FanGraphs Baseball

Comments

RSS feed for comments on this post.

  1. You pretty much say no value should be placed on the results so why hitch yourself to the SI waggon? First impressions and all that

    Comment by pssguy — April 24, 2012 @ 12:24 pm

  2. as far as pitching with the bases empty versus pitching with runners on base, aren’t some pitchers stronger in the windup versus the stretch? if the royals have a few of those, this could explain the disparity- to some degree.

    Comment by steven — April 24, 2012 @ 12:27 pm

  3. Excellent post.

    Comment by Baltar — April 24, 2012 @ 12:28 pm

  4. I’m sorry, but what on earth are you talking about? According to wRC+ and FIP, as taken from Fangraphs, the Royals should have scored 4.21 RPG and allowed 4.54 RPG. Hitting better in the clutch would have improved their runs scored, but they don’t get to an above-average rank without their +8 team UZR.

    Comment by Colin Wyers — April 24, 2012 @ 12:30 pm

  5. #6org

    Comment by sam — April 24, 2012 @ 12:32 pm

  6. This is just the worst. It’s sad, really.

    Comment by Billy — April 24, 2012 @ 12:33 pm

  7. smh

    Comment by sprot — April 24, 2012 @ 12:34 pm

  8. So if the Royals got their “situational BABIP” up they’d be a .500 team instead of 3-13 so naturally that means they’re #7 in all of MLB.

    Comment by Mr. Deez — April 24, 2012 @ 12:35 pm

  9. #7org

    Comment by John — April 24, 2012 @ 12:35 pm

  10. “And to me, that’s the entire point of advanced metrics – to shine a light on the non-obvious story, and help us understand the why behind the result.”

    Except when you subscribe to UZR, which is not a solid metric.

    Comment by Heisenberg — April 24, 2012 @ 12:35 pm

  11. Nerd.

    Comment by asdfasdf — April 24, 2012 @ 12:35 pm

  12. But they do have that +8 team UZR. You can’t leave that out.

    Comment by geo — April 24, 2012 @ 12:36 pm

  13. In other words, if the Royals had hit to their wRC+ and pitched to their FIP, they’d have a Pythagorean win expectation of .465. Or, sure, seventh best team in baseball.

    Comment by Colin Wyers — April 24, 2012 @ 12:37 pm

  14. Dave Cameron did, when he explained how the Royals have such a high team WAR. Instead he talked about clutch hitting.

    Comment by Colin Wyers — April 24, 2012 @ 12:38 pm

  15. I read a lot of advanced baseball articles and you my friend are the best around.

    Comment by LIz Phair — April 24, 2012 @ 12:38 pm

  16. You had the opportunity to go “yeah, this is a blip caused by small sample size UZR but in the long run this is a decent method”, but instead you decided to double down on the results and imply that you’re smarter than the people who disagree with you.

    Comment by regfairfield — April 24, 2012 @ 12:42 pm

  17. You mean your comment? Because I was thinking exactly the same thing.

    Comment by chuckb — April 24, 2012 @ 12:45 pm

  18. Interesting stuff, Dave. Thanks for posting.

    It goes without saying that we shouldn’t draw too many conclusions from a sample consisting of 10% of the baseball season.

    Comment by chuckb — April 24, 2012 @ 12:47 pm

  19. This is where the numbers can only take you so far.

    Anyone who has watched the Royals all year knows that they are willingly and serially sacrificing runs on the basepaths. Why is there no mention of that here?

    Hitting stats tell you they should be scoring runs. That they are not is far from just the lack of “clutch” and luck that Cameron suggests.

    Comment by Blob — April 24, 2012 @ 12:48 pm

  20. YEP

    Comment by sprot — April 24, 2012 @ 12:59 pm

  21. So should we expect there to be a win streak/ course correction (trending upward) in the near future for the Royals or is their wins/losses based solely on clutch performance – which has no predictive value?

    Comment by Ignorant Tool — April 24, 2012 @ 1:06 pm

  22. Except that is what SI suggests. It seems ridiculous to me.

    SI say

    <>

    Comment by RichW — April 24, 2012 @ 1:39 pm

  23. Except that is what SI suggests. It seems ridiculous to me.

    SI say

    By utilizing WAR, we can better identify which teams are actually playing well and will likely sustain their success going forward.

    Comment by RichW — April 24, 2012 @ 1:41 pm

  24. Not sure how anyone thought the Royals would be a good club this year when looking at their opening day rotation.

    Comment by M W — April 24, 2012 @ 1:45 pm

  25. Everybody in the rotation has been, overall, about the same as expected or better. Chen has been good. Hochevar has been mostly good. Sanchez has been Sanchez. Mendoza has been Mendoza (PCL ERA leader!!), and Duffy has been awesome. If I had known that would happen before the season began, I would have put them down for 82 wins.

