StatSpeak World Famous Roundtable: April 7

Welcome to a new wrinkle here at StatSpeak, one that we hope becomes a weekly feature.  Eric and Pizza Cutter will be inviting other Sabermetrically-inclined bloggers to join up with us and answer some pressing (and not-so-pressing) questions from the world of baseball.  We all toss out a question and we all answer each other’s questions… hopefully with answers that don’t involve “well he has a .300 AVG!”  Look for the World Famous Roundtable every Monday morning.
To get us started, we’re honored to have John Beamer of MVN’s Atlanta Braves blog Chop-n-Change and The Hardball Times to join us for our first roundtable.  Read on as we discuss managers, the frontiers of Sabermetrics, and the hitters who are due for a reality check.
Question #1: With pitch f/x having revolutionised baseball analysis over the last 12 months, what in your opinion are the three biggest unanswered debates in our National Pastime?
John Beamer: 1. Fielding. With the data we have we are probably at the limit of the potential of existing fielding systems. However, the different systems don’t correlate particularly well with each other mainly because of data errors. Over the next five years with more precise (computer aided) measuring techniques we should be able to understand the different characteristics of defense a lot more accurately and measure these.
2. Game theory. What is the balance between the mental and physical aspects of the game? Some players are good because they are able to out think their opponents; some just use brute force; some both. Game theory is a super critical part of the modern baseball game and we just have no idea how different players exploit it. Of course the challenge is that all this happens on largely a subconscious level, so how do you measure it? Again with pitch f/x we should be able to get a handle on exactly how pitchers and hitters respond to different situations and see the extent to which they employ game theory.
3. Stats and scouts. As the stats and scouting side of baseball start to merge a whole host of new, exciting development should pop up. For starter we should see a (small) jump forward in the accuracy of projection systems as we get more data to throw into the mix. Also we will be able to answer questions around the impact that a pitcher’s mechanics have on, say, the outcome of the pitch and the tendency to pick up an injury.
Eric Seidman: Out in the sabermetrics community there will always be topics that we find ourselves discussing; it starts with a cursory examination of a subject and soon flourishes into widespread debates with new metrics and systems like Pitch F/X.  The three topics I personally feel will find themselves at the center of the statistical spectrum are:
1) Clutch Performance – it has always been there, is still there, and will not go anywhere.  Too many people feel too strongly about this subject to just let it die.  Personally, I’m of the mindset that clutch situations exist but that no player is a “clutch player” due to the transience of this phenomenon and the lack of correlation year to year.  Ask people who is the clutchiest player in the league and David Ortiz will likely be the answer; Ortiz, who, as Pizza Cutter pointed out in his Boston YIR, was the 5th unclutchiest player in the AL in 2007.  I feel that studies on this subject will never die because, much like Hall of Fame voting, there is no set criteria for basing decisions on and many people like to cherrypick in order to prove themselves correct.
2) The effect of managers – while no metric will ever be able to account for team chemistry in baseball the way that Lenovo Stats attempt to measure in basketball (it’s the same players together 90% of the time) the skepticism surrounding just how much a manager actually does is only going to grow and grow and participate in a tug of war with those of the opinion that managers truly effect games.  When more casual fans look at Gassko’s managerial article or the recent WSJ article mentioning Gassko’s work it becomes hard to believe that a manager only cost his team 3 games during the course of one season; after all, they watched him make boneheaded move after boneheaded move and could probably point out twelve different games he single-handedly blew for his team.  They fail to realize that the managers actually had a part in their team winning games, too, only this goes unnoticed.  When a team loses the manager takes the majority of the blame however when a team wins it seems that everyone shares credit equally. I don’t think we’ve even begun to scratch the surface with this one.
3) Projection Systems – I’ll never truly understand the fascination with projection systems but said fascination does exist and baseball analysts will continue to find new ways to harness methods in order to provide the most accurate projections as possible.  Sometimes predictions are correct and other times they are not; it is as simple as that.  Despite this, I would fully expect 5-6 new projection systems that are supposedly “better” than what is currently out there by the end of 2010.  For me, personally, give me Tango’s Marcels.  They don’t try to do too much and you know going in not to treat them as gospel but rather a baseline as to what MAY happen.
Pizza Cutter: Pitch F/X hasn’t even gotten to the good stuff yet!  At some point, we’re going to be able to do some good detailed analysis of the mind game that goes on between pitcher and hitter.  What pitch gets thrown when and what’s the optimal strategy for mixing up pitches?  Is it movement that fools a hitter or is it location or is it providing the proper mixture so that the batter doesn’t know what’s next?   That’s the question that pitch f/x can start to answer that we don’t have any good solid quantifiable stuff about yet.
The second issue that Sabermetrics hasn’t yet gotten to is the issue of multi-variable explanations for things.  Usually, most Sabermetric studies have looked at one variable relating to another, which is a good first step.  But as the field progresses, we’re going to have to use more complex models and look at several variables.  For example, to have a breakout year (or even to improve slightly) a player must improve several different skills.  If he improves only one of them, it won’t show on the field.  If he improves several of them, he’ll get an effect that goes above and beyond what he otherwise would have gotten. 
The third issue goes back to the old scouts vs. stats debate.  This is something that would have to be done within a MLB organization itself (and it may very well be going on), but a good mixed-methodology study on player scouting and development would do wonders for our understanding of the game.  By mixed methods, I mean that you take the qualitative data that you have (scouting reports) and analyze them in a standard way.  There are techniques for that type of analysis (can you tell I’m a psychologist by training), but I think the biggest problem with scouting is that no one has made use of this type of analysis.  At least that I know of.  There will always be scouts (and there always should be).  Let’s see if we can improve our understanding of what they’re really saying.

