Why We Undervalue Defense

During the latter part of this week, Dave has discussed defensive metrics and how they are, or should be, treated moreso as inferential statistics than descriptive; that the results should be treated as data points as opposed to concrete and isolated performance indicators. As odd as it sounds, defense is still undervalued today. And when I say undervalued, imagine me putting a ton of emphasis to really hammer home the point that defense just is not considered as important as offense. If I were to compare Manny Ramirez and Carl Crawford (as I will below), those who do not really consider defense to be valuable with scoff at the results.

Here are Crawford’s and Ramirez’s runs projections for 2009:

Manny Ramirez: +30 batting, -7.5 adjustment, +20 replacement, -15 defense
Carl Crawford: + 3 batting, -7.5 adjustment, +20 replacement, +10 defense

Put together, this puts Ramirez at +27.5 runs and Crawford at +25.5 runs. Converted to wins, that is a mere 2.75 to 2.55 advantage for ManRam, and that isn’t even including baserunning metrics such as BP’s equivalent baserunning runs. Now, for a second, imagine if I were to go on air at ESPN and discuss how Crawford in 2009 is almost just as valuable, all told, as Ramirez: can you imagine the type of criticism I would receive? In fact, I expect to get some here as well, even though our fanbase tends to value defense a bit more.

Even in my own head it sounds odd that Manny’s production is only two runs higher than Crawford. I hate resorting to arguments centering around how the media shapes our views of events, but this is definitely a big part of why fans focus moreso on offense, or consider offense as worth much more than defense. Only when a player is considered a defensive wiz is the run prevention with the glove deemed important, and in cases like that, people almost go too far, saying that offense doesn’t matter at all because of how solid the player’s glove looks.

Shows like SportsCenter only show the spectacular plays, and this is why our views on the defensive resumes of certain players (See: Derek Jeter) are warped. He has made some fantastic plays in his career, and we have seen them over and over again, but he is not a good fielding shortstop in any sense of the word. This doesn’t show up in the form of errors or fielding percentage, though. Even though innate flaws exist in new age defensive metrics and our perceptions of their usage, we can all agree that fielding percentage is awful. Jeter won’t make errors on certain plays, primarily because his range prevents him from making these plays.

Watching a game, we tend to evaluate defense solely on whether or not the play was made, not if the player had enough range to reach said ball. If Chase Utley bobbles a ball at second base, even though he showed more range to get to the ball than any other second baseman in the league, many fans would think of it as a poor play. Was it? I personally feel it would be a great sign because he was able to reach the ball, which boils down to a big point: fans are spoiled, and defense is honestly one aspect of a game where watching a game hurts our opinions and evaluations.

Shane Victorino has terrific range in centerfield, but because he scurries to a Flyball A and Carlos Beltran is able to glide to Flyball A, many people would consider Shane the better fielder. Beltran made the play look easy, and therefore it becomes expected. He is expected to make that play and those viewing the games get spoiled into considering this normal, when in fact it is insanely impressive fielding. When we see enough of something, it gradually establishes the norm in our minds, and everything is treated relative to that norm as opposed to the average for the league. This is why many fans will consider an average or slightly above average defender miraculous with the glove simply because his predecessor had awful range and displayed poor defensive abilities.

Even with all of the defensive metrics available today, this is an area that still isn’t concrete from an evaluative standpoint. This fact makes it tough to justify defensive importance when statheads like us are under intense scrutiny to mathematically and logically prove our points. I’m not sure how fans, myself included, can get past being spoiled or falling into a comfort zone with what we watch as opposed to how the average performer would perform, but until we reach that point, comparisons like Ramirez-Crawford from above will continue to look ridiculous. After all, how could Crawford’s defense and baserunning come even close to what Manny can do with the bat? Perhaps if a defensive metric comes out that leverages the importance level of the situation we could begin to show this, or one that, with very educated estimates, shows what hit would have been recorded had the ball not been fielded (as in, factoring in the run values of what would have occurred had the play not been fielded), but this is still a tough analytical facet to sell.

