Skip to content

How not to criticize sabermetrics

Jun 28, 2011, 10:30 AM EDT

Straw man

I like the idea of Bill Simmons’ new Grantland site.  I think we need more deep-thinking, longer lede sports writing.  Someplace where folks who consume their sports news in 35 bloggy bits a day (ahem) can go and breathe a bit. But really, if you’re gonna go through the trouble of building such a site, make sure it has a bit more rigor to it, will ya?

Setting me off is Jonah Lehrer’s piece today in which he criticizes sabermetrics.

Now, don’t get me wrong: I’m not one of those people who get all angry when someone criticizes stuff to which I’m partial. Everything is better when put to constructive criticism.  Only, Lehrer doesn’t do that. He offers criticisms that are completely unsupported with any reference to observable fact:

My worry is that sports teams are starting to suffer from a version of the horsepower mistake. Like a confused car shopper, they are seeking out the safety of math, trying to make extremely complicated personnel decisions by fixating on statistics … coaches and fans use the numbers as an excuse to ignore everything else, which is why our obsession with sabermetrics can lead to such shortsighted personnel decisions.

Except there isn’t one example cited in the entire article where an “obsession with sabermetrics” has led a coach or a general manager astray.  Not one “team that has suffered from a version of horsepower mistake.”  He notes that the Mavericks got a lot out of some players who aren’t statistical darlings, but his citation to the Lakers — who he appears to be setting up as one of those stats obsessed teams, but doesn’t quite come out and say it — is horribly wrongheaded given that they, you know, had won the previous two championships and previous three conference titles.

In the end, this is an aimless, cranky complaint at best. It’s a misleading strawman argument at worst.  If you’re going to accuse sabermetrics of leading organizations astray, shouldn’t you be obligated to cite a single example?

  1. gammagammahey - Jun 28, 2011 at 10:35 AM

    This is a joke, right? The Mavs and Mark Cuban were relentless proponents of advanced metrics in basketball. Lehrer says that the Mavs had a team that didn’t look good on the scorecards. That may be the case if you’re only looking at the stats that show up on scorecards.

  2. Kevin S. - Jun 28, 2011 at 10:37 AM

    And he ignores the fact that the Mavs are on the cutting edge of statistical analysis in the NBA. That they used statistical analysis to determine the optimal crunch-time unit for defeating the Heat.

  3. Mississippi RBI - Jun 28, 2011 at 10:37 AM

    They are just trying to get more traffic…

  4. phillysoulfan - Jun 28, 2011 at 10:38 AM

    Craig, are you telling me that you need that guy to tell you an example of how advanced metrics have led a team astray? Seriously, you’re kidding right? Does the Oakland A’s ring a bell?

    • normb11 - Jun 28, 2011 at 10:41 AM

      The Oakland A’s won a LOT of games with a mediocre payroll, I believe they were ‘most wins per dollar spent’ organization of the ’00’s.
      Are you saying that because they didn’t win a World Series their belief in sabermetrics failed?

      • drmonkeyarmy - Jun 28, 2011 at 10:44 AM

        No, but winning a singular game in an ALCS would have been nice.

      • The Common Man - Jun 28, 2011 at 11:09 AM

        Indeed it would have, but you have to admit that building a strong team on that low of a payroll for that many years is a pretty strong vindication of how the A’s conducted themselves at the start of the 2000s.

      • phillysoulfan - Jun 28, 2011 at 11:12 AM

        The A’s have not won anything since Alderson left. And if you look at Beane’s drafts they are worse and worse. He uses advanced metrics as a be all and end all when it’s simply another tool.

      • thefalcon123 - Jun 28, 2011 at 11:24 AM

        To phillysoulfan:

        In Billy Bean’s tenure as GM, his draft has netted an average 25.1 WAR from 1998 to 2005 (any drafts later than that are kind of pointless at checking at this stage, since most of those players would have barely gotten their feet wet in the bigs)
        The *only* team that has a higher WAR from their draft picks are the Cardinals, mostly due to some data-skewing 13th round pick in 1999. Billy Beane’s drafts have been among the best in baseball.

      • drmonkeyarmy - Jun 28, 2011 at 11:27 AM

        Common Man,
        Perhaps you are right. I wise just making a snide comment that despite their regular season success, post season success eluded them in that era. I do, however, think that regular season success is a better indicator of squad strength and such.

      • The Common Man - Jun 28, 2011 at 12:10 PM

        @Phillysoulfan

        You know that Sandy Alderson is a pretty big advocate for sabermetric thinking, right? If I remember correctly, he’s the one who introduced it to Beane.

    • Kevin S. - Jun 28, 2011 at 10:44 AM

      Also, lots of teams are using them. They aren’t the only factor involved in winning, and if half the league is using them, all those teams aren’t going to make the playoffs. Picking out one team that uses advanced stats as an example of how they don’t work is pretty foolish.