    Comment by hawkinscm — April 24, 2012 @ 2:04 pm

  26. This is #6org all over again.

    Comment by Dr.Rockso — April 24, 2012 @ 2:19 pm

  27. As it stands, the Royals have scored 57 runs and allowed 81. With a more normal distribution on timing of hits, though, it’d be pretty close to 70 for both RS and RA. And, as everyone who has been beaten to death with pythagorean expectation over the past 20 years knows, a team’s runs scored and runs allowed are a better evaluator of how a team has played than simple wins and losses. Pythag suggests that the 57/81 split in their RS/RA means that the Royals have played more like a .313 team than a .188 team, and their underlying components of run scoring and run prevention suggest that they’ve played more like a .500 team than a .313 team.

    Isn’t it equally likely that the model is broken? While it’s been some time since the Angels outperformed their pythagorean totals, they managed to do so consistently for several years before coming back to earth. Maybe what the Royals are telling us is that these models don’t work well for small sample sizes, and cases where the hits don’t come in key situations.

    Comment by Rob McMillin — April 24, 2012 @ 2:22 pm

  28. Hope Rany reads this before he finds a high ledge.

    Comment by Wobatus — April 24, 2012 @ 2:32 pm

  29. It seems that all you are doing is replicating the traditional rankings and text power rankings, except your input is not W-L record but WAR. I guess this could be slightly more instructive, but the error bars on WAR have to be high enough that its going to take half a season before what you are producing is meaningful in any real way. You acknowledge this pretty explicitly when you say that the Royals aren’t actually the seventh best team in baseball. At least the traditional power rankings usually have common sense tilting the scales.

    If you are trying to tell people how good each team in the majors is, why not just regress heavily and weigh ZiPS projections for the first month or two of the season until there’s a decent sample size of games accumulated; it wouldn’t explicitly reflect how good the teams have been during the season, but it would much more accurately reflect how good the teams in baseball are, which presumably is the point of a power ranking.

    Comment by Anders — April 24, 2012 @ 2:34 pm

  30. The commenter who discussed baserunning is spot on. The Royals have had no problem creating run scoring opportunities, but have gotten thrown out with kamikaze baserunning all year. Those average with runners in scoring position numbers also don’t count the number of times that they have sacrificed with a runner on second to get him to third (this happened TWICE last night). It doesn’t matter if the team is built to play optimally based on stats if the manager refuses to follow sabermetric strategy.

    Comment by Roombaugh — April 24, 2012 @ 2:48 pm

  31. I know you like numbers, Dave, but to me it’s not so difficult to imagine a young team putting extra pressure on itself to perform. Double-digit losing streaks and players in slumps, especially young players, have their way of putting extra pressure on athletes to perform. It’s magnified when pitchers put runners on bases (“here we go again”) or when a hitter tries to do too much at the plate because he knows his team needs it. Even a guy like Pujols isn’t immune to this.

    It’s a small sample. The numbers will be more evenly distributed after they win a few. But it’s also not rocket science. Show me any up and coming young team with an 11 game losing streak and I’d bet you’ll see the same uneven distributions…unless the team was projected to really suck.

    It’s not a tabula rasa for the hitters…if it were, you’d see much more consistency…there’d be no reason to focus on anything but the mechanics and approach to hitting. Those things get all screwed up in slumps. The fact that it’s reflected in the numbers isn’t a surprise.

    Nice work though!

    Comment by Cosmicsniper — April 24, 2012 @ 2:54 pm

  32. Cluster Luck plays havoc at small samples on uber-derivative functions like W-L and ERA, so it’s useful to examine the component parts of such functions to ferret out the Cluster Luck or other anomalies (which you would expect to smooth out with more inputs).

    For that reason, most GMs probably ignore WAR, RC/G, ERA, W-L, ERA+ etc, because the GM is not ranking an entire player pool, the GM’s first filters (need, budget, health, availability) shrink his player pool to a tiny group of players. Ranking a tiny group is subject to Cluster Luck and small sample anomalies, so the GM focuses on the base components: bb/k, bb%, K%, etc and scouting. Sabermetrics provides quantum metrics for the front office to supplement its scouts, and broad derivative models for the historians, fans and fantasy players to sift large helpings of disparate data.

    I think.