Defense is very important, but how important is still uncertain in the eyes of many fans, primarily because of how we value errors, and how we get spoiled watching certain players.




Print This Post



Eric is an accountant and statistical analyst from Philadelphia. He also covers the Phillies at Phillies Nation and can be found here on Twitter.


33 Responses to “Why We Undervalue Defense”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Jim says:

    Well you make some good points….I guess the main reason offense gets the spotlight is because it is harder to quantify for the average fan….a homerun is exciting but it Jeter misses a ball because he has bad range, nobody cares.

    Vote -1 Vote +1

  2. Jim says:

    Even though they should.

    Vote -1 Vote +1

  3. Aaron says:

    Plus most of us have been viewing baseball through the prism of fantasy baseball which completely leaves out the defensive equation. I might even say that has a bigger effect than seeing highlights on Sportscenter because you’re training your mind to look at the game in one way all the time.

    Vote -1 Vote +1

  4. philosofool says:

    I think that another reason is that it’s simply difficult to see how many runs an out is worth. When I first started reading Dave’s stuff here and at USS Mariner about a year ago, I was incredulous. Surely a single out just isn’t worth that much, I told myself.

    Okay, let’s take a ground ball with no one on base and no outs. There are only too outcomes that we need to consider: we add a man to first or we add an out.

    The average number of runs scored with a this state of play is .555. The average number of runs scored with no one on and 1 out is .297. The average number of runs scored with a man on an no outs is .953. (You’ll find all this on tangotiger’s website; google “win expectacy matrix”.)

    What does that mean? It means that the defender saves .258 runs (.555 – .297) if he makes the play and “saves” -.398 (creates .398 runs for the opposition) if he doesn’t. That’s a difference of .656 runs. So a player that fields 15 balls in exactly this sort of situation creates 9.84 runs. That’s equal to the number of runs a .355 wOBA player with 600PA produces over the average hitter. There were just 65 hitters in the majors with a wOBA over .355 last season.

    It’s just hard to realize how much one out is really worth.

    Vote -1 Vote +1

  5. RollingWave says:

    that.. and because offense seems to be a much more stable skill than defense.

    if a guy saves 20+ run a year and hits ok, he’s valuble, but saving that many run has a lot of luck envolved, a lot more than say… BABIP or whatever other factor that may cause offesnive vairation. the same guy might save 10 run next year, which would make him just half decent, and a ok hitter is much more likely to you know… turn into a sucky hitter.

    various metrics differ, but even looking at just a few shows how almost every player show some pretty serious variation on his defensive skills from year to year.

    If we use PMR this year for example, and everyone’s favorite sabre defense scape goat Derek Jeter, PMR had him at 1% under expected ratio. this year, which combined with his declining but still above average hitting made him fairly valuable , while the year before that he hit better but was his usual terrible self with the glove. but in either case, Jeter was still a valuable player, where as we can look at the guys who dominated PMR in 07 SS was..

    Tulowitzki
    Pena
    Furcal
    McDonald
    Bartlett

    all 5 sucked pretty badly this year (well actually Bartlett remained pretty much the same), granted it’s not fair in Tulo and Furcal’s case. but Pena and McDonald illustrate why a great fielding poor hitting shortstop is still nowhere comparable to a great hitting poor fielding counterpart in the longer run. (according to PMR, both were WORSE than Jeter with the glove in 08)

    Vote -1 Vote +1

  6. don says:

    Well, you might be cheating yourself looking at Manny vs. Crawford because you’re thinking about .332/.430/.601 Manny 2008, whereas the projection (if you’re using Marcal) is .299/.396/.527. Solid to be sure, but hardly in the same league.

