

    Comment by Josh — June 27, 2011 @ 5:24 pm

  27. Good article. I still really like the triple slash. As you point out, it’s the perfect combination of brevity and descriptiveness. You get a lot of information about the style of player you’re looking at out of those 14 characters.

    Comment by Matt — June 27, 2011 @ 5:33 pm

  28. Shut up

    Comment by Snoth — June 27, 2011 @ 5:37 pm

  29. Does anyone know a quick way to estimate wOBA from a triple slash line? I know that two equal triple slash lines can have different wOBAs but I’m just looking for an estimate better than the second number

    Comment by bender — June 27, 2011 @ 5:39 pm

  30. Check the post again…

    Oh heck, I”ll give it to you now, again, because I’m a swell guy:

    ((OBP*1.7)+SLG)/3

    That’s a pretty rough guide, works pretty well.

    Comment by Matt Klaassen — June 27, 2011 @ 5:49 pm

  31. But SLG is affected by AVG.
    In 2009, Ichiro had a .465 SLG (11 hr’s) and David Ortiz had a .462 SLG (28 hr’s)
    Shouldn’t this also be considered a flaw to SLG (and in turn, a flaw to OPS)? I mean, what exactly, is SLG trying to tell you?
    Seems to me that not all SLG’s are created equal.

    Comment by Norm — June 27, 2011 @ 5:55 pm

  32. Why isn’t strikeout rate a part of wOBA? Seems to me the ability to put the ball in play has some value given that some percentage of the time the defense will muff the ball. Errors aren’t completely irrelevent in the sense that some players are more prone to benefit from them purely because some players create more opportunities for the defense to screw up. Probably not going to add a lot to wOBA, but seems worth including, there is definitely some offensive value in just putting a ball in play.

    Comment by corey — June 27, 2011 @ 6:06 pm

  33. Why did you allow the use of ‘PA’ in the latter but not the former? And how about a common stat like TB?

    ((H + BB +HBP) / PA) + (TB / AB)
    vs
    (.70*BB + .73*HBP + .89*1B +1.27*2B +1.61*3B + 2.07*HR +.25*SB – .50* CS) / PA

    Though with some creative rounding and dropping SB, CS and HBP

    ((H + BB +HBP) / PA) + (TB / AB)
    vs
    (7*BB +9*H +4*(TB-H)) / 10*PA

    Still says that Ledee was six points better than Kennedy in 2000 and that Braun and Fielder were about even in 2010 (a <2% gap, .376 v .370) and it is still a touch shorter than OPS [assuming you don't actually have OBP and SLG] :)

    Comment by Eric R — June 27, 2011 @ 6:13 pm

  34. Oh my bad sweet thanks

    Comment by bender — June 27, 2011 @ 6:18 pm

  35. I get that they’re expected run values for events on the field, but I don’t understand how those linear weights are being derived. What I’m really thinking about here is that under different types of game conditions, different events might have different values. So if I were interested in using wOBA for a college or a high school team how would I know what those coefficients (or linear weights) should be? Part of the insight of this stat seems to be that different events have a particular value towards run creation over other events. If that wasn’t the case then a SLG% would be sufficient. So where do those weights come from? That link you gave me didn’t seem to explain that.

    Comment by corey — June 27, 2011 @ 6:20 pm

  36. You could poentially add them in to wOBA. Depends on what you want to use it for. For example, FanGraphs includes SB/CS in wOBA, but the original version (http://www.insidethebook.com/woba.html) didn’t because in The Book they wanted to focus on pitcher/hitter match ups.

    That version did include Reach on Errors, which I like, but not everyone does. I’m not sure why we don’t use it here, but I think it has to do with the data sources used by the site.

    As for strikeouts, you could include those, too, but on average, a strikeout is about 2 thousanths (I think) of a run worse than other outs, so it wouldn’t make much of a difference.

    But yeah, you can get very detailed with linear weights and differentiate between not just strikeouts and other outs, but between ground outs and flyouts, infield singles versus others. Of course, as you get more and ore detailed, the gains in run estimation “accuracy” get smaller and smaller…

    Comment by Matt Klaassen — June 27, 2011 @ 6:21 pm

  37. unless “putting the ball in play” results in hitting into a DP…

    Comment by ToddM — June 27, 2011 @ 6:21 pm

  38. Sorry if my many comments here are getting tedious, but I’m skeptical of a lot of the sabermetric stats (particularly for pitching and defense), but I really like the idea behind wOBA and am trying to understand it a little more.

    Why haven’t I seen anyone use a wOBA for pitchers? I’m a big fan of WHIP, but WHIP is subject to the same problem that on base percentage is subject to for hitters, it treats all methods of reaching base as being identical. So why wouldn’t we use a wOBA for pitchers?