      • Mr. Jason "El Bravo" Heyward - Jun 28, 2011 at 10:56 AM

        Small correction: ALL of the teams in the MLB employ sabermetric analyses in their farm systems and when scouting talent. ALL of the MLB teams also employ actual humans to scout with their own eyes. This is the norm and it’s amazing that folks don’t get that. It’s like “hey, we use only stats and no humans” or “we only like players we regard with our own two eyes”…when in reality EVERY team has used both for awhile.

      • Kevin S. - Jun 28, 2011 at 10:58 AM

        You must have a pretty loose definition of sabermetrics if you think Ed Wade uses them.

        But seriously, you’re right. I didn’t mean that the cutting-edge teams were ignoring their scouts. I simply meant teams that were using more advanced statistics than others.

      • Mr. Jason "El Bravo" Heyward - Jun 28, 2011 at 11:07 AM

        Oh absolutely, you’re right. Some teams use them more or place a higher level of importance on sabermetrics. It always bugs me when someone brings up the A’s more recent struggles as if they are the only team ever to employ sabermetric analyses in baseball.

    • thefalcon123 - Jun 28, 2011 at 11:19 AM

      Oakland A’s: 1999-2006:

      1999: 89-75, 2nd Place
      2000: 91-70, 1st Place
      2001: 102-60, 2nd Place (wild card winner)
      2002: 103-59 1st place
      2003: 96-66 1st place
      2004: 91-71 2nd place
      2005: 88-74, 2nd place
      2006: 93-69, 1st place
      2007-Present: Other teams catch on that the type Sabermetric darlings the A’s are getting for cheap help win games. Players are no longer cheap. A’s stop winning 90 games.

      So…what the hell is your point?

      • b7p19 - Jun 28, 2011 at 11:23 AM

        Thank You.

    • explodet - Jun 28, 2011 at 2:14 PM

      Because they consistently lead the league in OBP and power and still lose in spite of it right? Oh wait, no, they’ve been near the bottom of the league in both of those since about 2006. Which merely strengthens the argument in favor of statistical analysis.

  5. drmonkeyarmy - Jun 28, 2011 at 10:42 AM

    I don’t like advanced sabermetrics. They rub me the wrong way. I don’t have any constructive criticism to levy, I just don’t like them. You know what, I actually do have a reasonable criticism to levy: given the sample sizes of the comparative statistics I would like to see them reported with standard deviations and p values so that the reader can assess the statistical significance of said sabermetric value.

    • Roger Moore - Jun 28, 2011 at 11:04 AM

      Three points:

      1) The question about standard deviations, p-values, and other error estimates applies to any statistic, not just advanced sabermetric ones. We all know that statistics from the first month of the season are more fun than useful in predicting full season performance, and that applies as much to BA as to VORP. I’ve even suggested (a long time ago on USENET) that MLB should replace their PA cutoff for batting statistics and use p values instead.

      2) Generating confidence intervals and p-values for some of those stats is inherently hard. How do you get the standard deviation of VORP? I guess there are probably some non-parametric approaches, but they’re going to be painful and time consuming. That’s especially true when you start to worry about error propagation. How do you deal with uncertainty in measured park factor? That’s not to say we should ignore the issue- a decent estimate of our uncertainty would already be a big advance- but it’s going to be hard.

      3) To some extent advanced statistics are already moving in that direction. One of the good things Baseball Prospectus has done is to move from providing a single prediction of players’ performance to providing a range of predicted values that reflect their uncertainty. That’s a lot better than most traditionalists have ever done.

      • drmonkeyarmy - Jun 28, 2011 at 11:12 AM

        Whether they are inherently hard to calculate or not is irrelevant. If one is going to promote the use of advanced comparative statistics such as wins over replacement, etc. then there has to be error estimates reported with that data. If such things were traditionally reported then I would have a lot more respect for the data. I have been trained to not take comparative data seriously unless “p” values or other error estimates are reported.

      • churchoftheperpetuallyoutraged - Jun 28, 2011 at 11:21 AM

        Whether they are inherently hard to calculate or not is irrelevant. If one is going to promote the use of advanced comparative statistics such as wins over replacement, etc. then there has to be error estimates reported with that data

        But most advanced sabr sites do argue this, in a way. Remember, it’s the ESPN/SI/Fox Sports/etc that write the articles “Johnny McAverage is on pace for 200 HR this year” after hitting 10 in the first 25 games, or that Frenchy has finally figured it out after hitting .350 in the first month. Guys like Tom Tango/Dave Cameron/Mike Fast/Colin Wyers/etc are the ones imploring people to let the games play out more so you can get a clearer picture of what’s going on.

        For instance, MGL’s UZR isn’t supposed to stabilize for three full years(!) worth of data and even then some question whether that’s enough. The problem isn’t in the statistics themselves, it’s the way people use them.