    I think the fact that 48 HRs and 145 RBIs (or whatever) end up on the stat sheet and the back of the baseball card have more influence on what makes people think a player is great than the more predictive statistics. Moreso than fantasy baseball – which may be widespread in the stat geek community but isn’t as ubiquitous as fantasy football.

    Vote -1 Vote +1

  7. Teej says:

    Plus most of us have been viewing baseball through the prism of fantasy baseball which completely leaves out the defensive equation.

    I love fantasy baseball, but this is spot-on.

    I think I — and a lot of other smart baseball fans such as those who read sites like this — have learned enough about what’s valuable in real baseball to be able to separate fantasy from reality, but I’m guessing the great majority of casual baseball fans who play fantasy have not. For a lot of people, fantasy is the only thing that keeps them interested in baseball during the grueling summer months, so they’re probably checking box scores more than defensive metrics.

    Fantasy baseball is good for MLB, but it’s probably not a great thing for the casual fan’s ability to evaluate players. But it appears to be here to stay.

    Vote -1 Vote +1

  8. lookatthosetwins says:

    So why hasn’t a better fantasy league appeared yet? There are so many of us out there that understand this. I would love to play if I felt like picking a better player would give me a better chance of winning. Dewan’s +/- can be found regularly during the year, correct? If so, it wouldn’t be that hard to convert these things to wins over replacement. When someone does that (and doesn’t charge an arm and a leg) I’ll start playing.

    Vote -1 Vote +1

  9. Teej says:

    Dewan’s +/- is available with a subscription to Bill James’ website. So you’re looking at a licensing issue. Plus the whole idea of one-year defensive stats being so raw that the flukiness could drive people mad. And quantifying +/- in real time is probably impossible.

    Vote -1 Vote +1

  10. Manhasset Paulie says:

    A quick thought which is worth exploring more.

    A run of defense SHOULD be valued less then a run of offense.

    The logic is simple.
    – Year over year there is higher variance (risk) in defensive performance then offensive performance.
    – A rational beings is risk-adverse.
    – Thus in an efficient market Manny lower variance (risk) +27.5 runs should be valued disproportionally higher then Crawford high variance defense driven +25.5 runs.

    Vote -1 Vote +1

  11. CJ says:

    I have never thought of defensive output as more variable than offensive output. You will have players with career years on defense, just like you have players with career years on offense. Conversely players can have down years on defense, just like it happens on offense. In fact, my general intuitive response is that defense has less variance, particularly during a player’s peak years. I do think that their may be a sharper decline curve for defense after a player’s peak years. But most of the premier defensive players are incredibly consistent during their peak years (look at Adam Everett’s age 25-28 defensive metrics, for instance), as are most of the premier offensive players during their peak years.

    I agree with most of the reasons that have been expressed about why defensive stats aren’t taken as seriously as offensive stats by fans. I would add another factor: runs saved are a joint output of pitching and defense. Good or bad defense shows up as an inseparable part of mainstream pitching stats, like ERA. Fans don’t realize that a lot of the credit or disparagement they give pitchers is really based on the defense of other players on the field.

    I think another factor is that many people refuse to believe advanced defensive metrics. They don’t understand how they work, and so they have false perceptions about why they are (in their opinion) worthless. Fans understand what fielding percent is, but UZR, PMR, or RZR…no. It doesn’t help that some players rate very differently on certain advanced defensive metrics, which only solidifies the opinions of those fans about the validity of the methodologies. I think it would be helpful if the creators of the various defensive measurement systems attempt to understand the reasons for the differences (and I realize part of the difference is just the data source).

    As for fantasy baseball with defensive metrics, I like the idea. RZR appears to be updated daily at HT.