    Comment by corey — June 27, 2011 @ 6:37 pm

  39. I still use the slash for quick and dirties. When I look at it I always think of what is “good” for each. So a .300/.400/.500 would be what I judge players off of. If one player is further off “good” than the other, he’s the lesser player.

    as far as the guy who says that both are equally difficult to calculate…you realize that slugging and on base are typically listed anywhere except in like pregame lineup cards. So you just add those. wOBA is never listed and you need a lot of calculations for that.

    Comment by Antonio Bananas — June 27, 2011 @ 6:40 pm

  40. O, and if you want to know how far behind the general population is, get this. In my Sports Economics class me and my professor were discussing player value (he was explaining how A-Rod is underpaid). I asked him if he looked into different performance valuations, like sabermetrics. He said “you mean the more complex stuff like slugging percentage?”.

    So, to my professor. Who is by no means “dumb”. Who will randomly use calculus derivatives to prove a point even though no one is arguing with him and he just wants to show off, thinks of slugging percentage as a “complex sabermetric statistic”.

    Comment by Antonio Bananas — June 27, 2011 @ 6:44 pm

  41. Not fair. You used PA for wOBA, but not for OBP. You also could use TB/AB for slugging %

    But yes, calculating OPS on the back of an envelop is probably not a common occurrence. On the other hand, explaining OPS is easier if you’re not talking to a nerd who questions the different denominators.

    Comment by TK — June 27, 2011 @ 6:46 pm

  42. is the difference in OPS and wOBA really greater than the systematic error/variance in wOBA over a full season??

    Not that wOBA isn’t more accurate, but 99 time out of 100 you are just splitting hairs.

    Comment by zenbitz — June 27, 2011 @ 6:48 pm

  43. I’m pretty sure wOBA doesn’t work outside of MLB, at least not with the same numbers. I’m guessing you could theoretically do it. On the other hand, stats like BA/OBP/Slugging are worthless in comparing talent across different levels without context (i.e. a HS player with a 1.500 OPS could be great or could play in a shitty league)

    Comment by TK — June 27, 2011 @ 6:51 pm

  44. I think FIP is fairly close to being a “wOBA for pitching”

    Comment by TK — June 27, 2011 @ 6:55 pm

  45. I disagree, because FIP doesn’t think singles, doubles, and triples exist, and that’s my problem with FIP. You could easily just calculate the wOBA of batters facing a pitcher, I don’t see why nobody does.

    Comment by corey — June 27, 2011 @ 7:13 pm

  46. I’m fairly stat-minded, and I can’t relate to a whoaba. I don’t know what a poor one is, I don’t know what a good one is, I don’t know how much better a .301 wOBA is than a .300 player (other than one “unit” of course). At least the other stats are actually measuring something, and I know from looking at history what decent OBPs are. Which is why I don’t understand why you get abstract with a stat. It’s a percentage of nothing, really, so why not go to a RC-type scale? It’s pretty easy to grok the idea of “this is how many runs a team would score with 9 copies of this player in the lineup”. And if I’m understanding wOBA, that’s all it is, just trying to take batter events and translate them into runs. To me going to a unitless percentage is a step back from RC.

    Comment by Dale — June 27, 2011 @ 7:18 pm

  47. Any saberist worth his salt has a pocketful of napkins with a chart of wOBA coefficients for every year printed on it.

    You do the calculation on the back of said napkin.

    To impress your date.

    But ESPN said we’re all going dateless. And they’re ESPN!

    Comment by MikeS — June 27, 2011 @ 7:46 pm

  48. I tend to think that in your example, Braun’s line would be preferable, given that they have equal wOBAs, I would think the one that is derived from more hits would be better, since a hit is a little more valuable than a walk.

    Comment by Date Mike — June 27, 2011 @ 8:00 pm

  49. How much better is wOBA as a predictor of runs scored over OPS?

    A poster on Orioles Hangout (http://forum.orioleshangout.com/forums/showthread.php/107778-Another-thread-about-stats) took four different ‘rate’ stats from all MLB teams in 2009, and 2010 (BA, OBP, OPS, and wOBA) to see how these four correlate with run production. He graphed each team’s run production in 2009, and ’10 as a function of their BA, OBP, OPS, and wOBA and calculated r^2 for a linear regression each time. The closer r^2 is to 1…….the better the fit…….herego the MORE likely that the given stat will be able to give you meaningful prediction power for run production.

    Calculated Comparing OPS to wOBA vs. Runs Scored

    In 2010
    R^2 for AVG = 0.46 (pretty poor fit)
    R^2 for OBP = 0.77 (decent fit)
    R^2 for OPS = 0.891 (pretty good fit)
    R^2 for wOBA = 0.908

    In 2009
    r^2 for OPS = .919
    r^2 for wOBA = .923

    So, is it fair to say that wOBA looks like a improvement over OPS, but the improvement is slight?

    Comment by Chris — June 27, 2011 @ 8:21 pm

  50. The coefficients are derived on average. Since whether a guy gets a hit with a guy on third or nobody on is pretty much random chance (or at the very least evens out over the course of a season) then as a rate stat it works.