Question #2: A hot topic in the sabermetrics community is being able to evaluate the efforts of managers.  Despite this it becomes very hard to separate player contribution from manager contribution.  Certain analysis, like the differential between actual and pythagorean records, are considered to be an indicator and Bill James even has stats kept on how quick a manager pulls an ineffective reliever.  My question for the esteemed panel is – when these statistics are revisited at the end of the season which manager do you foresee being ranked as the worst?
John Beamer: This is a toughie. The value the manager adds is still very questionable. Despite many claims there has never been much categoric evidence that teams can under/ over perform their pythag record consistently. People point to Bob Melvin and the Snakes last year … let’s evaluate that stat in 2010 and see if he is still outperforming. My view is that the manager’s role is to make the right strategic decisions (which sometimes might be to keep an ineffective reliever in if the game is already lost) and build team harmony.
I guess I must choose …. gee, I’m really hunting in the dark on this one. I’m going to plump for Dusty Baker. I know some of the managerial metrics rated him very highly when at San Francisco but from what I saw when he was in Chicago he made some appalling decisions. I hope he succeeds at the Reds because the game needs more black managers but I fear it will be a long ol’ road for Mr Baker.
Eric Seidman: It becomes very difficult to answer questions about managers without watching a multitude of games for each team, really studying the moves one makes.  I don’t miss Phillies games, and I watch every Greg Maddux start, as well as most Pirates or Braves games as I get FSN and TBS; outside of those teams and the ESPN games I always watch I am unfortunately left to rely on the media to tell me which managers are making good or bad decisions.  Due to this I cannot name one particular manager who may hinder his team’s fortunes but I will give a solid managerial attribute that will no-doubtedly lead him down this very road – – not understanding your team.
It seems simple enough but I have seen many managers come into a situation and try to marry their style with a group of guys either not yet ready for said style, or really in need of a different style.  Take Charlie Manuel of my Phillies, for example.  Charlie does a great job of connecting with the players but he micro-manages like crazy which tends to put pressure on his players to bail him out.  Manuel continually fails to realize that the Phillies bullpen is, pardon my french, suckay-voo, and so removing Pat Burrell in the 7th inning of every game is going to be detrimental; when the bullpen eventually gives up the lead it means So Taguchi will be batting with runners on 2nd and 3rd, not Burrell.
Then you have the managers like Dusty Baker, who come into camp really believing the Reds are going to compete for the National League Central.  It’s one thing to say that to the media in order to continue to be friendly Dusty but to give a starting CF job to Corey Patterson instead of letting Jay Bruce see what he can do at a major league level comes off as borderline ridiculous.  It’s one thing if you’re the Phillies in 2004 and you want Kenny Lofton to provide veteran leadership to a team on the cusp of stardom but the Reds, honestly, are not going anywhere this year.  I like to watch them and they are a fun team, but I would bet a gagillion bajillion dollars they finish in third place or lower this year.  Understanding that your team may be good soon but not quite yet is how someone like Dusty should be managing, giving guys a chance to play that could develop.  Corey Patterson will not lead this team anywhere.  Jay Bruce might not either, but I would be very interested to see what the kid can do as a manager of that team.  Reds fans may disagree and come to his defense but a manager who truly fails to understand the situation he is in is going to be the most detrimental to his team’s success.
Pizza Cutter: I’m not a fan of using the Pythagorean residuals to evaluate managers.  In fact, I think it’s a rather awful way of looking at these types of things.  Managers have three jobs.  They talk to the media, they are the putative leaders of the team (although the actual leader is not always the guy with the title) and they’re in charge of keeping the players in line and happy, and they run the game strategy.  The first piece isn’t a Sabermetric question.  The second could be (do some managers make their players statistically better consistently over time?) but Sabermetrics hasn’t gotten there yet.  On the third, the problem is that really most managers use the same playbook as all the other managers and have been slow to adopt several of the strategy improvements that have been suggested over the years.  There might be minor variations between managers, but try this: watch a game and “call” what the manager is going to do.  You can probably do it fairly well. 
