Keith Woolner is an author of Baseball Prospectus.
Keith Woolner: Hi everyone, let's get chatting!
jschmeagol (carlisle, pa): Can you tell me the differnce between VORP and RARP? I know that they measure the same thing, but what are the systematic differences between them?
Keith Woolner: VORP and RARP are very similar in concept -- VORP was developed by me, and RARP by Clay Davenport. Both are intended to measure how many runs a player is worth compared to a replacement level player at the same position.
There are a few methodological differences that contribute to the differences you see in the numbers. For VORP, I use Marginal Lineup Value (MLV), which is derived from Bill James' Run Created model of run scoring. Clay uses Equivalent Runs for RARP. We also use slightly different park factors, and definitions of replacement level.
David (Canada): Is there anyone now you would compare to 'Nuke LaLoosh' - a 'blow 'em away' lefty who may one day be able to gain control of his amazing fastball & become a dominating pitcher?
Keith Woolner: I'm not watching Bull Durham at the moment, but wasn't Nuke right handed?
Anyways, I'm not the prospect maven at BP, so I tossed your question out to some fellow BP-ers to answer your question: How about Bobby Jenks? Andy Sisco? Ryan Hannaman?
Cris E (St Paul, MN): Where's the next major new ground in stats going to occur: play-by-play? injuries? HS/college evaluation?
Keith Woolner: My guess would be translation of college stats. The focus that "Moneyball" has brought to college drafting has been dramatic enough that there's a lot of potential value to be had in determining whether college stats, properly translated, have the same predictive value as minor league equivalent stats do for major league performance.
Scott Peterson (a different one) (Not Modesto CA): Keith, are there any systematic divergences from linearity in VORP? I guess another way to ask is where in the model things start going seriously non-linear.
Keith Woolner: The Runs Created model is nonlinear (and, in my personal opinion, any good model of team run scoring ought to be, as run scoring itself is a nonlinear process).
MLV captures this nonlinearity by estimating how many runs an average lineup would score by replacing an average hitter with the player in question. This value comes from both the direct impact of the hits and walks he produces, and the added value from the other plate appearances he generates (or takes away) from his teammates by making outs at a lower (or higher) rate.
The actual deviation from linearity isn't tremendous, and linear models are much easier to work with, so my objection to them is more aesthetic than anything else. Within VORP for any single player, the effect of nonlinearity isn't great. But if you extend the framework of VORP to deal with real teams, rather than idealized average teams, the compounding effect on run scoring from adding multiple good hitters becomes significant.
Don (Minnetonka): When are you at BP going to standardize on park effects, level definitions, etc? It's maddening! Pick one and stick with it!
Keith Woolner: I sympathize, and there's a lot of work going on behind the scenes this year to standardize our reports.
Part of the differences, though, comes from genuine disagreements even within the BP team, about the best approach to use in addressing park factors, replacement level definitions, etc. So we're trying to walk a fine line between giving consistency to all the data reports we present, and not quashing the differences of opinion on debatable items that the individual writers and researchers have.
Greed E. Bastard (Park Ave, NY): Keith. Based on the CBA and current salary structure, why does any club bring up a young hitter, rather than letting them age as long as possible, thus having a better shot at their peak years while the club still has renewal rights? If there's no evidence that clubs can actually buy goodwill in terms of a discount when free agency comes, why not slap $300,000 on the player's ass for years 1,2, and 3, and then either lock them up for four or five more years or non-tender them upside the head?
Keith Woolner: There's always a risk in delaying bringing a prospect up -- he may get hurt in the minors (the level of care and the quality of the fields may not be as good as the majors), or the player may become frustrated and develop bad habits or lose motivation. Furthermore, it's always tempting to solve today's problems now, and bring up a player when he appears to be ready to help, even if it means potentially losing some value in the future (such as losing a key year or two at his peak to free agency). You'll see delaying a callup from time to time involving putting off arbitration (2-3 years of service time), but projecting how valuable a player will be 6 years from now (or more if you're considering holding him in AAA for a year), and the time-discounted value of that production makes it a more difficult choice that it may otherwise seem on the surface.
Paul Covert (Lynnwood, WA): Which "Hilbert problem" is claiming your attention most these days?
Keith Woolner: I've been more focused on improving some of the behind the scenes technical infrastructure to improve our stat reports lately than I have on new research (though I hope to do more of the latter as we get into the regular season). But some of my main interests among the Hilbert problems are:
6) Quantifying the value of positional flexibility -- I think that the emergence of two way players like Brooks Kieschnick makes this question especially interesting.
15) Clarifying the win/dollar trade-off preferences for major league decision-makers. -- The only way to really know evaluate whether it was rational or "worth it" to take on salary for a chance at winning is to understand this question.