      • drmonkeyarmy - Jun 28, 2011 at 11:30 AM

        Fair enough, but show me the error estimates. That is all I am asking. Don’t tell me that this stat will be in flux for 3 years….report the value and show me a “p” value. That way I can evaluate said stat.

      • Roger Moore - Jun 28, 2011 at 11:37 AM

        @drmoneyarmy:

        Why do you demand p-values of advanced statistics but not traditional ones? Where are the standard deviations for batting average and ERA? Where are the p-values for hustle and being a team player?

      • drmonkeyarmy - Jun 28, 2011 at 11:43 AM

        Mr. Bond,
        I only demand “p” values in comparative data sets. For instance, I wouldn’t demand a “p” value for OPS but I would for WAR or VORP.

      • drmonkeyarmy - Jun 28, 2011 at 11:47 AM

        Again Mr. Bond,
        I’m not sure hustle and being a team player are quantified intangibles. Neither are providing a veteran presence or being a team leader. Quick example, Dannys Baez for the Phillies. He is terrible pitcher. There are a couple guys in AAA who could do a far better job than him. However, he is somewhat of a bullpen mentor. He spots mechanical flaws, helps the young pitchers, and translates for the non-English speaking pitchers. Hence, they keep him around even though he is awful. He has an inherent value to the team that cannot be quantified or compared to others. Hence, attaching statistical measures of error to such a thing would be ridiculous and impossible.

      • paperlions - Jun 28, 2011 at 12:54 PM

        Within the way that most fans/media use statistics, error estimates are irrelevant, to some degree, as these things referred to as statistics are not statistics at all but are parameters. Statistics are estimates of a whole based on only a sample, parameters are the value of the whole….in general, “statistics” in baseball are based on the entirety of observation that meets the set criteria. It is true that they “estimate” talent or production, but there all available data are already in use…there is no true error term possible for many player-specific metrics.

        You could apply error terms at a larger scale. For example, is player X’s BABIP significantly different than that expected by chance based on the balls he has put in play (line drive rate, GB rate, FB rate, etc.)? But you have to have a hypothesis for the error terms (or associated p-values) to have any meaning…and even then, the meaning may be unclear (in part because there are too many variable and dynamics to make predictions that are more specific than “I expect his performance to regress”, which could mean that it could improve or worsen, depending on which end of the distribution his current performance is located in.

        In any case, you appear to be holding new statistics to higher standards than those with which you are more comfortable. Based on the ineffective nature of the pitcher-win stat to measure performance, I bet that having 10 pitcher-wins right now is not significantly different from having 6….and that batting .310 at this point in the season is no different than batting .280…because there really is that much luck involved in the measurement of those statistics.

      • Roger Moore - Jun 28, 2011 at 2:23 PM

        @paperlions,

        It actually is perfectly possible to put error bars on things where you’re sampling the entire population. There are many scientific experiments that depend on exactly that kind of error estimate, and a whole set of tests designed to work with such data. For batting stats, for example, if you assume that individual plate appearances are independent experiments you can use a multinomial approximation to estimate the sampling error. Even for things that aren’t easily approximated by multinomial statistics, like situation dependent value added stats, you can use resampling methods to generate error bars.

        The bigger question is what information you’re trying to extract from the data. Error bars and p-values like drmonkeyarmy wants are most interesting if you assume that the data is being presented prospectively, so that sampling biases are important. If you’re looking retrospectively at value added, you don’t really care if the value of a player’s contributions results from skill, luck, or some combination; you only care about how his actual behavior affected his team’s chances of success. In that case, the only error bars you care about are internal to your method related to how accurately you can estimate the value of different outcomes, which are going to be relatively small.

    • drmonkeyarmy - Jun 28, 2011 at 3:02 PM

      Roger,
      But you are using the past data to predict future performance. Is that not the whole point of using sabermetrics to as a player picking tool?

      • Roger Moore - Jun 28, 2011 at 4:12 PM

        Using past data to predict future performance is one use of sabermetric analysis, but it’s far from the only one. People also use statistics retrospectively to measure player value for things like giving MVP and CYA voting, HOF ballots, Gold Gloves, and the like. For that kind of pure value analysis, things like sample bias are less important. We care only about actual value added, not whether that value added was the result of skill or luck. The only source of error in measuring the value of past performance is error in our performance value modeling- which may still be large when comparing very different players. So what kind of error bars you use will depend on how you’re planning on using the data.

      • drmonkeyarmy - Jun 28, 2011 at 4:26 PM

        I like you Roger Moore. You present your arguments in a very clear and concise fashion. Also, you clearly know what you are talking about.