There’s hope in that a guy like Joe Torre has benched Juan Pierre or that Tony LaRussa (of all people!) is batting the pitcher 8th (he apparently read this), but managers have been notoriously slow in actually putting into play some of the suggestions that have been offered by Sabermetrics over the past few years.  It’s like a race of the slow-footed.  They’re all lagging behind quite a bit.  ”Who will be the worst?” covers the fact that baseball managers are generally much behind the times. 
Question #3: Time to play the fortune teller.  Which hitter who had a better-than-expected season last year will come back down to earth this year.
John Beamer: I’m going to take the path of least resistance and use good ol’ gut feel to answer this one. The academically correct method is to compare 2007 actual perforamce to 2008 projection but I don’t actually have 2007 data on my computer. And give this is Statistically Speaking I feel we a bit of “gut feel” will counter some of the stats wonky stuff that Eric will come out with (sorry Eric)!
I continue to be surprised by how well Maglio Ordonez hit last year. In 2007 he hit .363/.434/.595 — that is solid contact, a good walk rate and some nice power. I reckon he’ll do well to get within 60 points of those numbers this year.
Another player who is due a regression (and I say this every year and continue to be proved wrong) is Chipper Jones. Jones had a phenomenal year last year .337/.424/.604. Pundits perenially expect him to slide down the aging curve and this year I reckon they may be right — it has to happen sometime. He is prone to injury and a little knock might keep him in the line up but with far less production. A look at the Braves bench where there is a serious lack of decent bats will only encourage this.
Oh …. and finally, I’m also expecting some of Jack Cust’s glory to edge away this year.
Eric Seidman: This is a tough one to tackle because hitting regression can occur in many areas.  For instance, I don’t expect A-Rod to hit 54 home runs this year nor do I expect Magglio Ordonez to post a .363 batting average.  I don’t see Jorge Posada with a .338 batting average nor do I see Matt Holliday duplicating all of his 2007 numbers.  To answer, though, I’ll look at players who had very high slash statistics (BA/OBP/SLG) and perhaps a high total in another significant stat that could contribute to some viewing their final 2008 statistics as “disappointments.”
First is Curtis Granderson.  I’m a big fan of his but I just cannot see him duplicating a .302/.361/.552 with 23 3B and 23 HR this year.  Even before his injury I did not view him posting similar stats as a sure thing; the injury only aids my cause.  His slugging percentage went through the roof due to his 3B count and the fact that triples count as three total bases.  While he may be able to consistently hit triples in the 8-13 range, 23 is an incredible number and should be thought of as more of an outlier than the norm.  Also, when I re-did my Oddibe Awards piece to look at the specific league averages from 1981-2007, Granderson won the 2006 award.  His 2006 saw him go .260/.335/.438 with 9 triples and 19 home runs.  While the HR may stay in the range of 17-28 my gut feeling tells me he will not reach any higher than 14 triples this year, meaning his SLG will drop, meaning his OPS won’t be as high, which in turn means that fans may wonder where the 2007 Granderson went.  He never went anywhere though because I don’t think he was there to begin with.  2006 may have been an exaggeration on the lower side while 2007 was an exaggeration on the higher side.  I would expect his 2008 to he some sort of a combo in the range of .287/.346/.502, with 14 triples and 19 home runs.
The other player I anticipate having a dropoff is Aaron Rowand.  This was an incredibly easy one as I watched him closely for the last two years.  Rowand is your prototypical hard nosed leader that will make stellar defensive plays and hit in the .275/.360/.475 range with potential for 15-20 home runs.  Last year he had a home/road OPS split of .937/.843.  He was in a contract year and had been playing in a notorious hitters park.  Rowand also had plenty of little dribbler hits that luck prevented from turning into outs; this boosts the BA/OBP/SLG and those who do not actually watch the Phillies play only see Rowand posting great offensive numbers.  Rowand has virtually gone to a polar opposite team this year; the Phillies, a high-scoring, high-potential, playoff team in a hitters park to the Giants, a low-scoring, poorly constructed, cellar dweller in a pitchers park.  In my new formula, LS (low-scoring) + PC (poorly constructed) – 12*CD (cellar dwellar) – 5*PP (pitchers park) = severe regression for Rowand.
Pizza Cutter: The Washington Nationals, coming out of Spring Training made the funniest move of the year so far.  Cristian Guzman will be their starting shortstop.  Guzman, who bears a .302 career OBP(!) somehow put up a .380 OBP last year  (.328/.380/.466), and then at the peak of his prowess got hurt.  How did he do it?  This 60% groundball hitter had a lot of seeing eye singles that found their way through the infield.  His BABIP spiked to .364 and people believed that he had “turned the corner.”  I realize that the Nationals suffered through all those years of being the Expos and of being basically neglected, and it’s not nice to kick a man when he’s down, but… he’s hitting leadoff.  That dull thudding sound you hear from the direction of Chicago is my head hitting my desk.