16) Creating a framework for evaluating trades. -- This is, in some ways, a holy grail of sabermetrics. It brings together not only measuring on field production and contribution to winning, but projectability, replacement level theory, economic preferences of decision makers, and a host of other issues.
Rick D'Alessandrio (Las Vegas): Keith, I'm a big fan of your umpire reports. Have you done any analysis on how much of an impact the umpire actually has on the level of run scoring in the game? What statistical tool would you use to compare the effects of the hitters, the pitchers, the defense, the umpires, the park, and the weather?
Keith Woolner: I personally haven't done any analysis, but I think there was some work done in the SABR Baseball Research Journals a few years back.
If I were to study it, I'd probably look first at something the umps have direct control over, such as ball/strike ratio, and the percentage of pitches swung at (if the ump is calling a wide zone, batters presumably will have to swing at more pitches that they otherwise wouldn't have to avoid taking a strike).
If there was a measurable effect there, then I'd expand the study to look at run scoring, and see if there were any consistent trends. IIRC, there hasn't been a lot of evidence for a large "umpire effect" yet.
Elvis (Graceland): Thank you very much for answering my question, Keith-O. Next time you're down this way, we'll tip back some codeine cough syrup in champagne flutes. My question's about expansion, how long until it's talked about again, and what cities will be #31 and #32? Y'all are great. Love each other. This is called "Burnin' Love"...
Keith Woolner: Thankyouverrrymuch too, your Majesty. And I'm glad the true story of what happened to you has finally been told.
I can't predict what cities will get expansion teams (or even if expansion will even happen in the next N years), but the cities most deserving of a major league franchise are probably Washington D.C. and Detroit.
Greg Tamer (Indiana): New statheads are being brought into baseball, but none from BP. Where is the best work being done today?
Keith Woolner: What makes you think that BP isn't already involved in major league front offices?
We strongly support the trend of increasing use of objective metrics in MLB front offices. We're in favor of better baseball on the field. In terms of BP's professional services with MLB clubs, we do have several engagements with clubs, and we're happy and fortunate to say the number is growing all the time, including the addition of a new club this last week.
However, out of respect to our clients and contractual obligations, we don't focus on those engagements on the web site.
Paul Covert (Lynnwood, WA): I notice that, if you add up the VORP's for all hitters and pitchers, about 54% of the value comes from the hitters. Does that make sense, given that variation in team runs scores and allowed are about equal to each other? (Or does fielding make up the difference?)
Keith Woolner: The variation in team run scoring isn't directly related to the variation of player values above replacement level.
To use a contrived example, if the top 30 first baseman all hit exactly the same, and all the rest of the first baseman (replacment level) were 40 runs below that, the average VORP for starting 1B would be 40.
If half the starting first baseman were VORP=60, and the other half were VORP=20, the average contribution to VORP from a starting 1B is still 40, but the distribution of talent among players is still less.
In any event, a 54%/46% split in a single season from a system that isn't force-fit to a particular ratio isn't really that far different from the 50/50 you might expect if the variance in team run scored, and runs allowed were equal.
Huckabay's Receding Hairline (Stanford CA): How come you guys never mention economic utility functions? They're as important as any statistical analysis of performance in terms of actually using this information in front offices, and I would think you could derive a pretty good approximation of a utility function using available data.
Keith Woolner: Economic utility functions are pretty much what I had in mind in discussing win-dolllar tradeoffs in the Hilbert question earlier in this chat. For a given basket of wins and dollars, a given decision maker has a certain utility assigned to it.
Part of the difficulty is using existing data is the assumption that front offices are being rational in making these tradeoffs. Does anyone want to argue for completely rational decision making in the Bronx?
Marty (Elko, NV): What does TINSTAAPP mean?
Keith Woolner: There
(i.e. developing pitchers get injured or lose effectivness so often that counting on any individual prospect as a "sure thing" is almost folly).
Dr. C (Mobile): What is the one thing we think we know about baseball that really we don't?
Keith Woolner: Why people pay $7 for a beer.
Cory Glover (Los Angeles): Haven't you gotten too far away from real baseball? These numbers are all very distant from the real game. 'Hypothetical, Theoretical, Circumstantial Evidence, Irrelevance, look deep inside, empty."
Keith Woolner: Hypothetical, Theoretical, Circumstantial Evidence, Irrelevance, look deep inside, empty.
Saberfragrance... from Calvin Klein
"Don't feel the love... measure it."
xian (champaign, il): Looking at their catching choices, how many front offices do you think have read your articles on Catcher's ERA or at least have some inkling that "Field General" status might be overrated? Does the possibility of being a post-humously recognized groundbreaker terrify you?
Keith Woolner: The conclusions in "Field General" were so contrary to conventional wisdom that I expect it to take a long time to catch on. Furthermore, it was really the first research to suggest that "game-calling" may not be all that valuable. And for such a conclusion, it might make sense to wait on further research on the defensive value of a catcher before committing to any large scale change in player development programs.