    Comment by William Strunk Jr. — April 24, 2012 @ 2:55 pm

  33. Using small sample WAR (which uses ridiculously bad small sample fielding as a component) to make any kind of ranking list is assinine

    According to this website, the 2 most valuable defensive players by far thus far have been Alex Gordon and Alfonso Soriano. Gordon has been so dramtic in his fielding that despite a 256 wOBA he is positive in the WAR column. And despite Soriano’s 176 wOBA and 420 OPS he does not have a negative WAR.

    Its ridiculous plain and simple

    Comment by mister_rob — April 24, 2012 @ 3:30 pm

  34. Good article as usual Dave, but it shouldn’t surprise people that 3-13 is fluky. One doesn’t have to be a sabermetrician/ a scout / math major to know that the royals aren’t 3-13 bad. Your arguments were good though

    Comment by clint hulsey — April 24, 2012 @ 4:05 pm

  35. I am a big fan of both Rany and Dave’s writing and their thinking. This seems like an interesting way to promote the new “Fangraphs/SI Power Ranking”. However, I see a problem.

    If the Royals are just throwing away outs when they have men on base or in scoring position, wouldn’t that skew the results?

    Rany’s article seems to point out worse than questionable base running and sacrifice bunting choices. These problems can only occur with runners on.

    It seems that bad base stealing (worse than every other team except the Dodgers) and inefficient sacrifice bunting, the “more normal distribution on timing of hits” is not entirely random. To the extent that the disparity is due to bad manager/player decisions, wouldn’t it’s continuation be affected by whether the decision making process changes?
    Discounting the disparity in scoring as a lack of performance in the clutch seems to ignore the baserunning and bunting practices, which only apply when runners are on base.

    Comment by hoser — April 24, 2012 @ 4:28 pm

  36. The “bug” in these rankings is the failure to regress data. Simply tagging things with “sample size” and saying take with a grain of salt and then proceeding to use the #’s anyway is frankly…. ridiculous and a complete abuse and misuse of statistics.

    In addition to UZR, there can be issues with BSR, FIP (HR/FB outliers, early season park effects, opponent quality over such a small sample), and even the hitting stats (BABIP, parks again, etc…)

    This is why there is this thing called “regression”… the point of it is to account for sample size, not simply to have it both ways and say “the data isn’t clean, but hey let’s analyze it anyway as if it is”.

    Seems like having this published in SI with such an obvious flaw in methodology is not shining a good light on the SABR community.

    Comment by tom — April 24, 2012 @ 5:15 pm

  37. I think common sense would tell you that any team with a win % below .370 or above .648 is probably flukey due to small sample size.

    How much of every team’s curren twin % is influenced by the quality of teams they have played. Dodgers have 7 games against the Padres, 3 against the Pirates, 3 against the Astros, 3 against the Brewers, and 1 against Atlanta. That’s why they have such a good record. Not BABIP luck (although that’s probably there), not pitchers out-performing their peripherals (although that’s probably true too). Teams just haven’t all played similar competition. KC’s schedule has been fairly brutal. Of their 16 games, 13 of them have been against teams projected to be above .500 (Blue Jays around .500 in the AL East counts). Looking ahead, May looks to be pretty rough too.

    I love B-R’s SRS they have. It uses a similar approach that you’d use in college football. Measures the quality of team you’ve faced and the run differential. I think in a small sample size, this is something that could be useful.

    Comment by Antonio Bananas — April 24, 2012 @ 5:15 pm

  38. On the SI site they indicate nothing about sample size or regression. Is the casual Baseball fan supposed to take away from the rankings that Kansas city is a likely playoff team? If not then what is the purpose?

    Comment by RichW — April 24, 2012 @ 5:33 pm

  39. If you’re going to be so brash as to use “asinine” in a comment, please take the extra one second to spell it correctly. Otherwise, you look asinine yourself. I mostly agree with your statement; I mock your presentation of it.

    I am a huge fan of Dave’s, but the argument for ranking the Royals seven is unconvincing, apparently to the casual fan and learned statistician alike. In all fairness, Dave qualified his post. “No one thinks the Royals are really the seventh best team in baseball, including us. In fact, if the rankings were re-done to reflect last night’s game as well, they would have already fallen to 9th, and their .525 WAR Winning % puts them closer to 17th than to 8th.”

    The worst record in baseball might actually belong to the 17th-best team? Now that I can buy. Maybe.

    Comment by sportsczar — April 24, 2012 @ 6:16 pm

  40. oh, good one

    Comment by jim — April 24, 2012 @ 6:31 pm

  41. you seem surprised, are you not familiar with dave cameron’s work?

    Comment by jim — April 24, 2012 @ 6:31 pm

  42. Just read the last line of the post:

    “And to me, that’s the entire point of advanced metrics – to shine a light on the non-obvious story, and help us understand the why behind the result.”