  6. snowbirdgothic - Jun 28, 2011 at 10:42 AM

    Well, when he was in Vegas with his buddies Scooter, Nimrod and Sully, and they were talking about Bill’s dad’s fantasy basketball draft of Celtics bench players from the mid-50s by attaching each one to a snippet of dialog from “Swingers” – but only the extended edition from the DVD that debuted on Jimmy Kimmel Live, ’cause that’s how he rolls – Simmons decided that math was hard.

    • drmonkeyarmy - Jun 28, 2011 at 10:45 AM

      Simmons didn’t write the article.

      • snowbirdgothic - Jun 28, 2011 at 10:58 AM

        I know. But it’s his site, so he gets credit for editorial direction. And nobody would have gotten any Jonah Lehrer jokes.

      • Kevin S. - Jun 28, 2011 at 10:59 AM

        Who’s Jonah Lehrer?

    • b7p19 - Jun 28, 2011 at 11:26 AM

      Not only did Simmons not write the article, but he has written favorably about sabermetrics in baseball and basketball numerous times in the past. A good editor let’s his writers express their own opinions. Are you a communist?

      His only responsibility is to inappropriate content, not different opinions.

      • snowbirdgothic - Jun 28, 2011 at 11:41 AM

        If you’re going to be technical, Simmons didn’t climb on the stat bandwagon until 2010, at which point he embraced it noisily and started retrofitting a narrative about how he’d been a stat geek all along. That being said, he still calls statheads “dorks”, so your mileage on his conversion may vary.

        As a professional writer – and former product line editor – I’ve got a very good idea of what editorial voice is about and how it’s supposed to work, thanks.

        Thanks for the “communist” jab, by the way. Funniest thing I’ve read all morning.

      • drmonkeyarmy - Jun 28, 2011 at 11:51 AM

        Even though you may be right, don’t throw in the “I’m a professional writer so I know better” stuff. That is lame. It makes you sound as if you thing you know better solely because you are a writer. Furthermore, I’ve read a bunch of “professional writers” who don’t seem to know much about editorial voice and such.

      • Kevin S. - Jun 28, 2011 at 12:16 PM

        He also calls his friend Daryl Morey Dork Elvis, and he refers to his most committed fantasy leagues as the League of Dorks. I don’t think he uses the word as a put-down.

    • grizz2202 - Jun 28, 2011 at 12:29 PM

      Snowbird,
      Thank you so much for your reply. What you said is the exact reason I quit reading his crap six years ago. How many times can you read the same article over and over with a different date on it??

      Blah blah Vegas blah blah House blah blah Celtics (sounds of sucking ensue) blah blah movie reference blah blah I’m the most important person in the worldWAITI’MJUSTKIDDING blah blah blah.

      I guess if you’re from Boston, he’s quite dreamy. Otherwise, he’s a younger Rick Reilly. I’m sure he aspires to this, but I don’t find that a compliment.

      • rotaryfone - Jun 28, 2011 at 2:12 PM

        I’m from Boston, and I can assure you, a number of fellow New Englanders and I have grown tired of his shtick.

        His put downs of the Red Sox and baseball in general over the last few years have seemed disingenuous. He likes to talk about the length of games and refuses to acknowledge that maybe the problem isn’t baseball – maybe the problem is him. Which is fine, but don’t pretend otherwise. He also admits that stats are what soured him on the sport until recently. And yes, like many things he discovers late in the game, he writes as if he was there on Day 1. (sabermetrics, Anchorman, The Wire, etc) And his minions eat it up.

  7. vince9663 - Jun 28, 2011 at 10:43 AM

    An article about baseball, headlined with a picture of basketball, with an opening about buying a car. There’s nothing worth reading here, is there?

  8. Joe - Jun 28, 2011 at 10:48 AM

    I haven’t read this article yet, but there are two things that irk about the typical anti-sabermetric arguments, at least with regard to baseball.

    #1) The people making these arguments act like they know how front offices are using sabermetrics. They’ve heard of silly-sounding stats like VORP and make comments about what VORP might be missing. News flash – most MLB teams are using proprietary data, not just VORP, and a variety of metrics. You (writer) don’t know what they are looking at, so you’re in no position to criticize how they are using the data.

    #2) The horse-and-buggy argument. “Batting average was good enough for 100 years, it’s all you need today.” No, batting average was only good enough when it was what everybody was using. Once somebody noticed OBP and SLG drove run scoring more than BA did, then teams relying on batting average started falling behind. Just like when the internal combustion engine was invented and started moving people around faster, the horse-and-buggy fell behind.

    • spudchukar - Jun 28, 2011 at 11:02 AM

      All valid arguments but you fall prey to your own criticism. The notion that all teams only relied on BA to gauge player performances is merely anecdotal. TB (Total Bases) was and still is a very valid stat. I remember how important it was as far back as the late fifties, and recall many sportswriters’ use of it to evaluate players. Do not get me wrong, Sabermetrics has been a useful tool and a great addition to baseball, but to be fair, the notion that BA was the only metric used to measure performance is revisionist history.

