StatSpeak World Famous Roundtable: May 26
May 26, 2008 1 Comment
Welcome to the Memorial Day edition of the roundtable. This week, we welcome back an old friend to StatSpeak, Mr. Mike Fast. Mike’s been hanging out lately at The Hardball Times and at his own personal blog Fast Balls. Today, we discuss the Marlins, re-aligning baseball, and what’s been going on in Sabermetrics for the last twenty years.
Question #1: In his 1987 Baseball Abstract, Bill James rates baseball statistics in terms of how meaningful they are on the basis of four criteria: (1) importance–does it correlate with winning?, (2) reliability–the extent to which the statistic truly reflects ability, (3) intelligibility–can the average baseball fan make sense of this information, and (4) construction–where James applied minor penalties for poor construction, such as Runs Produced giving only half credit to home runs. James rated ERA, OBP, and SLG as the most meaningful traditional single-season statistics. Among the more recent statistics, James did not evaluate his own creations, but rated linear weights the highest. Which of the new statistics since 1987 are the most meaningful, and which are the most meaningful overall, including the traditional ones?
Mike Fast: What are the best or the most commonly used of the new stats over the last twenty years. I’m not sure whether BABIP originated with Voros McCracken’s DIPS work or not, but it’s been one of the most influential stats over the last decade. It tells us a lot about whether pitchers and hitters are getting lucky or not. Or at least we think it does, on the whole. We’re still learning. Along the same line, we have Fielding Independent Pitching (FIP), which is a good ERA predictor based on similar ideas to DIPS. Bill James introduced Win Shares, and it’s had some refinements, like Win Shares Above Bench, but I’m not sure it ever really caught on to the point that people understand it, and it has its flaws. There’s Value Over Replacement Player (VORP), by Keith Woolner, denominated in runs. Hmm…there’s lots of versions of the same kind of things, and I’m not sure how many of them are improvements. I do like the batting runs linear weights system from David Smyth, Base Runs. On the fielding side, we’ve improved a lot with the play-by-play data. Without it, the best of the bunch is probably Dan Fox’s Simple Fielding Runs (SFR). With the play-by-play data, which is privately held, we get systems like Ultimate Zone Rating (UZR), Probabilistic Model of Range (PMR), and BIS’s Plus/Minus.
It’s hard for me to evaluate the fielding metrics since I’m not privy to the data. On the pitching side, I’ll credit FIP as the best new stat, and on the hitting side, Base Runs. FIP is on the same scale as ERA, which we understand well as fans, and it includes only three main rates: HR/IP, BB/IP, and K/IP, which are also well understood by fans. It also correlates well with winning and true talent of the pitcher. Base Runs is an improvement on Pete Palmer’s linear weights. It has all the same advantages in terms of reporting runs relative to league average, which we fans understand fairly well, but it does a little better job of correlating with actual run scoring. In the end, though, I have to stick with the same choices Bill James made: ERA, OBP, and SLG. For all their flaws and supposed flaws, they do a pretty good job, and we fans have years and years of context stuck in our brains with which to interpret them.
Eric Seidman: FIP is my favorite statistic primarily due to most of my research dealing with pitchers and my fascination with ingratiating non-statheads into the saber-community. Many are intimidated by statistical analysis and so statistics like FIP, EqA, and wOBA work so well, in my eyes, because they explain more than the barometers they represent while staying on a familiar scale. A casual fan knows that a 3.00 ERA and .288 BA is solid; when they see these numbers represented in a statistic that accounts for more variables they are more likely to understand and appreciate them.
The best part about using FIP in comparison with ERA is we can see a measure of actual results vs. skill based results to see which pitchers are outperforming their skills or underperforming their skills. Unearned runs are a result of errors, which come from those official scorers making decisions, so they are a little fishy in and of themselves. Walks, strikeouts, and home runs, the key ingredients of FIP are not arbitrary.