    Here I was thinking the point of advanced stats was to better assess and objectively measure performance (and in some cases project it). When the primary purpose (“the entire point”…. really?) is to be contrarian or ahead of the curve or be viewed as smarter than the layman, it leads to the advanced stats being misused.

    Yes it is boring to read an obvious conclusion based on advanced statistics, but the over-riding need to be “non-obvious” has led to some really poor application of advanced stats and an increasing trend of bad analysis and misuse of statistics

    Here’s a crazy strawman…. maybe the variation is in the #’s with noone on base and the men on base #’s are more representative. Given the absurdly small samples we are dealing with there is no way to know. Why is the assumption that the men on base #’s are the outlier? Well because that’s what fits the story being told…. a more fundamental approach to the analysis (including regression) would sort out some of this out.

    Comment by tom — April 24, 2012 @ 7:07 pm

  43. The Royals have been caught stealing 8 times already. I also feel like I can remember them getting thrown out in a few other cases. Probably wouldn’t make a huge difference though….

    Comment by Metsox — April 24, 2012 @ 8:01 pm

  44. I’m not entirely sure what you mean. If you regress to the mean, you maintain the ordinal ranking of the teams.

    And saying “sample size” and posting the data anyway isn’t an abuse of statistics. It’s a sample. You’re saying “here’s a sample of data of the 2012 KC Royals”.

    Ranking these samples in order of some “performance” also isn’t terrible. Pretty meaningless, though.

    Comment by CJ — April 24, 2012 @ 8:26 pm

  45. Forget “sabermetric” strategy. Talk to any guy in the street who watches baseball and he’d say that they’re throwing away runs.

    I fully expect insane baserunning to regress to more normal levels.

    Comment by CJ — April 24, 2012 @ 8:29 pm

  46. Which fielding system do you prefer and why?

    Comment by CJ — April 24, 2012 @ 8:31 pm

  47. “Why is the assumption that the men on base #’s are the outlier?”

    Because there is a lot of data to suggest that teams performing very differently with men on versus with the bases empty tend to regress toward a more common level of performance somewhere in between.

    Tom Tanger summarized some results here:
    http://www.tangotiger.net/clutch.html

    The premise is that clutch exists, but a lot of data is required to show a little clutch. The conclusion is that much of the spread in performance in various situations is due more to “luck”, or randomness, than to a true difference in skill ability.

    So basically we have to adjust our null hypothesis to the overall expectation–that teams will regress toward past expectations–and wait until there is enough evidence to reject that notion.

    Comment by Matthias — April 24, 2012 @ 8:38 pm

  48. I should clarify, Tango’s research was in reference to individual players. However the concept still applies. Larger samples are needed before we should accept that the Royals can’t hit or pitch “in the clutch.”

    Comment by Matthias — April 24, 2012 @ 8:40 pm

  49. My point is not whether or not there is clutch hitting…. it’s the assumption that the noone on base # is the “real” # and the Royals are just getting unlucky and that of course is the # things will regress to.

    Given the size of the sample…. where is the “in between”?

    The tone of the article suggest it’s obviously closer to the noone on base # and it’s a luck thing – given the sample size I don’t see how that can be assumed.

    Comment by tom — April 24, 2012 @ 9:02 pm

  50. “But, just adding 11 runs doesn’t just fix a 3-13 team, right?”

    If six of those 13 losses are by one run, I’d say adding 11 runs in the right places could do more than fix a 3-13 team.

    Comment by TecJug — April 25, 2012 @ 12:29 am

  51. They all suck pretty hard. In a few years when FieldFx is in full force, everyone will be talking about UZR being the RBI of fielding statistics. There is opportunity value in UZR, which is why a single season’s worth of data is not accurate.

    Comment by UZR is a Joke — April 25, 2012 @ 8:32 am

  52. agree. these are good arguments for showing that 3-13 isn’t necessarily deserved, but to me wins and losses will always be the best measurement of how good a team is.

    Comment by joele — April 25, 2012 @ 10:07 am

  53. I don’t think Sanchez saw you staring at him. That was an easily winnable game for KC last night…if Sanchez hadn’t walked 7 guys in 4.2 IP. The Tribe has been the opposite with the timely hitting so far, amazing KC didn’t get blown out.

    Comment by Tom — April 25, 2012 @ 10:14 am

  54. When Colin Wyers yells at Dave Cameron it’s like hearing Mom and Dad argue all over again.

    Comment by JDanger — April 25, 2012 @ 2:31 pm

  55. I have no interest in engaging you in any kind of fight, Colin, so this will be my only response to you.

    The Royals record, through Sunday, was 3-13. Their expected record based on total team WAR through Sunday was 8-8. The gap we’re talking about is five wins.