      • Joe - Jun 28, 2011 at 11:11 AM

        BA was meant to be shorthand for “conventional stats.” But it usually the one cited in the horse-and-buggy argument, so it’s the one I chose to use.

    • Joe - Jun 28, 2011 at 11:04 AM

      And so, cases in point:

      “Relying on just the horsepower number will lead you to bad decisions.” That’s equating sabermetrics as relying too heavily on batting average.

      “According to one statistical analysis, the Los Angeles Lakers had four of the top five players in the series. The Miami Heat had three of the top four.” Why do you assume that the stats cited are the ones the Maverics or any other NBA team relies on?

      Also, he has at least one example of a team being penalized for ignoring Sabermetrics (Giants vis Aaron Rowand), and at least one example of a team being rewarded for ignoring conventional stats (Mavs vis JJ Barea).

      And he illogically concludes that teams ONLY rely on the numbers without regard to personalities, intangibles, etc. Of which there is very little evidence.

      Overall a train wreck of an article.

  9. terencemania - Jun 28, 2011 at 10:51 AM

    The article also seems to focus on the playoffs which, any sabrmetrics nerd knows, represents a small and irrelevant sample size.

    • drmonkeyarmy - Jun 28, 2011 at 10:54 AM

      Even in season long comparative statistics they don’t report p values or standard deviations. It is impossible to assess differences without knowing a p value.

      • Kevin S. - Jun 28, 2011 at 10:55 AM

        That’s actually not entirely true. PECOTA provides percentile projections. I wish other systems would do the same.

      • drmonkeyarmy - Jun 28, 2011 at 10:58 AM

        I need “p” values. The “p” value helps to sort out whether the difference is statistically significant or whether the differences could be due to random chance.

      • Kevin S. - Jun 28, 2011 at 11:01 AM

        I don’t know if they explicitly provide p-values and standard deviations, but if one had all the percentile projections (I think they do them on the 10s), couldn’t one reverse-engineer those numbers?

      • drmonkeyarmy - Jun 28, 2011 at 11:05 AM

        Probably, I’m not a statistician so I don’t really know. What I do know from my assessments of various drug studies is that when comparing data sets the best way to see if there is statistical difference between the sets is the “p” value. It is always reported so I would have no need to calculate it myself.

      • Roger Moore - Jun 28, 2011 at 11:32 AM

        I think that expecting p-values for everything is unrealistic. A p-value relates to a specific comparison or observation, not to something generic. So you can reasonably expect p-values that show whether a specific player’s performance is better than replacement level, but being able to compare any two arbitrary players would require somebody to do every pairwise comparison you might want to consider. Even if somebody is going to do the work to do that, it’s going to need a special tool to do the reporting.

        Even standard deviations have some assumptions baked into them. For example, a standard deviation for VORP requires not just error estimates for the player’s performance but also for replacement level at his position, his park factor, the runs per win factor for the season, and even for your run value estimator. But when turning around and comparing the player to somebody else, some of those terms may wind up being unimportant because the error will affect both players equally. The runs-per-win error will only apply when comparing players from different seasons, the park factor error will only apply when comparing players from different teams, etc.

        The whole process of reporting uncertainties is probably doable- and it could certainly be done better than it is now- but it’s harder than it looks.

      • drmonkeyarmy - Jun 28, 2011 at 11:55 AM

        Mr. Bond,
        I’m sure it would be an extremely tedious process. I just think that it would lend more credibility to the data. That is all I am saying. Honestly, I don’t think that we are saying that vastly different things.

      • paperlions - Jun 28, 2011 at 1:10 PM

        Okay, what are you going to do with the p-values? What arbitrary level are you going to set for alpha? The p-values are pretty meaningless here. WARs of 5 and 6 are not too different, and the player that put up 5 one year may very well out-perform the player that put up the 6 next year. For the one year, that doesn’t matter much….one player was worth 1 more win than the other. Obviously, a great deal of chance is involved in those calculations, and the closer they are the more likely it is that going forward the players will be of similar value…but there is no natural cut-off line…just arbitrary one you would define….giving you a p-value (which is just a likelihood of similarity score, really) would just transform the metric to another scale….and it would only hold value within the context of future predictions, not within the context of evaluating what happened, because we already know that…and that there was a 1 win difference between the players.

        What do you think you’d do with some p-values that you couldn’t already do with a basic understanding of the metrics?

  10. thefalcon123 - Jun 28, 2011 at 10:55 AM

    Here’s the thing: *Everyone* throughout the history of baseball was evaluated players and made decisions based on statistics. Sabermetrics just showed that a lot of those statistics weren’t the best and offered up some better ones. What the hell is the problem?