9 Responses to StatSpeak World Famous Roundtable: April 7

  1. John Beamer says:

    I’d actually be surprised if we saw many (any?) new projection systems before the end of the decade.
    I reckon some (PECOTA probably) will become more accurate as scout data is thrown into the mix but there isn’t that much more one can do to increase the accuracy of projections. By virtue of the limit on observations there is a (theoretical) maximum on their accuracy …
    Good round table guys — keep up the good work

  2. My point is that they should not be taken as gospel and should be treated more along the lines of a baseline of what may happen based on what has happened. You might be surprised how many people really take stock in these systems. If I recall correctly I remember a study done recently wherein fans made their own projections without any sort of system and they came pretty close. If I’m wrong and that study didn’t take place, it should take place because I’m sure the fans would come pretty close.
    Perhaps I worded that wrong in my answer. The “sometimes they are correct and sometimes they’re wrong” was more along the lines of meaning that none are infallible and should not be taken as gospel.

  3. Sky says:

    Pizza hit the nail on the head with Pitch f/x. It’s not just one new thing, it opens the door for hundreds of new things. It’ll be as important as retrosheet eventually, providing the data to do whatever people can dream up.
    I also have a bone to pick about Erik’s view of projection systems. I totally disagree that “sometimes predictions are correct and sometimes they’re wrong”. Even if Marcel outputs a 39 HR prediction for Pujols, there’s an assumed distribution around that 39 number. As systems get better, we won’t celebrate them nailing exact numbers more often, we’ll celebrate being way off less often and on average closer. It’s not an exciting business, but it’s extremely important. Helping people understand that projections aren’t hard numbers is critical.

  4. Pizza Cutter says:

    Eric, I think that there’s a bit more of an understanding of these types of projection systems as a baseline for expectations than you might believe. BP is good about providing percentiles for their PECOTA projections and a place like fangraphs puts all the systems on one page. Most of the systems are within shouting distance of one another on their projections.
    John’s right in that we have a hard time gauging a player’s true talent level based on so few appearances, and then projecting it out over a sample size that is itself a poor gauge of true talent level. But then, we are in the probability business.

  5. dan says:

    I see Sky is still alive… I was getting worried.
    Pizza- is there a reason you haven’t gotten into the pitchf/x data yet? I figured you were leaving it up to Mike while he was here, but do you have any intention of following up on some of the questions you asked in that section?

  6. I’ve actually begun looking into it as I have some aspects of batter vs. pitcher and pitcher w/umpire I would like to investigate.

  7. Pizza Cutter says:

    There are a couple of reasons that I haven’t gotten into it. One is that I have absolutely no programming skill so Joseph Adler’s Baseball Hacks scares me. Another is that most of the work being done so far is simply pitch classification and working out some of the bugs involved, which is best done by people who understand things like air resistance and rotation and things like that. It’s been 10 years since I read a Physics textbook. The last one is as simple as I’ve got other things on the docket now. At some point, I would like to start investigating pitch sequencing, but to be honest, at this point, I’d have to bum someone else’s database.

  8. Sky says:

    Dugout Central put out feelers for “creating a new stat” a few months ago. I suggested incorporating Mike Pags’ scouting data into a projection system, but they didn’t seem to think that was a good idea. They wanted to reinvent the wheel with metrics for OF arms and baserunning. When I pointed out things like that had already been done, the project was scrapped.

  9. Pizza Cutter says:

    It would take a lot of standardization, and I mean from the ground up to see what was actually predictive and what wasn’t. But, it’s doable.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: