Happy Thanksgiving! Regularly Scheduled Articles Will Resume Monday, December 1
May 26, 2011
Answers from a Sabermetrician, Part 2
Believe it or not, most of our writers didn't enter the world sporting an @baseballprospectus.com address; with a few exceptions, they started out somewhere else. In an effort to up your reading pleasure while tipping our caps to some of the most illuminating work being done elsewhere on the internet, we'll be yielding the stage once a week to the best and brightest baseball writers, researchers and thinkers from outside of the BP umbrella. If you'd like to nominate a guest contributor (including yourself), please drop us a line.
You asked, he answered. Below is the second and final batch of responses to the questions BP readers submitted for sabermetrician Tom Tango. All questions are presented in their original form.
TOPIC #4: Team management
Subtopic: Relevancy of sabermetrics
The harm is really in the inefficiency of making the same mistakes. The problem really is that the overall structure is not set up to support a shift in mindset. Rather than being a holistic community, each team, each media outlet, each fan, seems to have more comfort in its own world. I guess the real-world analogy would be Homeland Security trying to get the CIA, FBI, and the other agencies to be more cooperative.
As to how to combat this, my preference would be for MLBAM to create some sort of saber-wing, one they'd support as an investment (not as an expense). The reason that OBP became prevalent is because MLB declared it an official stat about 30 years ago. It's really that simple. If MLB decided to support RA9 instead of ERA, then everyone would simply move to RA9. And if Wins Above Replacement (WAR) and Fielding Independent Pitching (FIP) were recognized as official stats, then we'd make the impact that way.
There are probably plenty of them. The assistant GM of the Twins, for one, did an interview last year where he confirmed it. But given how successful the Twins have been in the last 10-15 years, and doing it on a small payroll, it's hard to argue that they are not efficient at what they are doing. It's like arguing that Vladimir Guerrero could be a better hitter if he wasn't so aggressive. Perhaps the reason he is such a good hitter is exactly because of his aggressiveness.
Subtopic: Manager skills
I agree that I would prefer a manager who manages egos to a manager who manages by rote. That said, there's nothing wrong with having a tactical bench coach who should be aware of the percentages. This is the way it works in corporate America and in the White House and everywhere else. So, you get a strong leader that everyone respects, who is surrounded by experts in their particular domains.
Subtopic: In-game strategy
Morgan Ensberg has had several posts on his blog on this topic, and he quite enjoys discussing the various percentage plays. I think there's a definite place for players to learn here (dropping a flyball foul out with a runner on third base, for example; where to play for the bunt; which base to throw to). I don't know how prevalent all of this is across MLB. They probably learn enough that their intuition will guide them correctly most of the time.
Subtopic: Past Discoveries
Phil Birnbaum asks:
My "10" would be regression. As for any question regarding teams I've worked with, I always issue a no comment. And I have no knowledge of what other teams have learned.
Subtopic: Future Discoveries
dREaDS Fan asks:
PITCHf/x, FIELDf/x, HITf/x (or Trackman, or any of the real-time tracking systems) are the gold mine. What those systems will give us is the convergence point for performance analysis and scouting observations. That's the pinnacle of sabermetrics.
What sabermetrics is about is trying to infer the true talent levels of players based on performance data. If we had god tell us, then we wouldn't need any performance data. What we have right now is human beings (scouts) who try to tell us this. This is a great benefit at the minor-league, college, and high-school levels, and somewhat beneficial at the major-league level. The real-time tracking systems are a sort of scouts+. We don't need to rely on scouts to tell us how much a pitch breaks, since we can measure that. We can free up scouts to tell us things we can't readily measure. In the end, quantitative analysts and scouts are after the same thing: determining how good each player is.
Subtopic: Other discoveries
I think the game within the game would be one area: the game theory of pitch selection and location on the pitcher side and swing/take on the batter side.
1. Regression toward the mean.
TOPIC #5: Outsiders
The most important thing is for people not to misinterpret the numbers they see. If they were completely unaware of the numbers, that'd be great. If they knew exactly how to interpret numbers, that'd be great. The problem is people ascribing meaning to numbers without knowing why they are ascribing that meaning. Felix Hernandez and Jered Weaver were both 13-12 last year. We know exactly why they went only 13-12, and it has nothing at all to do with their talent levels. If people want to presume that that 13-12 must necessarily be linked to their talent levels to some degree, that's where the problem is. Basically, people know enough about numbers to be dangerous.
Richard Bergstrom asks:
I don't know that my opinion really matters here. Anything that pushes us in a direction of less gasbaggery is a good thing. Anything that gives good analysts a viable path to a vocation is a good thing.