As far as the traditional statistics, I like to use ERA but only in the comparison mentioned above; I also like looking at the slash line BA/OBP/SLG. I don’t like any of the three on their own, with the newer statistics available but looking at all three side by side explains more than any on their own. Jason Giambi has a .236 BA this year… he also has a .384 OBP and a .516 SLG. Looking at those three on their own I would determine, in order: Giambi is a terrible hitter this year; oh, ok, Giambi is either walking a lot or hitting well and walking some of the time; and Giambi is a power hitter. Put together and we can see that his OBP and SLG are leagues higher than his BA, meaning he walks a ton and, while he doesn’t get hits quite often, they are usually extra-base hits.
Pizza Cutter: The research methods professor in me chuckles at Bill’s choice of words (The first criteria is actually “predictive validity” and the second is “construct validity”). Sabermetricians have focused on those first two, often at the expense of #3. I’m not so sure that’s a bad thing, although it does mean that we can’t really whine that the general public hasn’t quite gotten the hang of this yet. Of all the ones that are out there, the win probability added suite makes the most sense to me as something that’s intuitive to the general public, can be easily presented, and clearly tracks winning. I’d agree that the linear weights approach is the most informative and probably best meets criteria #1 and #2 for evaluating individual players, although it’s not really intuitive to explain to someone. Of the traditional stats, it seems like all that’s really needed is for people to understand rate stats per PA. So, perhaps K/PA or BB/PA or HR/PA could become fairly standard. That would be ideal. (If it could, then linear weights could probably gain a mainstream foothold.)
Question #2: In my recent interview with David Pinto, he mentioned that he thinks the divisions should be radically redesigned. So, springboarding off of that, in keeping with the same type of format–5-6-5 in the NL and 5-5-4 in the AL–redesign the divisions and explain the reasons behind why certain teams were moved.
Mike Fast: I have a hard time coming up with a better alignment than the existing one in the National League. Even if we weren’t constrained to the 5-6-5 alignment, I’m not sure I’d change it. I wasn’t a big fan of the Brewers moving to the NL in 1998, but now that they’ve been there for a decade, I wouldn’t change it back. You need an even number of
teams in each league, and there isn’t anyone else I’d switch at this point. I grew up when the leagues still meant something; they even had their own league offices. So any proposal that switches teams between leagues just doesn’t fly with it.
The existing NL divisions divide up pretty well between time zones, except for Cincinnati and Pittsburgh. Maybe you could switch Pittsburgh to the NL East and make that the six-team division, that might make as much sense as anything. Each of the divisions ought to have a balance between expansion teams and some of the original eight. I don’t care as much about the natural rivalries, but I’d try to preserve Giants-Dodgers and Cards-Cubs. Those make geographic sense, too.
In the AL, it’s also tough to come up with an improvement over the existing alignment. Texas and Tampa Bay are the outliers in their divisions geographically, but you can’t leave the West with only three teams by moving Texas into the central, and where else would you put Tampa Bay? If I were going back to 1998 to do it all over again, I might have moved Colorado to the AL West (or alternatively, placed the D-Backs in the AL), and left the Brewers in the AL Central. Then take the Royals and Rockies and make a six-team AL West. Given 15 years of history with Colorado in the NL, I wouldn’t change it now, though.
Eric Seidman: When David mentioned this, it didn’t hit me until afterwards to ask him how to realign them but I couldn’t really figure it out myself right away. I then remembered that some Pittsburgh Pirates executives came to Penn State a couple of months ago and a statistics professor friend of mine asked if they believed a competitive balance existed. The execs said yes. The professor then followed up with, “Well then is there any division you wouldn’t want to play in?” The exec responded the AL East. The professor then shot back that if a competitive balance existed and every team had a shot, there should be no divisions in which a team shouldn’t want to participate. In terms of evening things out, I’m not a fan of putting small market with themselves and big market with themselves. Pittsburgh-Cincinnati bugs me a bit with time zones but, otherwise, with the divisions keeping the same titles these basically make sense.
If we were to rename the NL East into something different and the divisions weren’t based on geography, it might be interesting to see random re-alignments every so often, but with the geography based divisions named the way they are, perhaps moving Pittsburgh or Cincinnati somewhere else would work; everything else seems just fine.