Voros McCracken's thesis on pitchers not being able to affect balls in play much is similarly radical, but can be implemented much more quickly (by changing how pitcher acquisitions are evaluated) and tweaking the offense/defene balance in roster composition. Compare that to taking several years to convert promising hitting prospects to catcher, and hoping that they'll pan out.
Furthermore, the conclusion was really that major league catchers don't differ significantly, not that all players, regardless of their skill set, could play catcher. The speculation of someone like Frank Thomas catching was far fetched (and I think was properly called out as an extreme implication in the original article).
Dusty (Chicago, IL): Keith. Hi. How hard can I safely ride my starters, in terms of a pitch limit per game for each one?
Keith Woolner: Here's my recommendation: After 200 pitches, tug real hard on their arm at the end of each inning to see if you hear a popping sound. If not, send 'em out the next inning.
But to play it safe, better have the bullpen ready as he approaches 350 pitches in any given game.
Mario Van Peebles (Solo, CA): Keith, holding all other things equal, if one hitter has a 25.0 VORP at age 25, and another has a 25.0 VORP at age 23, what's the expected difference in their peak and overall career value?
Keith Woolner: That would be a better question for Nate Silver, and his work on PECOTA. As is, there probably isn't enough information to answer the question. Even after holding position, the proportion of offensive contribution coming from OBP vs. SLG, baserunning and playing time constant, I think there's probably a big difference in projection between two players who post 25 VORP in 200 PA (which is MVP level production) and 25 VORP in 650 PA.
If you really want a WAG, probably a 15 VORP delta at peak, and 75 runs more over the course of a career.
Jordan Lyall (Orange County, CA): Will A-Rod finish out "the" contract with New York? How big of a contract is he likely to sign after he finishes this one?
Keith Woolner: My guess is that he will finish this contract still with the Yankees. IIRC, he still has a no trade clause in his contract, which pretty much means that he can stay in New York as long as he's happy there.
Of course, the events of the past couple of years have shown that's far from a certain thing, if the Yankees revert to 1987-92 form.
Tim (Walnut Creek, CA): Do you think you could talk a little more about baseball, and a little less about undergraduate concepts in econ?
Keith Woolner: This question will not be answered as doing so would not be Pareto efficient.
Bill Bavasi (Seattle, WA): Which of my many mistakes this offseason has been the worst? Letting Cameron go? Signing Ibanez for three times his value? Saving room in the rotation for Ryan Franklin by putting Raffy Soriano in the bullpen? Offering Pat Borders arbitration? My hands are getting tired, so I'll stop there.
Keith Woolner: I'll take "Offering arbitration to 40 year old catchers who played 12 games last year" for far-too-many thousands of dollars, Alex.
(Oh, but I forgot that he got in 2 games at third base, too. See, we need that multipositional value research!)
Hank Kingsley (Look-Around Cafe): Hey now! Which is a better expected return? Vernon Wells to win the MLB HR Title at 500-1, or KC Royals to win the AL Pennant at 28-1?
Keith Woolner: Wells.
Reede (Boston MA): Hi Keith. On the radio, Gary said that the Red Sox still looked like a better team than the Yankees for this year, and that the division would be very close. Do you see it the same way?
Keith Woolner: Keeping in mind that, as a Red Sox fan, I may have a teenie bit of bias, but I can state unequivocally that *this* *is* *the* *year*.
One more question..
Ryan (Dearborn MI): In BP 2004, I enjoyed your analysis of the tactics of stealing bases. The main problem with it is that you're never actually running on "an average catcher". Each pitcher/hitter dyad is different in terms of both their ability to prevent the stolen base, and in terms of the risk/reward calculation involved. Against a Terry Mulholland that's impossible to steal off of but who can't pitch real well, the calculus is different than if you're facing Pedro. Don't you need different decision trees for every pitcher/hitter combo on the opposing team in order to make an informed decision?
Keith Woolner: Thanks for the comments about my article. Yes, the individual situation faced by a runner is different, but that doesn't mean that we can't use aggregate results to get useful knowledge. The same could be said about the expected runs base-out matrix first popularized by Pete Palmer in "The Hidden Game of Baseball"
The expected probability of success certainly depends on the battery in the game at the time. Of course, the pitcher on the mound also has an impact on how valuable a successful stolen base would be (runs are scarcer against Pedro Martinez than Jeremy Bonderman), and thus the marginal win value of a run changes too.
But aggregating over every steal opportunity faced by a runner evens out the differences in individual opportunities. Not completely, not perfectly, but enough so that we can determine whether what the stealing propensity faced by a given catcher was.
Keith Woolner: Thanks for the questions everyone! We'll be posting the schedule for the next chat session soon!