    Four of those five wins can be explained by the distribution of hits. Just using expected RS/RA (as your own site does with third order wins) would have had the Royals at 7-9. At most, you could potentially ascribe the effect of the Royals positive UZR to one win.

    Now, about that +8.2 fielding rating. +1.6 runs saved have been attributed to garnering outfield assists, +1.2 runs saved to their efficiency in turning double plays, and +1.6 runs saved due to their low number of errors. Hopefully even you would agree that these events are outside the realm of your criticism of the way fielding metrics have been collected and calculated, and it’s far less likely that there’s any kind of systematic bias interfering with these particular runs saved. In reality, your point of contention with UZR has always been with RangeRuns, of which the Royals were only credited with +4.8 runs saved. To quote the +8.2 fielding rating and pretend that those runs should be discarded in their entirety is disingenuous.

    If we zero out those +4.8 range runs, their expected record through Sunday would have still been 8-8. They would have placed a few spots lower in the ordinal rank because of the large cluster of teams all hanging out in the .500 range, but the overall evaluation of the Royals wouldn’t have changed much at all.

    And, of course, ordinal ranking isn’t something that one should get overly worked up about, especially when the scale is provided. If all of the teams between #7 and #15 are projected to have won and lost an equal number of games, the idea that the ordinal ranking is actually telling you that team #7 has performed better than team #15 is an insincere attempt to actually look at what the ratings are saying.

    Comment by Dave Cameron — April 25, 2012 @ 4:19 pm

  56. The expected number of wins is included in the table. Every team between 7-15 was expected to have an equal number of wins and losses.

    Comment by Dave Cameron — April 25, 2012 @ 4:23 pm

  57. As explained above, this is not a blip caused by small sample UZR. This is a blip caused by small sample situational hitting. I wrote about the actual effect that has driven a great majority of the gap between the Royals actual record and their expected record (by whatever kind of component winning% you want to use – they all return values very similar). Those who are attempting to ascribe the gap in between the Royals win% and expected win% primarily to their fielding rating are simply incorrect.

    Comment by Dave Cameron — April 25, 2012 @ 4:26 pm

  58. Of course – any “snapshot” style power ranking that uses only current year data is not going to reflect a team’s true talent levels two weeks into the season. We don’t claim that these do that, and neither does SI.

    Comment by Dave Cameron — April 25, 2012 @ 4:28 pm

  59. The existence of one outlier doesn’t prove a model is broken.

    Comment by Dave Cameron — April 25, 2012 @ 4:29 pm

  60. We’re not trying to tell people who the best teams in baseball are every week – SI specifically asked for a “snapshot” system that would evaluate teams where they are through a given date. Power Rankings columns are almost always a reaction to current performance as measured by wins and losses. This simply switches the measure of current performance to the underlying components that drive wins and losses.

    Comment by Dave Cameron — April 25, 2012 @ 4:31 pm

  61. When we did our Power Rankings here on the site last year, we regressed them very heavily against preseason projections. We certainly understand that any measure that is attempting to answer the “who is the better team?” question would include much more than just current season data. That is not the question SI is trying to ask with their power rankings, however. Their list is backwards looking, asking how teams have played up through the date of publication.

    Comment by Dave Cameron — April 25, 2012 @ 4:34 pm

  62. Gotta defend the orthodoxy at all costs, eh Dave?

    I think it was Emerson who said “A foolish consistency is the hobgoblin of small minds.”

    Comment by objectiveobserver — April 26, 2012 @ 12:30 am

  63. From the power rankings themselves:

    “Welcome to the SI.com Power Rankings, powered by FanGraphs. You will probably notice that these Power Rankings don’t look like most other Power Rankings you’ll see around the web, as our system is based not on the current standings or a gut feeling about team quality, but on how well they’ve performed at the underlying traits that predict future performance better than wins and losses.

    Those traits — getting on base, hitting for power, running the bases effectively, getting strikeouts while avoiding walks and home runs from their pitchers, turning hits into outs on defense — are summed up in our Wins Above Replacement metric, and then we’ve translated each team’s total WAR into an expected winning percentage based on the number of games they’ve played this season. By utilizing WAR, we can better identify which teams are actually playing well and will likely sustain their success going forward.”

    So is it a “backwards looking” list or is it about “success going forward?”

    Comment by Colin Wyers — April 26, 2012 @ 2:43 am

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.

0.132 Powered by WordPress