    And seriously, how is batting average (Times on base-walks-HBP/PA-BB-SH-HBP) considered to be simplier and better than that scary On Base Percentage (Times on base/Plate appearances)?

    • drmonkeyarmy - Jun 28, 2011 at 10:59 AM

      I think your formula for calculating batting average is a bit off.

      • Kevin S. - Jun 28, 2011 at 11:02 AM

        The formula’s fine, it’s just a deconstruction of H and AB. Deconstructing AB makes sense, not so much for H, although his point was to show how much the formula excluded.

      • drmonkeyarmy - Jun 28, 2011 at 11:18 AM

        Yeah, ok I see what he was doing there. I mistook the “-” signs for dashes. My mistake. Carry on.

      • paperlions - Jun 28, 2011 at 1:12 PM

        He forgot errors, and sacrifices should also be subtracted from the numerator (they aren’t hits)

      • Kevin S. - Jun 28, 2011 at 1:17 PM

        Actually, sacrifices aren’t under times on base.

    • churchoftheperpetuallyoutraged - Jun 28, 2011 at 11:24 AM

      Adding on to this, no one seems to have a problem with the absurd formula for QB Rating in football, but people bash FootballOutsiders.com’s DVOA (essentiallyl WAR for football).

      • Andrew - Jun 28, 2011 at 12:47 PM

        On a related note, check out PER (player efficiency rating) for basketball. When Kevin Garnett was with the Wolves, the local newspapers often cited Garnett’s PER to demonstrate how good of a player he was (he typically led or was near the top). The formula is so complex that its Wikipedia entry uses a horizontal scroll bar just to fit it all onto one page. http://en.wikipedia.org/wiki/Player_Efficiency_Rating#Calculation

  11. yankeesfanlen - Jun 28, 2011 at 10:58 AM

    I am sabermetrics-neutral.Sabermetrics can play an important role in evaluation of comparable players, and may lead to better adjustments to a player’s game.
    Some things they don’t cover:
    Physical limitations that can limit a players range
    Mental focus
    Ability to get and pay for the proper fit within the roster.
    This takes the judgment of coaches, scouts, managers, GMS and front office.

    Think of it as taking a photograph: A combination of aperature, shutter speed, lighting, ISO, lens light (speed)capability, then (the mgmt part) taking the photograph. All cameras now have an “auto” setting which will do everything but snap the shutter. But all these combinations have to be put together to obtain the best photo.

    Well, enough. Sabermatrics can make excellent composition, but it takes a good management staff to finesse a great final result.

    • Joe - Jun 28, 2011 at 11:13 AM

      Thumbs up for the photography analogy.

  12. Jonny 5 - Jun 28, 2011 at 11:13 AM

    I don’t understand the “I don’t like sabermetrics” mentality. Sure some are better than others, but it’s merely as a group “more tools to do the job”. Go ask anyone building anything if less tools are better than more tools to get the job done the best way possible. And here it’s coming into play in competitive sports. “Competitive” being the key word here. Any extra tool at your disposal, well you better use because the next guy is and he’ll get a leg up on you. For the casual fan though, I can see why it may annoy you. But to say any team is just using Sabermetrics to evaluate talent is basically saying you don’t know what the heck you’re talking about. No team would limit itself in such a way, it just won’t happen. Remember “more tools are better than less tools”. Scouts are still doing their job, but with more tools now, that’s all.

    • drmonkeyarmy - Jun 28, 2011 at 11:20 AM

      As I have said numerous times above, I don’t like the comparative sabermetrics because I don’t think they report values that signify the statistical differences between data sets.

      • Jonny 5 - Jun 28, 2011 at 11:38 AM

        I wasn’t referring to you actually. It was more directed at the generic group of people against using past results to gauge talent. I didn’t read any above posts before commenting, but I can see why you thought I was pointing you out, now that I have read above. It was coincidence. And if you read between the lines of my comment, I don’t have a problem with people not liking it anyway. I have a problem with criticizing those who are only implementing every possible tool to evaluate talent for competitive play. Their job security depends on the best result. Using every tool at your disposal is how you do your job best. That’s all.

      • Ari Collins - Jun 28, 2011 at 11:38 AM

        I’m getting the impression you might care a little about p-values. : P

        It would be nice if it were included with every statistic, but most any sabermetric writer will say that the data might not be statistically relevant yet, and there are countless articles about sample size. FG even had an article this year detailing exactly when each statistic reached reliability thresholds.

        Now, p-values would be nice, but having to figure out the reliability yourself does not make the statistic meaningless. Sabermatricians stress the importance of reliability; “Small sample size!” is practically their rallying cry. (Side note: getting a “Small Sample Size!” tattoo would be possibly the most hilariously nerdy thing you could do.) Perhaps someday we’ll get to the point where p-values are on everything, but until then, you can look up the reliability or figure it out yourself, and even if they’re not displayed, most saber types are taking them into account and even stressing their importance.