    Vote -1 Vote +1

  12. The Ancient Mariner says:

    I don’t think fielding evaluation is ever going to be “concrete,” really; I don’t think it can be. Runs are scored via a linear series of clearly-definable events, so that can be measured more or less concretely. We can evaluate pitchers based on strikeouts and walks because those are also clearly definable events, with only two players involved. Whether a ball that lands in play becomes a hit or not, however, is not in the least linear; it depends on variables such as hit type, location, positioning of fielders, and even weather in addition to the quality of those fielders. As well, since in trying to gauge the value of players, we’re trying to measure them against average and replacement level, we’re necessarily comparing the play of those fielders to what other fielders would have done had they been there instead–which means we have, of necessity, gotten into might-have-beens. This is never going to be anything more than inferential; the best we can do is to sharpen the inferences.

    Vote -1 Vote +1

  13. philosofool says:

    Rational beings are not risk averse. Risk averse people are subject to reverse dutch books, which means that there are bets that guarantee that they will make money that they nevertheless won’t take.

    Here’s what risk aversion gets you in the context of baseball: this guy has a 50% chance of adding to wins and 50% chance of costing me one win. But costing one win is twice as bad as adding a win is good. So I won’t take him. It’s obviously stupid reasoning. (The expectation in the above scenario is adding half a win, so you should do it.)

    Rational beings consider risk exactly the way they consider reward: according to the expected utility calculations regarding each.

    Vote -1 Vote +1

  14. robneyer says:

    I don’t understand how these numbers work:

    Here are Crawford’s and Ramirez’s runs projections for 2009:

    Manny Ramirez: +30 batting, -7.5 adjustment, +20 replacement, -15 defense
    Carl Crawford: + 3 batting, -7.5 adjustment, +20 replacement, +10 defense

    Put together, this puts Ramirez at +27.5 runs and Crawford at +25.5 runs.

    Vote -1 Vote +1

  15. Manhasset Paulie says:

    Because of diminishing marginal utility, risk causes expected utility to decrease. Maybe utility is not marginally diminishing in baseball.

    I suppose you are arguing that the utility function in baseball is not diminishing but linear because a “win is a win.” If we want to nick pick (which we should we could say it is spiked in the 86-93 win range. As most teams are below this range should we expect most baseball GM’s risk-seeking?

    This would be easy enough to test. How do contracts for suspended players (Mota/Cameron) compare with injury prone players (Pedro/Alou)?

    Vote -1 Vote +1

  16. Eric Seidman says:

    Rob, taking their projections for 2009, ((wOBA – lgwOBA)/1.15) * PA to get batting runs. Then take the + – projections for 2009, and multiply the plays by 0.8 to convert to runs. For leftfielders, the adjustment is -7.5 runs, and then we add +20 runs so it is all above replacement level, not average. I have Crawford at about +11 or +12 on defense next year, Manny about -17 or -18, which turns into -15 runs and +10 runs.

    Vote -1 Vote +1

  17. robneyer says:

    Thanks, Eric. I’m still hazy on the -7.5, though; that’s a defensive adjustment? Hitting? (Sorry I didn’t just e-mail you, but for some reason gmail isn’t helping me retrieve your address.)

    Vote -1 Vote +1

  18. kensai says:

    http://www.insidethebook.com/ee/index.php/site/article/complete_war_2008/

    The positional adjustment is there. It’s in wins though, so multiply by 10 to get runs.

    Vote -1 Vote +1

  19. Eric Seidman says:

    Rob, yeah the link above is good, but the adjustments are done primarily to level the replacement field of play. Here is another link that gives good insight:

    http://www.hardballtimes.com/main/article/replacement-level-article/

    For Catchers, add 12.5 runs, for SS, add 7.5, for 2B/3B/CF, add 2.5 runs. For LF/RF subtract 7.5 runs, for 1B subtract 12.5 runs.

    Vote -1 Vote +1

  20. snowshoe says:

    Manny Ramirez is an interesting choice of example to bring up regarding the evaluation of defense quantitatively. There’s been a great deal of controversy over how to measure defense for left fielders in Fenway given the skewed results seen in that park, particularly by zone rating and its variants.