    Fangraphs has a leaderboard for weighted plate appearances that give you more of what your looking for.

    Comment by Deadpool — June 27, 2011 @ 8:33 pm

  51. Wow that’s beautiful. The difference between a 0.891 and a 0.908 R-Square isn’t at all enough to be conclusive. That’s cool.

    The way I look at OPS is that it’s production. High slugging or high on base kinda is HOW he gets his production.

    To use an example, the article comparing Halladay to Koufax. They’re both awesome, but they do it in different ways. Two guys with .980 OPS are awesome players, if one has a high OBP and the other a high SLG then you kinda gotta say that they’re still about the same, just do it differently.

    Did he also do a study on pitching statistics?

    Comment by Antonio Bananas — June 27, 2011 @ 8:43 pm

  52. I keep track of the Royals pitcher wOBA on a sidebar here: http://kcbbh.blogspot.com/

    The main reason *not* to use it is because it is defense-dependent, but I still find it interesting. Would be great if Fangraphs added it. Tom Tango advocated for it here, saying he uses it all the time: http://www.insidethebook.com/ee/index.php/site/comments/pitchers_woba_leaderboard/

    Comment by Royal_Heritage — June 27, 2011 @ 8:44 pm

  53. I think the run environment plays a part in the calculation, so that relationship might vary slightly in extreme offensive seasons.

    Comment by Phantom Stranger — June 27, 2011 @ 8:45 pm

  54. He did run them all seperate right? Not multiple? Coz then there’d likely be multicollinearity seeing as a lot of the stats contain the others.

    Comment by Antonio Bananas — June 27, 2011 @ 8:45 pm

  55. Because a pitcher has relatively little control over balls in play aside from influencing whether they’re on the ground or not. Even if you want to throw the idea if luck or bad defense (the traditional arguments for DIPS theory) you at least have to admit that a pitcher could throw the same pitch to the same spot to two different batters with wildly differing results.

    Comment by Deadpool — June 27, 2011 @ 8:46 pm

  56. Tango designed it so league average OBP was equal to league average wOBA. not sure if fangraphs is the sane but it shouldn’t be too far off.

    Comment by Deadpool — June 27, 2011 @ 8:49 pm

  57. I’m not sure I follow why that would be. Could you explain more?

    I ask because whether it’s 1 run or 10 runs, the correlation should be the same. The stats may be lower numbers (lower AVG, SLG, etc), but the correlation between the stats and number of runs should be the same.

    Comment by Antonio Bananas — June 27, 2011 @ 8:49 pm

  58. Antonio – My understanding was that he did run them seperate.

    Comment by Chris — June 27, 2011 @ 8:50 pm

  59. If you know what a decent OBP is, you know what a decent wOBA is. It is supposed to be on the same scale as OBP.

    Comment by Adam — June 27, 2011 @ 9:02 pm

  60. That’s kind of like saying if one guy has 10 dimes and another guy has 4 quarters, the guy with the quarters is doing better, because even though they have the same value, a quarter is worth more.

    Comment by Adam — June 27, 2011 @ 9:05 pm

  61. I think one thing that keeps fringe/newbies from embracing wOBA is that it’s scaled to OBP rather than something more familiar like AVG. OPS is different enough from AVG that it’s not confusing. OPS also has the convenient 1.000 at the upper echelon, and tiers formed by the .900, .800, etc. levels which make it really accessible, even if you don’t know the calculation or why it’s good/bad. AVG doesn’t have the same convenience as OPS but it’s got far more familiarity than OBP, especially for the folks that would be abandoning something like OPS.

    That aside, what I’d like to see is a new triple slash. When I read the piece I wasn’t paying enough attention and thought that Braun’s and Fielder’s SLG must have been transposed when I read that Fielder had more power. I would rather see ISO in the 3rd spot and not SLG since I routinely make that mistake. I do like OBP in the 2nd spot though, for some reason. So how about AVG/OBP/ISO all normalized to 100:

    Prince Fielder: 102/123/143
    Ryan Braun: 118/112/135

    Comment by jmr — June 27, 2011 @ 9:07 pm

  62. Maybe…it just makes sense intuitively that the guy with more hits would be more valuable in general since a hit usually moves runners on base further than a walk. I know that players should never be evaluated based on if guys are on base when they come to bat, but given their equal performances it makes sense.

    Comment by Date Mike — June 27, 2011 @ 9:16 pm

  63. Any stat that says a homer is only twice as valuable than a single is flawed.

    Comment by supgreg — June 27, 2011 @ 9:18 pm

  64. If the run environment is high, then a given individual event is more likely to lead to a run, so two players in two different environments can put up different wOBAs with the same OPS.

    Comment by Deadpool — June 27, 2011 @ 9:19 pm

  65. I see what you’re saying I think. A single in one run environment, could just be a single, but the other could drive in a run.

    Still though, that study should at least lead people to believe OPS and wOBA aren’t too different. Which I believe. Where did they get the numbers for wOBA? Like the .7, .73, etc.