      • drmonkeyarmy - Jun 28, 2011 at 11:58 AM

        I don’t claim to read or research sabermetrics, so I am glad to hear that they are taking such things into account. It makes their work more meaningful to me. Thank you for the information Ari Collins.

      • paperlions - Jun 28, 2011 at 1:17 PM

        Monkey, what do you think you’ll do with some p-values? They are just “likelihood of dissimilarity” scores that give you an idea of how chance affected the outcomes. They won’t affect what actually happened (i.e. a player worth 1 more win was still worth one more win, even if that difference falls within the realm of chance)…all it can do is tell you if the difference is likely a reflection of true ability or not….and you can gain the same level of understanding just by browsing through some player histories to get an idea of how values vary over time for individual players. And if you want them because you think there is something magical about .05….well…then, you are better off without them.

      • drmonkeyarmy - Jun 28, 2011 at 1:28 PM

        Paperlions,
        I hear what you are saying. You kind of made my point. I would like to know the probability of whether that one more win was a result of chance or something inherent to the player.

  13. churchoftheperpetuallyoutraged - Jun 28, 2011 at 11:26 AM

    They look good on paper — so much horsepower! — but they fail to satisfy. The dashboard is ugly, the frame squeaks, and the front seats make our ass hurt.

    This is largely the fault of sabermetrics.

    How can you be a paid writer and juxtapose those two sentences together.

  14. spudchukar - Jun 28, 2011 at 11:39 AM

    For the past couple of weeks I have been formulating a new theory. I believe sabermetrics has been a factor in the lowering of run production is today’s game. Just so I am clear, I also consider that a good thing. Like I said it is a hypothesis, so criticism is both expected and warranted. One of the main tenets, and perhaps the least controversial of sabermetric stats is OBP. I know that stat has been tweaked and isn’t an advanced metric, but the emphasis here is how a metric effects the game. Teams that paid attention to OBP seemed to have advantage over others that didn’t. So teams began to both employ hitters who had good OBP numbers and taught developing players it was vital to their success. The result I believe was an increase in runs for the teams that adhered to that strategy. But baseball is competitive, and as soon as players began to take more pitches, driving up pitch counts, putting pitcher’s in the stretch, shifting infielders, starting runners etc., teams needed to counteract the strategy of more baserunners, by adjusting, primarily by demanding pitchers throw more strikes. Soon stats that showed the importance of strike one began to emerge.

    With strike one, batting averages dip to the Mendoza line, ball one makes .300 hitters out of most everyone. The recognition of the importance of getting ahead in the count, which also decreases walks significantly, was emphasized and slowly the significance gained more traction. Once the minor leagues and developing players were also instructed and indoctrinated, the acceptance and use became the new dogma. The result is fewer and fewer walks are issued, hitters that used to enjoy the advantages of being ahead in the count now find themselves behind, pitch counts are lowered, etc, and the thus fewer runs. Yes, this is all arm chair philosophy, and is meant to be simplistic. It also isn’t mind warpingly new but once stats backed up old axioms, teams began to demand pitchers practice what had always been preached, and command has now, become THE primary concern, and walks taboo.

    • Roger Moore - Jun 28, 2011 at 11:57 AM

      A more likely explanation is that advanced statistics have changed a lot since the Moneyball days. The biggest difference is that Voros McCracken’s DIPS has shown that fielding is much more important than sabermetric types believed just a decade ago. There has been a much greater emphasis on fielding, which has hurt hitting two ways. On the one hand, increased emphasis on fielding has meant that batters are facing better defensive lineups than they were before, which hurts their hitting directly. On the other hand, teams are more willing to sacrifice some offense for better defense, so the average lineup is less hitting oriented than it was. There have been some rule changes that help defense, too, like reducing the maximum allowable bat diameter from 2.75 to 2.61 inches.

    • kingkillerstudios - Jun 29, 2011 at 10:58 PM

      “I believe sabermetrics has been a factor in the lowering of run production is today’s game.”

      Your argument is interesting…but if by “today’s game,” you mean “since 2005,” I think there’s something else lowering the run production. And it ain’t math.

  15. kopy - Jun 28, 2011 at 11:42 AM

    53 comments and not one man told Craig to get off his lawn? What the hell?!

  16. sdelmonte - Jun 28, 2011 at 12:01 PM

    Grantland hasn’t impressed me so far. It seems to be like a lot of those New Yorker articles no one gets around to reading.

    • drmonkeyarmy - Jun 28, 2011 at 12:09 PM

      Yeah, I’m not a big fan of Simmons. His articles are often tedious. Furthermore, I don’t like how is mixes popular culture into his stuff. It seems like he is trying to hard to be Hunter Thompson sometimes.