    In this post, Manny is described as a -15 defensive player which is awful.

    But PMR had Manny as only slightly below average at -4.3 runs for 2008.

    Chris Dial had Manny as -10.8 in the AL while playing for Boston and +2.2 while playing in the NL for the Dodgers.

    +/- had Manny as -13 in 2008 and -38 in 2007.

    How bad a defensive player was Manny in 2008? What do the metrics tell us about his innate defensive ability?

    It’s a real challenge to try to ascertain what Manny’s true defensive value and ability is. He’s clearly not a good defensive player. But quantifying how bad he is seems very problematic. In turn, it’s difficult to equate his value with that of Crawford’s.

    Part of why defense is undervalued is due to this variation in these kinds of results.

    Vote -1 Vote +1

  21. Chris says:

    I believe that smaller stadiums make slower outfielders look worse than they actually are and bigger stadiums make fast outfielders look better than they actually are. Fenway was probably the worst stadium in terms of defense for a guy like Manny. The jutting stands down the line, the monster and the weird angles make his lack of speed much more apparent. If you put a guy like Crawford in Fenway’s LF, he would do much better than Manny, not because he’s necessarily a better fielder, but he has the speed to get to more of those balls that Manny couldn’t. In LA, with more of a standard outfield layout, Manny is able to get to more balls because he simply has more time and more room to do so, which makes up for the lack of speed.

    Also, I wanted to comment on the 1B adjustment. I watched a lot of Morneau during his career and he doesn’t necessarily save a lot of runs fielding batted balls, but saves a TON of runs bailing his fielders out on bad throws. He’s the best “picker” I’ve seen in the AL. I wonder if that is taken into consideration when fielding stats are formulated?

    Vote -1 Vote +1

  22. philosofool says:

    Eric,

    Once we factor in defense, the replacement level stuff makes a lot less sense to me. If we’re already considering a player against the collection of people that can play defense at their position, why are we then adding a replacement level factor? I mean, you can play anyone you want a shortstop; to know how wise that is, just look at his run prevention while playing there. Why do we still need a positional adjustment when we’re looking at players that play the same position?

    Vote -1 Vote +1

  23. Fritts says:

    The +/- system compares players to others at the same position. The positional adjustment used is because a +5 LF isn’t as valuable as a +5 SS.

    Vote -1 Vote +1

  24. Cheers says:

    Eric, I agree with what you’re saying about defense being undervalued. However, and maybe this is a small point, I don’t see why you are saying that defensive metrics like ZR, UZR, PMR, etc. are not “concrete and isolated performance indicators.” They are, as Snowshoe described in his comments on Dave’s post. They provide an indication of the result of every fielding play during a season — and as such are no different from batting or pitching counting stats, or rate stats for that matter.

    What you choose to infer from them is up to you — which again, makes them no different than batting or pitching metrics. They have the same sample-size concerns: Casey Blake’s 1.000 ZR at shortstop this year tells us as much about his true fielding skills as Felix Hernandez’s 5.000 OPS tells us about his true batting skills. They have the same capacity to produce single-season outliers: Utley’s +47 PMR season looks just as strange and unlikely as Brady Anderson’s 50 HR season, yet no one denies Anderson actually DID hit 50 bombs.

    Utley is not going to put up a +47 every year, but there’s no denying that he DID put up a +47 in 2008 — and it’s up to us to analyze what that means and put it in context. I think that’s pretty much the same point you’re trying to make when you talk about using defensive metrics as “data points”. And I agree — but that applies to every batting or pitching metric as well.

    Vote -1 Vote +1

  25. The Ancient Mariner says:

    They provide “an indication of the result,” yes; but the thing about fielding is that the performance is more (or less) than the result.

    Vote -1 Vote +1

  26. Sky says:

    Paulie, in fact, defensive runs are MORE valuable than offensive runs, given that the team is better than .500. Take a team that scores 100 more runs than it allows. Would you rather scor 900 and allow 800 or score 700 and allow 600? Pythag would say:

    High runs: .554, 89.7 wins/162
    Low runs: .570, 92.3 wins/162

    (I’m using straight Pythag with a 1.83 exponent. Maybe something like Pythagenport would close the gap a bit.)

    Vote -1 Vote +1

  27. Sky says:

    philosophool – If two players have different playing time, they deserve a different pro-rated piece of the position adjustment (same for the replacement-level adjustment). But you’re right, comparing players at the same position otherwise doesn’t require the position adjustment. It’s just that the resulting number is unitless — only the relative differences will mean anything. Which is fine for some uses.

    Vote -1 Vote +1

  28. CJ says:

    Re: Effect of Fenway Park on LFers.

    Left field at Fenway, and to a lessor degree, Minute Maid Park, produces distorted defensive stats. At least one of the defensive play by play data bases fails to accurately identify balls which are playable off the LF wall at those parks. Both MMP and Fenway have short LF walls, and the data doesn’t tell you what balls off the wall are catchable. I have seen analyses indicating that you should subtract as many as 10 runs off Manny’s Red Sox results and 4 or 5 runs off the Astros’ LFer’s results.

    PMR controls for ballpark in the sense that players are compared only to opposing fielders in their home ballpark. Both Ramirez and Carlos Lee rate significantly better on PMR than most other defensive metrics.

    Vote -1 Vote +1

  29. Clark says:

    Sky, in your example, the “low runs” team has the advantage of having a higher ratio of runs scored to runs allowed than the “high runs” team. A fairer question would be “would you rather add 100 runs or save 100 runs?”

    Starting with a team that scores 750 runs and allows 750 and using the Pythagenpat formula (x=((RS+RA)/G)^.285), adding 100 runs of offense would give them 90.7 wins—one win for every 10.3 runs added—while subtracting 100 runs from their defense would give them 91.7 wins—one win for every 9.3 runs saved. This is consistent with the oft-cited figure that a run saved is 10% more valuable than a run scored.

    Vote -1 Vote +1

  30. Sky says:

    Clark, your fairer question has the same “flaw” as my example, fyi. You’re keeping the run differential constant, but changing the ratio. You’re just comparing 800/700 to 700/600 instead of 900/800 to 700/600.

    At any given run differential adding a minimal number of runs to RS or subtracting a minimal number from runs allowed will have the same value. But if you get to keep the same run differential, you’d rather have a low-scoring team, meaning teams should be on the lookout for good pitching and good fielders, meaning saving runs are more valuable.

    Vote -1 Vote +1

  31. Clark says:

    No, not really. I used a clearly defined starting point (750 runs) to measure the value of a run scored relative to a run prevented. You used two different starting points which overestimated the value of a run prevented. Had you used 900/800 and 800/700 you’d have been fine, but changing the run environment in the second example (700/600) skewed the results and made a run prevented look 20% more valuable relative to a run scored than it really is.

    That said, I don’t know how significant any slight differences really are considering that over the past two seasons, only half (31 of 60) of teams finished within three wins of their expectation and twelve finished at least five games away. In 2007, there were three teams—the A’s, Mariners, and Diamondbacks—who “should” have won 79 games; they actually won 76, 88, and 90. When a Diamondbacks team that was outscored by twenty runs wins as many games as a Rockies team that outscored its opponents by triple digits, you can pretty much throw all expectations out the window.

    Vote -1 Vote +1

  32. doug says:

    are you guys aware that there IS an on-line fantasy game that values defense VERY highly?
    imaginesports.com

    Vote -1 Vote +1

  33. Reggie says:

    I doubt this will be answered, but does UZR take shifts into account in any way and if so how?

    Milwaukee had the poor range vs error prone defender at 2B in 2008. Many people likes the poor range guy better because he made fewer mistakes.

    Vote -1 Vote +1

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>