Shaun P. asks:
My boy isn't old enough to have any such insights in baseball, but he does have them for many other things. For example, he’s great for creating a random game with the other kids, coming up with rules on the spot that require just a little tweaking. So, I think his insight is how to balance the strategy to make a game playable. I look forward to him one day telling me that MLB should have a penalty for mid-inning relief changes, once teams go with a 14-man pitching staff. He won't be constricted by the inertia that seems to trap the adult baseball fan (but that somehow doesn't constrain that same fan in the other sports he follows).
Subtopic: On insiders
Excellent question. I wish I could talk about examples that on the surface look questionable but are revealed to be rational once you know more. I would give wide latitude to the idea that the front office is knowledgeable, intelligent, and efficient. And you should think only in some small circumstances that management misvalued the situation.
TOPIC #6: Statistical Theory
All data is a sample of the true rate. And the more data you have, the less uncertainty you have. You actually need one million trials in order to be certain at the .001 level. Take the case of a hungover or injured player: if we don't know his condition, then even if he has an 0-fer, we wouldn’t be able to distinguish that from his other 0-fers, and therefore our mean estimates and uncertainty level would be similar in both cases. But, generally speaking, yes, you should look at more recent performance and give it greater weight, and the further away you get from that date, the less weight you afford it. Mariano Rivera's true talent level on July 1, 2006 is going to be established, to some degree, even by how he performed on May 16, 2002.
Subtopic: Forecasting Minor League Players
I can't speak to the arrogance of anyone. Just my own. I presume you are talking about players that tear up the minor leagues, but are for whatever reason not in MLB? This brings up a good point regarding the so-called "Minor League Equivalency" (MLEs). Several years ago, I pointed out several problems with MLEs (please read that). The basic issue is that we are testing our models against a sample of players that have been preselected to play in MLB. But we don't know the reason they were selected. Was it because their scouting profiles were good, or was it because they tore up the minor leagues? By ignoring this parameter, it would seem based on MLEs that anyone who has a great Triple-A season will be a good candidate for MLB. That selection parameter is important. You can apply the same reasoning with Japanese League players.
So, if you are asking me to score this one, I'd say that scouting is an important parameter, and that without it, you have to dial down your enthusiasm for Quad-A sluggers.
K rates and BB rates are the best predictors because those metrics have less noise than other metrics.
Subtopic: PED Footprints
You may be able to find it based on the spread in performances, but you'll never be able to link it to any one player.
Subtopic: Park factors
Unless a park's configuration has changed, the climate has changed, or you have drastically different kinds of personnel from year to year (such that different players take advantage of different parks in different ways), any and all fluctuations would be random variation around the park's true factor. As for how ESPN calculates it, I think they base it on games played, rather than PA or IP.
Slots? No idea.
Subtopic: Games over .500
This used to bug me too. The key is to not take it too literally. Games over .500 is really won-loss differential. Wins above a .500 team would be won-loss differential divided by two.
TOPIC #7: The Rest
Subtopic: The Book
Thanks for the support. Writing a book is a horrible ordeal. The three of us probably collectively spent over 1000 hours on it. And in return, we earned between minimum wage and a starting salary out of college. And that's for a book that was a moderate success. So, to justify spending that amount of time, a second time, with the risk that it won't be as well-received, is going to be pretty difficult.
Another disheartening thing was the proof-reading. We spent alot of time proof-reading each individual chapter (which we saved as separate files). The two editors were whoever did not write that chapter, so that worked out great. When we were all happy, we then merged it into one big file, and we gave it a final walk-through. We found at that point in the process, no exaggeration, over one thousand errors. It was a disaster. So, we spent two weeks fixing all of that, and we gave it a second final walk-through. We found several hundred additional errors. It was a horrible experience. We had a third and fourth "final" walk-through. At one point, we agreed, this is our final-final, and regardless of what we find, we won't fix these typos. And when we finally published, guess what, we found an error on page two of the foreword (and more errors later). It's a thankless job, a non-productive job, one that we spent two months on. So, when I see people complaining about typos in BPro and other books, I just shake my head, because I will never criticize someone for that.
Phil Birnbaum asks:
Shoes! Women love shoes. They spend an inordinate amount of time accessorizing, shopping, or otherwise selecting shoes. If you have no idea what to say to a woman, talk about her shoes, because you know she spent a non-trivial amount of time deciding which pair of shoes to wear.
Otherwise, I'm as clueless as you are as to why they can't see the benefit of arguing over whether to bring in Mariano Rivera or Rafael Soriano with a three-run lead in the ninth inning.
Subtopic: English language
Jason Wojciechowski asks:
Did you know it used to be called "Base Ball?" And some bright and enterprising person decided that since those two words were used together so often, they should either by hyphenated or become a single word. If lawyers can decide that here to fore can be spelled as heretofore, I'm not going to wait for the word police to recognize that “a lot” makes sense. Finally, the English language is flexible enough that if new words can get created, they will. If you are really interested, I wrote alot more on the subject here.