    • kopy - Jun 28, 2011 at 12:23 PM

      I haven’t been all that impressed either. Every story on Grantland has been written like the writer knows it’s going to be their only story (or some more crap you only care about if you’re from Boston), and they’ve published it accordingly.

      We’ve seen stories about the greatest sports moment ever personally witnessed ever, the funniest sports moment ever personally witnessed ever, the personal reason for becoming a sports writer, etc., and of course this: the “I have one shot at 2000 words to nationally grind this ax I’ve been holding for years and here it is”.

    • Kevin S. - Jun 28, 2011 at 12:26 PM

      Until Grantland gets Joe Posnanski writing for it, it’s destiny will remain unfulfilled.

  17. churchoftheperpetuallyoutraged - Jun 28, 2011 at 12:31 PM

    For drmonkeyarmy, here’s a better way to write what I was trying to explain above [courtesy of Tom Tango]:

    It’s not a drawback! What sabermetrics does is explain the numbers. Give the saberists the numbers, and he’ll tell you what it means. A saberist will NOT tell you anything else. What the saberist is going to do is tell you the LIMITS of numbers, of how far you can take the numbers. AFTER that, after the numbers have been parsed and exploited, THEN that’s where your scouts and your guts come in. And those are IMPORTANT activities that take place.

    See, after I tell you what the numbers mean, and how much uncertainty there is (say, I’ll tell you that Justin Verlander (entering 2011) is a .575 (+/- .050) winning % pitcher if given average support), then the scout can come in and say “A pitcher with these kinds of tools is way better than a .575 pitcher.” Great, that’s perfect. I ACCEPT that.

    If he follows that up with: “Just look at how many strikeouts he has (through 2010).” Then I have to stop him. I’ve already parsed the numbers. I know what the numbers mean.

    What sabermetrics does is allow the scout to focus on things OTHER than the numbers. On the pitchers tools and his guts. Let the scout quantify that into a 20-80 scale. I already quantified his actual past performance.

    http://www.insidethebook.com/ee/index.php/site/article/jonah_lehrer_this_is_largely_the_fault_of_sabermetrics/

  18. nicosamuelson2 - Jun 28, 2011 at 1:00 PM

    “Grantland?” More like Suckland.

  19. ditto65 - Jun 28, 2011 at 2:23 PM

    All this talk of p-value is making me sick. If you like sabermetrics, fine. If you don’t, good for you. There are so many other things we should argue about. Like, is Beep-Beep’s intangibles worth the low production and limited range? And why is A-Rod a jerk?

    • drmonkeyarmy - Jun 28, 2011 at 3:05 PM

      If you don’t like the discussion then don’t read the comments. Pretty simple. A-Rod is a jerk because he is and has always been allowed to be. Jeter’s intangibles are worth his dip in production….and stop calling him Beep Beep.

      • halladaysbicepts - Jun 28, 2011 at 3:10 PM

        Who’s Beep Beep?

      • drmonkeyarmy - Jun 28, 2011 at 3:16 PM

        This person calls Derek Jeter “Beep Beep”.

      • halladaysbicepts - Jun 28, 2011 at 3:24 PM

        If he’s calling Derek Jeter “Beep Beep”, maybe I should start calling Jimmy Rollins “Honk Honk.”

      • ditto65 - Jun 28, 2011 at 3:58 PM

        How did you know I was referring to Jeter? And what is the p-value of his intangibles?

      • drmonkeyarmy - Jun 28, 2011 at 4:24 PM

        Because I’ve read where you have done so before. Was the comment “what is the p value of his intangibles” meant to be funny?

      • bigxrob - Jun 28, 2011 at 4:44 PM

        It’s making him sick, yet he returned for more, this guy sounds like a troll

    • ditto65 - Jun 28, 2011 at 4:48 PM

      Of course the p value of Beep Beep’s intangibles is meant as a joke. And for the record, this is the first thread where I refer to Jeter as Beep Beep. I believe we have Yankeesfanlen to thank for that one.

      • drmonkeyarmy - Jun 28, 2011 at 4:53 PM

        Ok. Fair enough. Sorry then. Don’t mock the “p” value though. It is a valuable tool in statistical analyses.

      • yankeesfanlen - Jun 28, 2011 at 5:11 PM

        Who? What? Leave ARod Alone!

Leave Comment

You must be logged in to leave a comment. Not a member? Register now!

Featured video

Orioles turn AL East on its head
Top 10 MLB Player Searches
  1. G. Stanton (3884)
  2. R. Castillo (3080)
  3. A. Rizzo (2470)
  4. A. Pujols (2153)
  5. H. Ryu (2113)
  1. E. Gattis (2090)
  2. J. Hamilton (1923)
  3. C. Davis (1920)
  4. B. Belt (1903)
  5. C. Young (1805)