Pizza Cutter: Why not a 5-5-5 split in both leagues? It makes no sense that the teams of the AL West have a 25% chance of making the playoffs, while the NL Central teams have a 17% chance. Doesn’t seem very sporting. Plus, that would force MLB to have interleague play every day. At one point, there was talk of moving Arizona into the AL West (right after their World Series win, meaning that the defending NL champions would have been playing in the AL!) and having Houston move to the NL West. I liked that plan, but then I like things that are symmetrical. Usually when these designs are floated, they revolve either around grouping teams together geographically (the Mariners, Angels, Dodgers, Padres, A’s, Giants, and Diamondbacks in one big West Coast division? So what?), or for competitive balance reasons (put all the small market teams together, so at least one of them will make the playoffs. Sounds like the BCS.) I’m not opposed to some limited restructuring, but I don’t see any need to start radically shifting teams all over the place. Not for any traditionalism reasons. I just don’t see the point.
Question #3: Last week, we discussed how “real” the
Devil Rays are. What about that other surprisingly-high-in-the-standings-team from Florida, the Marlins?
Mike Fast: It’s not really a surprise that the Marlins are mid-pack (9th in the NL) in run scoring. Ramirez, Willingham, Uggla, and Hermida are a good group of hitters, even though they don’t have much major league service time. What’s a surprise is that they are also mid-pack (7th) in pitching. Most of their pitchers were expected to be mediocre at
best, and they didn’t appear to have any real stars. If you look at walks and strikeouts, several pitchers appear to be getting lucky in the early going (Scott Olsen, Renyel Pinto, maybe even Kevin Gregg). Who’s a real breakout candidate among the pitchers? Possibly Andrew Miller? Justin Miller looks okay, but it’s reaching to call him a star. I don’t expect their pitching to perform as well June-September as it did in April and May, and being mid-pack in offense and toward the back of the pack in pitching won’t be enough to stay on top of the division. It’s always nice to have a 29-20 start in the bank, though.
Eric Seidman: They are as real as the Nationals were a few years ago, staying in first place until the all-star break and then fading to an 81-81 record and last place. The Marlins have talent, and will not be walked on, but they are not very likely to win this division yet. In high school, we used to play some bottom of the barrel teams comprised of kids with a lot of talent but lacked most fundamentals. If we could score in the early innings and take upwards of a 6-0 or higher lead they would not come back; they would take themselves out of the game and that would be that. However, if we kept them in the game, suddenly their adrenaline would arrive in hordes and they would play to our level. While I’m not using this on a game by game basis, if nobody in the NL East starts to step up and run away with the division, and we get to August with the Marlins still ahead, they could very well “cowboy up” and skimper away with it.
From a personnel standpoint, I mentioned the other day at Fangraphs that, quite simply, Mark Hendrickson is actually pitching well. I can’t imagine him sustaining this performance all season because, well, he is Mark Hendrickson. Uggla is currently riding an insane month of May and Hanley is Hanley. They haven’t even played the Phillies yet, so it isn’t as if this is a fair sampling of their schedule with which to base judgments.
Give me Hamels-Myers-Moyer and Johan-Perez-Maine over Hendrickson-Olsen-Miller. I can definitely see them finishing upwards of 83-79 this year but that is not going to win the NL East.
Pizza Cutter: The Marlins’ rotation is awful and is getting 5.4 innings to the start. So, the bullpen is pitching 3 or 4 innings per night. Thankfully they’ve been good. The problem is that their top three relievers by WPA (Gregg, Pinto, Waechter, all have BABIP’s below .240). The offense consists of Dan Uggla and Hanley Ramirez with occasional interjections of Josh Willingham… and that’s about it. And so far, they’ve spent an inordinate amount of time playing against Washington, Pittsburgh, and Milwaukee. The Marlins are going to end up as proof that if you put 25 major leaguers near a baseball field for 162 games, they can generally contrive to win 65 or 70 of them. If they’re having a good day, maybe they sneak up to 75.