    Comment by Antonio Bananas — June 27, 2011 @ 9:27 pm

  66. Yea, but if that pitcher knows the hitter inside and out, and sequences his pitches right, that variance is a lot smaller. Let’s say you study tape and the hitter is always super aggressive on 3-1. You position your defense to pull, purposely get the count to 3-1, then throw a change. I bet he rolls it over to the defense.

    You can’t just discredit any pitch not resulting in a HR, BB, or K. It’s why I don’t like FIP. Some pitchers actually do pitch to contact successfully.

    Comment by Antonio Bananas — June 27, 2011 @ 9:43 pm

  67. no one is debating that the ability to induce ground balls is a skill. the defense will make an out out of a ground ball about 72% of the time, given an average runner, which sounds like a pretty good bet to me. the reason FIP ignores that, however, is that the pitcher does not control if the defense actually does field it. Are you suggesting that this hypothetical instance you suggest (which includes deliberately wasting three pitches, shifting the defense, and hoping that this batter is really frigging stupid) is A) at all common, and B) repeatable?

    Comment by williams .482 — June 27, 2011 @ 10:01 pm

  68. 1) I was mostly joking around.
    2) PA’s isn’t the denominator of OBP. SH’s aren’t included. Though, in all fairness, they shouldn’t be included in wOBA either.
    3) I have an excel spreadsheet on the back of the napkins I mentioned.

    Comment by Yirmiyahu — June 27, 2011 @ 10:15 pm

  69. You realize that fact (i.e., moving runners over) is included in the underlying values for wOBA? According to wOBA (2010), a walk is worth .70 and a single is worth .89.

    Comment by Yirmiyahu — June 27, 2011 @ 10:26 pm

  70. Agreed. Fortunately, neither OPS nor wOBA is that stat.

    Comment by Yirmiyahu — June 27, 2011 @ 10:29 pm

  71. You call it a triple slash, yet there are only two actual slashes. How can I trust your crazy maths?!?!

    Comment by Rick_S — June 27, 2011 @ 10:32 pm

  72. while i like this, i think a more appropriate state would be to somehow normalize ISO on the same scale that OBP is, where the average ISO is scaled to be the average OBP, so that the relative OPS, or in this case OPI, would be something that people could look at and definitively understand and consider one number to be an accurate picture of a players value at the plate.

    Comment by Rex — June 27, 2011 @ 10:42 pm

  73. *stat

    Comment by Rex — June 27, 2011 @ 10:43 pm

  74. Antonio, it won’t let me reply directly to you for some reason, so I’ll have to do it this way.

    “Where did they get the numbers for wOBA? Like the .7, .73, etc.”

    That’s the question I’ve been trying to ask all day. I haven’t gotten a satisfactory answer. I think the answer is that somebody out there is running regressions on events in a game and whether a run is scored. The coefficients on those events are the linear weights applied in wOBA. If that’s not the answer then I’m baffled.

    Comment by corey — June 27, 2011 @ 10:53 pm

  75. “You can’t just discredit any pitch not resulting in a HR, BB, or K. It’s why I don’t like FIP. Some pitchers actually do pitch to contact successfully.”

    I completely agree with this.

    Comment by MB — June 27, 2011 @ 11:05 pm

  76. Because not allowing singles or whatnot is not a repeatable skill.

    Comment by Will — June 27, 2011 @ 11:12 pm

  77. Accurate to what?

    Sure, an OPS comp is probably better for quick, “who is better” questions, but there’s also the scaling issue to consider.

    Comment by Will — June 27, 2011 @ 11:14 pm

  78. I think it’s important to remember that wOBA is about the value of each offensive event and not the difficulty of achieving it. Because of that, it is the premier stat to use to determine the MVP, but if you are looking for which player is better, you can’t rely on it. Reaching on an error and getting hit aren’t repeatable performances, yet they are valued higher than walks and singles respectively, which are performances that more directly portray a player’s actual skill. While this isn’t a flaw in wOBA, a user must keep in mind what the stat is intended for. And the difference between looking for the best player and looking for the most valuable player is slight and using wOBA is a safe bet either way, but using OPS wouldn’t be that different from wOBA either, and since that distinction was the point of the article, I figured this distinction should be mentioned.

    Comment by Max — June 27, 2011 @ 11:14 pm

  79. First off, xFIP uses FB% instead for that very reason because getting GBs is a skill and keeping HRs in the park seems to be less so (for most pitchers).

    So xFIP gives the pitcher credit for giving his defenders the superior chance to field the ball.

    In addition a wOBA style stat for pitchers wouldn’t work because, again, even if you give the pitcher credit for good contact versus bad contact the difference between a good contact single and good contact double or HR is still largely on the hitter and defense.

    Comment by Deadpool — June 27, 2011 @ 11:19 pm

  80. Exactly, since hits are valued higher he has to get more walks to equal the wOBA of the guy with the singles. He’s getting equal value because he’s actually getting on base often enough to outweigh the other guys singles.

    Comment by Deadpool — June 27, 2011 @ 11:30 pm

  81. That’s kind of what I meant, but more over the course of a season. Like this season. Singles are more scarce, so they’re more likely to contribute to a win and thus worth more on the wOBA scale, but OPS treats it the same as other seasons.

    Comment by Deadpool — June 27, 2011 @ 11:31 pm

  82. Are you saying Biggio wasn’t skilled at getting hit? Heretic!

    Comment by Deadpool — June 27, 2011 @ 11:35 pm

  83. Will- Tell that to Mariano Rivera.

    Comment by delv — June 28, 2011 @ 12:12 am

  84. Sorry, that should be Win Probability added. It means what I said, but for some reason I see wpa and my brain substitutes the wrong words in.

    Comment by Deadpool — June 28, 2011 @ 12:44 am

  85. Wow your printer prints on Napkins. Now you can pre-print all the pickup math you use in bars! Any girl worth knowing can understand the equations.

    Comment by kick me in the GO NATS — June 28, 2011 @ 12:59 am

  86. Chris:

    There are a lot of complex issues here, and while that’s an interesting way of doing it, I think the method to compare run estimators used in this older two-parter by Colin Wyers is superior, and he explains why therein:

    http://www.hardballtimes.com/main/article/the-great-run-estimator-shootout-part-1/

    http://www.hardballtimes.com/main/article/the-great-run-estimator-shootout-part-2/

    You can see that wOBA and other linear-weights based measures are far superior in matching the actual weights of in-game events, while OPS underrates walks and heavily overrates extra-base hits.

    Thanks for the comment.

    Comment by Matt Klaassen — June 28, 2011 @ 1:09 am

  87. Perhaps you should have taken an English class as well.

    “Me and my professor” should be “my professor and I.”

    Comment by AJS — June 28, 2011 @ 1:35 am

  88. A big part of why GPA didn’t catch on because originally it was called “Gleeman Production Average” (this was on his blog, several years before HBT came into existence). He was already seen by many on rsbb and baseball primer as somewhat of a pompous ass, so there wasn’t a lot of support in the online sabermetrics community to use something named after Gleeman, which is why I always thought it never really caught on. Eventually he changed it to “Gross Production Average” but by that time the conversation had moved on to other metrics.

    Comment by walt526 — June 28, 2011 @ 1:43 am

  89. face it, if there were more girls who understood sabremetrics, we wouldnt be here commenting on fangraphs, we’d be avidly hunting them down and marrying them

    Comment by jim — June 28, 2011 @ 3:14 am

  90. what is it good for?

    Comment by jim — June 28, 2011 @ 3:20 am

  91. will, that thought has been pretty debunked by now, i would think, or were you not around for the dave cameron posts on matt cain over the offseason?

    Comment by jim — June 28, 2011 @ 3:25 am

  92. maybe one reason wOBA isn’t catching on is because people actually say it “whoaba.” nothing that you pronounce like that could be taken seriously, it’s “doubleyou-oh-bee-aye,” it sounds so much more scientific and legitimate

    Comment by jim — June 28, 2011 @ 3:29 am

  93. you do know that wRC+ is the comparative equivalent of wOBA, right?

    Comment by jim — June 28, 2011 @ 3:34 am

  94. Unfortunately wOBA inlcudes SB and thus is not 100% a hitting metric. OPS is 100% hitting.

    Comment by pft — June 28, 2011 @ 3:37 am

  95. The coefficients for wOBA change with run environment though.

    Comment by pft — June 28, 2011 @ 3:42 am

  96. Great Read. You can learn alot about the type of player with the slash line, wOBA only gives you if they are a good player.

    Your part on Braun and Fielder gives a perfect example of why for instance the Brewers should have Braun first in the batting order. While they both have very similar wOBA, you can easily see that Braun isn’t as patient but gets more hits.

    Therefore, giving him protection from Fielder will give him better pitches to see.

    Comment by Derek — June 28, 2011 @ 8:27 am

  97. Because we’ve developed this intuition around BA and SLG, we think that they’re actually measuring something. But they’re really just made-up numbers. One problem is the concept of the “at-bat” (i.e. only some PAs “count”). If BA was H/PA, it would be measuring how frequently the guy gets a hit, which would be at least somewhat interesting. But it isn’t.

    For SLG, there is the additional issue of TBs. By adding 1*1B, 2*2B, etc. I suppose you’re actually measuring something (i.e. the # of “bases”), but it isn’t something that tells you anything in particular. Point being that folks think that the triple slash stats are the elements of offensive production, but they’re not. OBP is, but the other 2 are just numbers.

    Then there’s the issue of the “error” and assigning 100% credit/blame for discrete events….

    Comment by Pierre — June 28, 2011 @ 9:49 am

  98. SLG tells me that, generally, the more powerful batter is going to wind up with a substantially higher one than a player w/o power, but if you hit .352 on the season with 297 TB, you’re going to crack the SLG leaderboard.

    That’s an outlier season by Ichy. His career SLG is .425 while Papi’s is .545

    “slugging and batting average aren’t good for valuation”
    That rational for the use of wOBA is why I doubt it ever catches on. I assume there are very few people outside the saber community that actually believe that statement, and even those who do, do not cheer for their team’s best hitter to earn a walk while down 2 runs in the 8th inning.

    Comment by Sultan of Schwinngg — June 28, 2011 @ 10:39 am

  99. I still see no excuse for a site like FanGraphs to use the traditional triple slash. I groan every time I see it. Why not just use wOBA except in contexts when on-base vs. power matter and then use OBP and ISO, just as you might use HR’s or BABIP’s in contexts where they matter.

    Comment by GiantHusker — June 28, 2011 @ 11:24 am

  100. That’s what ISO is for.

    Ichiro .113
    Ortiz .224

    His point still stands. If the run environment has drastically changed, the linear weights need to be adjusted to accurately reflect what is valuable.

    Comment by Matt — June 28, 2011 @ 12:12 pm

  101. Why is a HBP worth more (0.73) than a walk (0.70)? I’ve asked this question in multiple places and never gotten a concrete answer.

    Comment by Corey Seidman — June 28, 2011 @ 12:22 pm

  102. This might be a completely ignorant question, but here goes:

    Since wOBA is based on linear weights (which implies linear regression), were the assumptions required for linear regression to be valid ever verified? In other words, did the analysis pass a D’Agostino-Pearson normality test, and a runs test?

    Once upon a time at my job, I got hammered for assuming linear regression was valid without actually proving it. I’ve never looked into whether wOBA proved it, though.

    Comment by gonfalon — June 28, 2011 @ 1:16 pm

  103. All the linear weights are based off of real life historical data. Going back inning-by-inning through baseball history, comparing HBP’s to BB’s in situations that were otherwise the same, the innings with the HBP ended up producing slightly more runs than the situations with the BB’s. That’s just what the numbers say.

    The practical answer is that BB’s are quite often calculated risks. Say, the cleanup hitter is up with first base open. So the pitcher avoids throwing strikes, at the risk of walking him. HBP’s, on the other hand, are usually random mistakes that are not tailored to the situation.

    Comment by Yirmiyahu — June 28, 2011 @ 1:23 pm

  104. You can’t do laundry with dimes though.

    Comment by Ben — June 28, 2011 @ 1:32 pm

  105. You may be reading “linear weights” and thinking “regression coefficients”. There actually isn’t any regression involved.

    Comment by Pierre — June 28, 2011 @ 1:44 pm

  106. Wait, there isn’t? Then I’m back to square one, where are these linear weights coming from? I’ve been assuming they were regression coefficients. If there’s no regression involved then how do we know that’s the appropriate weight of a given event?

    Comment by corey — June 28, 2011 @ 2:25 pm

  107. I’d sugeest reading up. I think it’s explained on this web site as well as Tom Tango’s. Basically, they’ve examined all the baserunner/# of outs permutations (e.g. runner on first, one out), toted up the # of runs scored for each event (strikeout, single, etc), and then weight-averaged the results.

    Comment by Pierre — June 28, 2011 @ 2:44 pm

  108. Williams, my example of the 3-1 count I was basically just saying that with a smart pitcher who knows what hitters do and has decent stuff and can locate (Halladay, Maddux, most guys to a much more mediocre extent without great stuff), actually do stuff like this. They’ll set guys up. You say “intentionally waste 3 pitches” like it’s crazy. You DO waste pitches. Although they’re not “waste” they’re “set up” pitches. Inside, outside, fast, slow, etc. You move them around and change speeds, change eye levels. It’s how you pitch.

    Comment by Antonio Bananas — June 28, 2011 @ 6:03 pm

  109. Who gives a shit. The english language isn’t based on actual rules, it’s a clusterfuck of stuff we’ve pieced together. It’s why “lose” and “nose” even with the same vowell-consanent structure, have totally different “o” sounds. It’s a language pieced together over years. People who nitpick little semantics are losers who have nothing better to do with their time. It’s like that douche bag who corrects you for how you spelled “effect” and not “affect” or “who” instead of “whom”. Language is about understanding and communication. If someone understood everything you said and it’s not a formal setting where perfection is necessary, then there is nothing wrong with that was said.

    Comment by Antonio Bananas — June 28, 2011 @ 6:06 pm

  110. I think wOBA is incomplete because if you want to look at the run producing ability of a person, an out isn’t always equal as any other out.

    If you want to make an offensive stat that shows the true run producing ability of someone, you’d have to include just about everything.

    total up all ground ball outs, a ground ball up the middle to a SS will score a run from third many times. A K never will. A deep flyball will score a run more often than a ground ball to the SS.

    So, scale how often a K will score a run (0 times) the chances of a ground ball up the middle will score a run (probably super low), the chances a ground ball to the corners will score a run (probably even lower) the chances a shallow flyball will score a run (maybe the same as GB to middle), the chances an average flyball scores a run, the chances a deep flyball scores a run, the chances a bloop single scores a run, the chances a little squibble between infielders scores a run, the chances a line drive single scores a run, the chances a double up the line scores a run, the chances a double in the gaps scores a run, the chances a walk scores a run, and the chances a home run scores a run.

    You plug in the amount of times a guy hits each one of those and bam you have you “likely run production” for that player that year. Otherwise you’re either discrediting stuff as “luck” and so you don’t include it, which is stupid, or you’re putting weights on things that aren’t even the same. An strikeout is worse than a deep flyball because you can actually score on a deep flyball and a single isn’t always a single.

    My biggest problem with sabermetrics is the amount of times you go into extreme detail and then just discredit everything else. Putting weights on FIP for BB, K, and HR, but then saying “fuck it” for everything else. It’s really REALLY stupid IMO. It’s like a car company building a car and being very precise on alignment and putting the doors on, but just kinda flopping the windshield on. Okay, you got some stuff down pretty good, but you totally forgot the rest.

    Comment by Antonio Bananas — June 28, 2011 @ 6:15 pm

  111. O, and also look at the probability a guy scores based off what he does too. Steal, single, etc.

    Comment by Antonio Bananas — June 28, 2011 @ 6:16 pm

  112. I was wondering, if batting events are weighted, shouldn’t events like GIDP and fielder’s choice be weighted more negatively than a single out since through poor batting the hitter caused someone already on base to get out? Hitting behind runners is a batting skill. Just a thought, I really like the wOBA stat. On the same note, shouldn’t being caught stealing at third be more negative than a CS at second since you were in scoring position already? And should successfully stealing third be worth more than second since you are now easier to bring in, even on a sac hit? Perhaps, in the grand scheme it wouldn’t make much difference on individual outcomes, but a thought?

    Comment by flyingelbowsmash — June 28, 2011 @ 7:23 pm

  113. wOBA is runs per plate appearance, according to linear weights, multiplied by 1.15 so that the mean (about .330) is the same as on base percentage.

    Comment by Brian Cartwright — June 28, 2011 @ 7:28 pm

  114. I don’t think GIDP should be, but maybe weight a ground ball to the shortstop with a negative value X, a ground ball to the third basemen (presumably harder to turn 2) as another negative but not as much.

    I really think there should be a study that looks at every time of hit, where it was hit, outs, BBs, Ks, etc, every possible outcome of an at bat. Put “run potential” values to that, both the value of a run scoring on that even and the chances of that event causing a future score (say 1 batter?).

    You can’t really say ‘well the difference between a groundout to short and a hit that’s a grounder between short and third is luck” because everything is luck. You get a walk on a bad call, you’re lucky it wasn’t a better ump, you hit a home run in a small park with the wind blowing out, you’re lucky there too. You blast a ball but the wind is blowing in and it’s a huge park, you’re unlucky.

    I say divide the field up into zones. Look where each ball was hit and with what speed and what kind (ground, fly, liner, etc) and do a season long study of what the run potential of each hit in those places produce. Don’t look at singles, doubles, homers, etc. Just look at distance, speed, and type. Assign values to each zone and each type and each speed, add all that up and that’s your player’s “run potential” for that year.

    Much better than being really detailed about some things then discrediting others as being “luck”.

    Comment by Antonio Bananas — June 28, 2011 @ 7:58 pm

  115. Corey: go to my site (tangotiger.net) and read “How Runs Are Really Created”.

    If that’s still not enough, go to your library, and ask for The Hidden Game of BAseball (Pete Palmer).

    You can also try The Book, which is available for free at Amazon’s Look Inside.

    I know, I know, I’m giving you work. But, that’s how you should catch fish.

    Comment by tangotiger — June 28, 2011 @ 8:08 pm

  116. That’s what RE24 is. It answers your question exactly. That runs by the 24 base/out states.

    There’s also WPA/LI which is like RE24, except it also includes inning and score.

    I think the problem is that people think there’s a problem with one stat, when really, it’s been handled in a more advanced stat. Think of wOBA as SABR 201, but you are asking for more details, something a stat in SABR 301 or SABR 401 will answer. We’ve got that. You just have to look.

    Comment by tangotiger — June 28, 2011 @ 8:12 pm

  117. Alot of the issues being asked, very legitimate issues, is already being handled by other stats.

    You can’t blame OBP (a SABR 101 stat) for valuing a HR and BB identically.

    You can’t blame wOBA (a SABR 201 stat) for treating the run value of a HR the same across all 24 base/out states.

    You can’t blame RE24 (a SABR 301 stat) for not distinguishing between a close and late game and an early inning blowout.

    You can’t blame WPA/LI (a SABR 401 stat) for, well, for not understanding it.

    The best thing to do is rather than saying that a stat doesn’t cover it, instead say: “which stat DOES cover this scenario”. And, then I can tell you which one does.

    To just out and out presume that because a SABR 201 stat doesn’t cover a scenario without understanding or knowing that there’s a SABR 301 stat out there just keeps us in circles.

    Ask where you can find it.

    Comment by tangotiger — June 28, 2011 @ 8:16 pm

  118. Genius isn’t making simple ideas complex but making complex ideas simple.

    That, in a nutshell, is my biggest gripe with Sabermetrics. Not what things do and do not cover.

    Comment by Antonio Bananas — June 28, 2011 @ 8:24 pm

  119. Stop having a gripe, and instead tell us what complex idea of baseball you have, and we’ll tell you if someone has made it simple.

    Comment by tangotiger — June 28, 2011 @ 9:38 pm

  120. For example, you idea about looking at where a ball was hit and not worrying about whether it actually was a single or double, etc does get handled in some form. Things like xBABIP, etc, have the basic idea for it.

    Again, I think your gripe is that it’s too hard to find the stuff. We’ve made complex things simple.

    Comment by tangotiger — June 28, 2011 @ 9:40 pm

  121. FIP seems worthless to me. I think it’s definately taking one of the those simple things and making them complex. I hate that you take anything not a BB, K, or HR out. You could do my zone idea with pitching and instead looked at it from the pitchers’ perspective.

    Comment by Antonio Bananas — June 28, 2011 @ 10:24 pm

  122. A lot of thingts that don’t really tell you much more than simpler stats. Like up earlier in the thread, the guy with the study of OPS and wOBA. OPS is very easy to understand and it tells you scoring pretty much as accurately as wOBA. No need to create this weighted equation that has different weights for everything (but doesn’t include everything). Just use OPS and don’t shit on it because it actually is pretty much just as good as wOBA.

    Comment by Antonio Bananas — June 28, 2011 @ 10:29 pm

  123. Antonio, if you can’t see the redeeming qualities of FIP, and if you don’t see it as taking something complex and making it simple, then maybe sabermetrics just isn’t for you.

    Comment by tangotiger — June 28, 2011 @ 11:54 pm

  124. Maybe not. I mean, I think it’d be a lot more simple to just use things like variance but that’s out of the question.

    I see some use in FIP, but people use it as gospel around here. Which is annoying.

    Comment by Antonio Bananas — June 29, 2011 @ 1:41 am

  125. Then in that sense, your gripe is with users, not the discipline.

    Comment by Tangotiger — June 29, 2011 @ 11:40 am

  126. Antonio, send me an email:

    tom~tangotiger~net

    Comment by Tangotiger — June 29, 2011 @ 11:42 am

  127. I’m interested in getting more detailed wOBA, tailored to player types.

    I’m going to go out on a limb here and suggest that a walk to Prince Fielder is not going to have the same run value as a walk to Brett Gardner.

    I’m not talking about “clogging the bases”, but I am talking about run values per event base on speed and base-running skill.

    I wonder how much that changes things?

    Comment by CircleChange11 — June 29, 2011 @ 12:13 pm

  128. at what?

    Comment by Antonio Bananas — June 29, 2011 @ 7:52 pm

  129. I do have a problem with the users because most of them are smug and the stats they use aren’t really that effective.

    The other main problem I have is that they didn’t think to make the saber stats easily integrated. like wOBA, what’s a good wOBA? I do like WAR and I do like the RC stuff.

    Comment by Antonio Bananas — June 29, 2011 @ 7:58 pm

  130. Replace the ~ as appropriate.

    x at y dot z

    Comment by Tangotiger — June 30, 2011 @ 11:09 am

  131. wOBA is scaled exactly like OBP. If you know OBP, you know wOBA. If you don’t use OBP, then you won’t like wOBA.

    Comment by Tangotiger — June 30, 2011 @ 11:10 am

  132. Let’s take a step back – what do you want the stat to explain? What question are you trying to answer?

    Are you trying to isolate a player’s skill, or are you trying to measure what someone did? The two are very, very different, especially over a game, a month, or even a year. Your explanation here makes me think you’re more interested in the latter. FIP, as an example, is more interested in the former. It may be you don’t like it and don’t like readers speaking reverently about it because it’s got nothing to do with the question you’re trying to answer.

    You don’t like the fact that a bunch of things are excluded from stats like FIP because they’re quite volatile (something that’s often conflated with luck). That is, they’re isolating things that are stable from year to year. We can quickly get at an approximation of a pitcher’s skill in the measured variables, and we isolate this approximation from his teammates and random variation.

    Your proposed stat is all about random variation – so this may be of use as way to see if someone was horribly unlucky in the past (souped-up xBABIP)? Did you want it to be “for” something more than that, or is that what you wanted?

    Comment by marc w — July 8, 2011 @ 7:51 pm

Leave a comment

Line and paragraph breaks automatic, e-mail address never displayed, HTML allowed: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>


Current day month ye@r *

Close this window.

0.161 Powered by WordPress