Happy Thanksgiving! Regularly Scheduled Articles Will Resume Monday, December 1
January 25, 2010
The following is an edited transcript of an in-house discussion among the Baseball Prospectus team about a hypothetical pitcher capable of delivering a guaranteed performance.
Kevin Goldstein: Hi everyone. I was having a talk with a front office guy today, and after talking business, things got a bit philosophical.
Let's say there was this right-handed pitcher, Joe Consistency. Every time out, Joe is going to do the same thing. He's going to go exactly seven innings, and give up exactly four runs. All of his rate stats-hits, walks, strikeouts, home runs, ground balls, platoon splits, whatever-are pretty much what you'd expect from a guy with a 5.14 ERA. He's never going to get hurt. He's going to take the bump 32 times, and throw seven innings every time (224 for the season), and give up four runs (128) every time.
He's a robot, and he's looking for a three-year deal before moving back to his home planet, and that performance is guaranteed for those three years. What's that worth on the open market?
Matt Swartz: Well, a 5.14 ERA, if seven percent of runs are unearned, which is about average, corresponds with a 5.50 ERA. I don't know what league we're talking about here, but 5.50 RA seems really, really close to replacement level, so I'd say it's pretty close to zero. When the PECOTAs come out, check out how many fifth, sixth, and seventh starters are right around there. It really is just about replacement level, so I don't think I'd guarantee him a contract. You can wear out a new Rodrigo Lopez every fifth day and get four earned runs in seven innings on average. I'm not sure that there is positive value on the guarantee at all.
But if we're trying to go for 5.14 RA, then that's different. A 5.50 run average over seven innings is 4.3 runs every time. If you give up 0.3 runs a game for 32 games, it's 9.6 runs. That's about one win, and teams pay about $6 million a year for that on the free market (ironically, considering Bil's e-mail as I type stuff).
It really depends on whether you mean RA or ERA and whether you mean AL or NL, because I can see that this Joe Consistency is worth somewhere between zero and one win, which means he's worth somewhere between $0-6 million, which covers 99.9999 percent of the nation's salary range.
Colin Wyers: Good call on the league question. The difference between the AL and the NL is typically about .5 runs per game on average (ah, the wonders of the DH).
Factoring in a normal distribution of unearned runs, it should be a 4.73ish ERA. Of course, if all of his runs are earned runs, that simplifies things a bit.
Regardless, I'd pay him a bit less than a typical starting pitcher with those numbers. Consistency is really only desirable from a starting pitcher if he's consistently good, at least as far as his rates are concerned. Typically, a pitcher who is less consistent will put his team in a position to win more often than a pitcher who is equally good but more consistent. I know that's counterintuitive, so I'll walk through it real quick.
The relationship between run scoring/prevention and winning isn't exactly linear. In other words, some runs are more important than others. As you go from average toward zero, each run becomes more valuable. As you go from average toward Shawn Estes, each additional run becomes less valuable. By being inconsistent, you're trading a number of middling starts for a couple of very good starts, starts in which your team's chances of winning are better than you would predict by treating all runs equal. You're also trading some middling starts for some disaster starts, of course. But the difference in value between those very good starts and those middling starts is greater per game than the difference in value between those middling starts and those disaster starts.
So, on a per-game basis, he's actually less valuable than his 5.14 (E)RA would imply, as odd as that may sound. Now, he's still taking the mound every fifth day and pitching deep into each game. Is there any value to that, above and beyond his poor rate stats? Some. That's actually precisely the question that something like WARP was designed to answer. What's a bit murkier is the question of bullpen chaining and saving wear and tear on your middle/late relievers. Just thinking about bullpen chaining makes my head hurt again.
KG: Both the guy and I (admittedly not guys who don't understand the numbers as well as you, and we may be wrong here), put a lot more value than you guys seem to on the consistency. In the most general sense, he never gets ripped, and keeps you in every game. The other aspect was the fresher bullpen. My gut says he's worth more than your average 5.14 RA (yes, ERA was the wrong number there) guy, but it looks like the reality disagrees.
Tommy Bennett: I think, on average, he would be worth the same as a 5.14 RA pitcher. However, I have to follow the Swartz/Wyers position here, since inconsistently bad starters are worth more than their consistently bad counterparts. Additionally, I think he would be worth more to a high-scoring team and less to a low-scoring one.
A final thought is that he would be worth more in relief than as a starter. Since his RA is guaranteed, a manager could deploy him to lock up a game with more consistency.
On the other hand, doing a Baseball-Reference player index search of guys with ERAs in the 4.7 to 5.2 range since 2000, picking a random few out who seem to sort of match to check on their WARPs, you've got Ryan Dempster's 2001 (4.94 ERA in 211 innings equaling -0.5 WARP), Jeremy Guthrie's 2009 (5.04 ERA in 200 innings equaling 2.0 WARP), Josh Beckett's 2006 (5.01 ERA in 205 innings equaling 2.4 WARP), and the first guy who came to mind, Livan Hernandez's 2007 (204 innings with a 5.11 ERA equaling 2.5 WARP). None of those guys reached seven innings per start, but you can see that, based upon ballpark and league context, there's the potential for a value of a couple of WARP above replacement, which is worth real money.
Steven Goldman: Tommy raises a very good point, in that the value has to be determined by context. If you have a top offense, the idea of locking in that RA-knowing that it's going to be sufficient for you to win on most nights-is perhaps worth a premium. The 1976 Reds, just to name one example, won 102 games getting about a third of their starts out of guys who were around the replacement level, but they also had a .290 team EqA. If you have the 2008 Giants' offense, his consistency is actually a curse. A less consistent pitcher with the same RA will throw the odd low-scoring game between beatings, giving you a chance to win. With this guy, you lose almost every time out.
Ben Murphy: Seems like it would also depend on the composition of the rest of the pitching staff in addition to the offense. If you've already got a bunch of consistent, boring, average guys, then he's probably less valuable on the field and taking up roster space than if you've got more volatile pitchers and the robot consistency can help you recover when the wild child throws up an ugly line.
If this is a team-specific question, we can put a much finer point on the answer than if it's a generic "oh about $4-million-per" type of thing. You'd be able to look at expected performance distributions like Colin and Matt touched on and see how the robot profiles as a potential complement to those other guys, instead of having to approximate value based on the league averages.
MS: I think it's closer to what you wrote about when you checked the persistency of FLAKE and found that within-year consistency was inconsistent with respect to multiple years. I don't remember the details of your findings there, though, beyond consistency being inconsistent.
But in terms of the value of consistency within a game, I actually don't know that a consistent pitcher is actually a good thing, since you'd probably do better by concentrating your runs all in one game. Think about it like this: The average pitcher with a 5.14 ERA probably gets blown out a lot, right? Maybe his starts might look like this:
A: 2 runs 7 innings
That would be 14 runs in 25 innings-a 5.04 ERA-which is pretty close. But that team probably wins about 60 percent of starts A/B/C and 5 percent of start D, so that's a 1.85-2.15 record on average for a .463 winning percentage. What about a starter who gives up four runs in seven innings four times in a row? Each start is probably less than 46 percent.
I think runs are more likely to be clustered together than spread equally between games, so even bad pitchers probably have distributions that are skewed in this way. So I think consistency has a negative value for starters.
What if we just looked at the correlation between FLAKE and SNLVAR in different FRA buckets of 2.00-3.00, 3.00-4.00, 4.00-5.00, and 5.00-6.00? Are they positively correlated in all four buckets? In some? That would answer the question, wouldn't it?
ES: The gist of the findings in my article were that FLAKE had a very low ICC, meaning it wasn't consistent for the same pitchers over a four- or five-year span (and with certain sample sizes of innings, etc.), making consistency itself inconsistent. Additionally, consistency did not correlate strongly to any win-based metric, which jibed with Michael Wolverton's idea that the inconsistent pitchers can actually add more to the championship dreams of a team.
Consistency is valuable, as I wrote, in the sense that execs, managers, and fans need not pull their hair out or buy a yearly supply of Mylanta-which makes sense why Kevin's contact would be really into it-but it doesn't make a guy a better pitcher for exactly the same reason Matt stated.
If you have two pitchers, each of whom makes five starts, going 36 total innings with 18 runs surrendered, but break down like this:
P1 is likely better, even though his FLAKE would be worse, because when people value consistency, they want consistent goodness. They don't want The Livan Special. They want the 7 IP/3 R performances over and over again.
Guys like Kevin described fit that Jeff Suppan mold, guys who could certainly fill out the back end of a rotation, but are bound to be overpaid based on a perceived quality that is deemed to be more valuable than it is.
Clay Davenport: Assuming that "consistency" and "flakiness" are properly defined, we should find that:
So this pitcher you describe is somewhat below average, and perfect consistency is not the ideal trait. It is true he won't get you blown out of any games, but he won't win any "by himself" with an outing of one run or less, either. It is good enough to earn positive WARP; the replacement level for a pitcher is around 5.5, and I do think at times it should be around 6.0. On the order of 19 runs above replacement, that's roughly 1.9 WARP. That's worth $3 million-$5 million, assuming that it's $2 million/WARP +/- 0.5.
But seven innings every time out is going to be pretty helpful to the staff as a whole. It isn't just the resting the bullpen side-which I don't think has been quantified well-but in letting the manager go to the bullpen a little earlier than he might in the next day or two (rather than sticking with a struggling starter because the bullpen needs rest). That's one I've wanted to tackle for a while.
Russell Carleton: Y'all are being far too rational. I guarantee this guy/alien/robot would get a "what were you thinking?" contract above $10 million a year (cf. Meche, Gil; Silva, Carlos). The key word in there is "consistency." People will do really illogical things (like pay extra) to not worry. A lot extra. It's not hard to see that this guy is below average, but people will be blinded by the comfort of knowing.
Consider the following thought experiment (stolen from the book Iconoclast by Eric Berns): The fate of the universe/your life/a chocolate donut/a date with your favorite celebrity (whatever gets you going) is dependent on you pulling a white marble out of one of two urns, which are hidden behind a curtain. You know one urn contains 20 white marbles and 20 black marbles. You know the other one contains 20 white marbles, but you don't know how many black marbles. Which urn would you pull from? Keep that answer in mind.
Now. Suppose that the universe/your life/that donut/a date depends on pulling out a black marble from the same urns. Which urn would you choose? Let me guess. You picked the 20/20 urn both times. Or you were at least tempted to. It's illogical to do so. If you picked the 20/20 urn for the white marble, you are implicitly saying that you think/guess that the unknown proportions urn has more black than white, so your odds are better at 20/20. So, when it came time to pick a black marble, why did you go back to the one that you implicitly said had more of a chance of landing on a white marble (comparatively)? Because you hate uncertainty. It's a human failing.
There is also the issue of "seven innings." That actually has some value, both real and psychological. Seven innings guaranteed means that you only have to get two innings out of your bullpen, which can be concentrated on the good relievers. An extra inning of work by the starter buys not only three outs, but also an inning of not having to use one of the lesser relievers in your pen (so your anticipated "bullpen ERA" is based off a better group of guys). It may also buy not having to tax your good relievers for an extra batter or two, and a few extra pitches. Psychologically, there are probably really only two or three relievers that managers really trust in each pen. The other guys are just too unpredictable. Now, imagine that I could guarantee that you wouldn't have to live through that.
Now I want a donut.
ES: Eric Berns sounds like a cheesy alias I would use ... or my porn name.
David Pease: Yeah, no, I picked the 20/20 the first time, the 20/n one the second time. When n (black marbles) is a random number, there are a lot more numbers above 20 than there are below 20. Then again, I thought it was silly that millions of people in the DC area felt paralyzed by a sniper that shot a few people over a few weeks.
MS: I might not really understand the rules of the game, but it seems that people have assigned a probability distribution to the uncertain urn that isn't really clear. It could very well be that their choices are smart given their assumptions. Urns have finite space and they might have assumed that the experimenter went cheap on the experiment, figuring that you only need to buy 60 marbles to conduct this experiment. I say that somewhat tongue-in-cheek, but I think that most people have some inherent thought process where they think that it would be really silly for the experimenter to buy extra marbles and mentally skew the distribution towards fewer than 20 marbles.
I also think that markets are much more rational than you do, but that's maybe an issue of Psych vs. Econ backgrounds (it's a PhDuel!). I don't think everyone is rational, but I think that, on average, people are rational with several notable exceptions that you can find in behavioral econ literature. When businesses have sales and discounts of various kinds, they are generally taking advantage of known irrationalities, but that's far from the case here, since the pitchers aren't intentionally offering tricky discounts, etc. I think that in this case, markets might not be so irrational. Do GMs really value consistency? I would like to see some evidence of that, because I'm not sure it's true, and even if it were, I'm not sure that it's irrational because it could very well be that the GM signs the sure thing so that people can't blame him for signing Oliver Perez and call for his head. It's the same reason managers use the sacrifice bunt and follow "the rules" so much. They aren't necessarily irrational as much as they are avoiding losing their job. As far as Meche and Silva, I don't really think of either as being inherently predictable. I just think that the value of a win is so high that these deals are worth it on average-Meche worked out well, Silva didn't.
I do think baseball is one of the kinds of industries where markets wouldn't be perfectly rational, because there is no firm entry/exit, and the consumption value of owning a team is idiosyncratic enough that the owner who can run the team better won't necessarily buy it, so you don't have the same kind of Darwinian-type market corrections that you would see if firms could be knocked out of business if they behaved irrationally. But I'm not sure that you can make the claim that this is really an example of irrationality, because I don't really see consistent evidence of teams overpaying for consistency. It doesn't seem like one of the traditional examples of imperfect markets, as long as the pitchers can't control their consistency very much.
RC: The point isn't the process by which people assign a probability assumption to the "mystery urn." You could use a reasonable method like you did (makes as much sense as anything else, given the available info), or just guess. The point is that when you do this experimentally, most people pick the same urn both times, and specifically, they pick the one with the 50/50 shot. It's impossible for the same urn to be the better chance on both white and black. It's telling, though, that people pick the known quantity as well.
I don't have direct evidence for the fact that GMs (over)value consistency in any systematic sense. I can tell you that humans in general overvalue not feeling anxious, and that the unknown is a great provocateur of anxiety. There are other situations in which baseball teams systematically overvalue certainty (my third base coach example) and have other motives (blame avoidance l ... the Perez example is spot on). It's not much of a stretch. Consider how often you see a team with a promising youngster who's MLB-ready, but unproven, who is unseated by a clearly inferior free agent that the team signs because "he's got a track record." It is circumstantial evidence, but I think it's a good case.
As far as the market goes, I'd agree that if there were a large number of baseball teams, the inefficiencies would probably wash out in the aggregate. Individual humans have varying degrees of tolerating uncertainty, though, and in a small sample (30 GMs), you just need one outlier to get a Silva contract, especially on the quasi-auction model that is free agency.
Ken Funck: I agree almost entirely with what you're saying, Russell. A GM will definitely place value on the certainty aspect of the robot's performance, whether that's rational or not, just because that's one less uncertainty to worry about. And I understand the point being made with the "mystery urn" experiment, but I'm not sure I agree that the reason most people will choose the 50/50 urn both times is illogical and caused by "people hating uncertainty." I think it could also be risk-avoidance. Close to uncertainty, but not quite the same thing.
If the consequence of picking the wrong marble is that I don't get a chocolate donut, I'm more likely to perhaps guess which urn has more of the correct marbles and stick with that guess, since the consequence of my guessing wrong is merely not getting a donut (sad as that may be during lunchtime). If the consequence of getting it wrong is universe destruction, or being fired from my job running a major-league franchise, then I'm more likely to take the known 50/50 chance. Putting Joe Consistency in the rotation means I avoid the risk of having someone worse out there. I'll never have to explain why I put my chips on young Teddy Unready for my rotation, and suffered through his terrible starts. Instead, I've essentially chosen to pay to be certain I avoid the worst possible outcome. In the process, I've also made certain I avoid the best possible outcome, but as others have said, depending on the team and the circumstance, that might be a logical choice.
Will Carroll: I think Russell's next article has to involve this and a discussion with a couple of GM-types to see if perhaps there is (and I'm almost sure there is).
Rany Jazayerli: This might be too obvious to warrant mentioning, but given the unique nature of this player, I think the value is highly dependent on the team's existing roster, and their other options.
Let me put it this way: You sign Mr. Joe Mediocre (I prefer the name "Kevin Gross") to a three-year, $30-million contract. Then you find out that Gross has a twin brother capable of exactly the same level of performance. Do you sign him to the same deal?
I think the answer to that is clearly "no." Otherwise, taken to the extreme, you could sign five Kevin Grosses, and suddenly you have a below-average rotation that you're paying $50 million a year for. Even given the money you could save in the bullpen since you'd only need about 325 relief innings, that's not cost-effective.
The point is that the replacement level changes with each new Kevin Gross you add, and because the value of this pitcher is highly sensitive to the exact replacement level RA, that value is going to change significantly based on the team's existing roster. The Atlanta Braves wouldn't want to sign this guy for $5 million a year, but a team with three or four good starters but a hole in the fifth slot such-the Dodgers? Cardinals?-would probably fork over $10 million per in a heartbeat. Well, they would if Frank McCourt had any money.
ES: My head has been spinning lately so forgive me if I am mentioning something already stated, but a player with a 5.50 RA and 5.15 ERA just is not very likely to get 224 innings in a season. There may be true talent guys of that ilk who have better years-like Colin pointed out their ERA/DERA discrepancies-but you don't want 224 innings of 5.14 ERA pitching. At least, you don't want to go for that on the market. As Matt said, there are tons of Rodrigo Lopez-types available for $600,000, so why you would pay $10 million for that type of pitcher is beyond me.
It seems as though the choice here is between the Unrealistic Robot, since 224 IP/5.14 ERA is not really desired, or cobbling together 224 innings from Sidney Ponson, Bruce Chen, Rodrigo Lopez, and Eric Milton, which would cost 25 percent of the robot's fee. I'll go with PonChen MiltLopez every time.
KG: From a front office official: "I think there is some value in it, but for me it's pretty limited. I'd far rather sign a guy who at least has a chance to make an impact. When you are trying to win a championship-and, in theory, we are all trying to do that-you have to roll the dice. You can find consistency-that guy is out there-but there's a reason Jon Garland isn't signed and Rich Harden is."
Gary Huckabay: This runs right into a number of x, y, z twisting singularities of the baseball player market.
First, I want to answer before reading the massive number of responses, so forgive me if this ends up a little bit redundant on some fronts:
It's an interesting question. Intuitively, it seems like someone's going to get slammed with the winner's curse on Robo(Knuckler? Livan? Garland?) I think that someone will give him three years and $24-30 million. I just wouldn't want it to be me. I think his real value to a team is way lower than that. Way lower. I don't think I'd bid for his services at all, or I'd value him at very close to the league minimum. The whiff of "completely replaceable" is pretty strong. Some risk-averse GM would pay him way too much and cripple his team.
KG: Two front office people have brought up Garland, and one scout replied "So can I call this robot of yours Joe Blanton?"
ES: Heck, I don't even know if calling him Garland is accurate, let alone Blanton, who at 200 innings and a 4.30 ERA or lower is very valuable. The guy we're talking about now is like Suppan, not Garland or Blanton.
GH: The really interesting and compelling question is, "How much do you spend for the 15-percent chance to find the pitching coach who has a 20-percent chance to improve his performance by .05-.30 ERA, and a five-percent chance to degrade his performance by .02-.15 ERA?"
Christina Kahrl: I guess I'm amused by how quickly people have made some time-wasting inferences. Kevin made the point at the start: four runs allowed, period, and seven innings, period. The rush to insert distinctions over whether there are additional unearned runs allowed is just so much self-assigned make-work. Leave it to the official scorer to decide what's unearned from among those four runs, since that's too often a matter of opinion anyway. The posited theoretical seemed pretty cut and dried: four runs, seven innings, every time.
Nagging facts: Chen hasn't allowed less than six runs per nine in the major leagues since 2005. Lopez has bested the robot exactly once since 2005, in his injury-shortened Rockies' season; and he couldn't do it over a full year. Eric Milton? He's "beaten" the robot once since 2005 as well, and that was in last year's whopping five-start sample. Ponson hasn't beaten the robot since 2003; talk about sure things. The only relative guarantee is that their track records make it clear that they'll do worse than the robot, by a run per game or more, and probably worse if you try to get as many innings out of them as you know you'll get from the robot. I don't imagine keeping Ponson shackled to the mound for seven innings isn't going to keep the opposition to under four runs scored, certainly, but I certainly agree with a suggestion that those are the kind of entirely fungible fifth-starter types you'll find floating in the free-talent pool. The problem is that they're virtual locks for failing to fulfill both halves of the proposition, because they can't deliver on the workload or the RA/9; that's why they're making around the minimum.
In my pedantry, I figure it might be more useful to look at who had between 5.0-5.5 RA/9 last season, and 30 or more starts, just for the sake of argument and comparison:
5.0-5.5 RA/9: Derek Lowe (5.04 RA/9), Brad Penny (5.30), Carl Pavano (5.37), Joba Chamberlain (5.38), Jeremy Guthrie (5.40), Ricky Nolasco (5.40), Mike Pelfrey (5.47), Livan Hernandez (5.49)
Obviously, just about all of these guys underwhelmed the expectations attached to them. That's why they got 30 or more starts, out of the expectation they'd do better than the robot. To a man, they threw less than seven innings per start, and you could expect that forcing them to try would lead to worse RA/9 marks.
For a point of comparison, how about the fleshy analog who comes to mind to the mechanical man, Jon Garland? He's been at 4.68 RA/9 in '09, 5.31 in '08, 4.92 in '07, 4.77 in '06. He's better than the robot, and he's going to make a lot more than, say, Hernandez. I'd probably use Hernandez as my guide here, because of the durability-related part of the proposition. The Twins gave him $5 million for 2008 on the basis of his robot-like performances in 2006 and 2007. Like the robot, signing Livan involved very little upside.
Depending on team context making him useful to you at all because you need the innings and have an offense that can make four runs stand up more often than not, I'd stick with that $5-million mark for Livan after 2006-07 for the sake of argument. Happily, you can safely add incentives to beguile his agent, like $5 million for making the All-Star team or another $5 million for the Cy Young, because you're safe in the knowledge that RoboPitcher won't earn those distinctions.
KG: An assistant GM chimes in. I think this guy speaks well to the devil you know, risk-averse thinking Russell discussed:
First of all, I would not expect someone who gives you 224 innings a year to have an ERA of 5.14. I don't think you would keep running him out there every fifth day with that performance. Nonetheless, the reliability in terms of innings and 'performance' does have some value and allows you to plug in the rest of your rotation knowing you have one piece of certainty. Somewhere between $5-7 million, only because you have guaranteed me he will not get hurt.
JJ: And just like that, the news breaks that Blanton gets three years and $24 million, a breakdown that see him make $7 million in 2010 and $8.5 million in both 2011 and 2012 after he made $5.47 million last season. Buying out his first two years of free agency suggest that Gary's three-year, $24 million-$30 million estimate for our robot is too high, given what should be a clear differentiation between him and the slightly above-average workhorse Blanton.
Meanwhile, Vicente Padilla got one year and $5.025 million.
ES: As a Phillies' fan, I am perfectly fine with three years and $24 million for Blanton. Blanton is not the robot.
Guy Worse than Robot: ChenSid MiltLopez
KG: Supposed robot? The Robot is real! Love the Robot!
ES: For the record, "supposed robot" should be read as if I'm impersonating an impersonator of Johnny Cochran.
CK: This is where I ask Clay if we know who in history would be the most similar to the robot.
KG: Can we really know without going by game logs? I really think the consistency makes him very unique.
CK: True. We've got almost 40 years of complete game logs. I guess my anticipation is that: Robot:Don Sutton::Trabant:Corsica
CD: Without getting into game logs (those are going to be messy across eras), I went looking for players who exemplified the robot with their season totals. Using the translated pitching stats, so that context could hold up relatively well across time, I looked for guys with between 200 and 250 innings for the season (3625 pitcher-seasons), then looked for ERAs between 4.50-5.00 (which got me to 872), and then looked for pitchers who made the list most often.
There was a tie for the top spot, with six. I declare the winner to be the pitcher who did it in six consecutive years, from 1927-32: Freddy Fitzsimmons.
Second place goes to the frequently-mentioned-in-this-thread Jeff Suppan, who turned this trick six times over a nine-year period-four straight from 1999-2002, then again in 2006-07.
Other entrants were Bob Groom (1909-10-11-14), Earl Whitehill (24-25-30-32), Don Sutton (67-79-83-86), Bob Forsch (77-78-82-86), Todd Stottlemyre (90-91-95-96), John Burkett (90-93-95-98), and Brad Radke (95-96-03-05).
TB: OK, I'm going to throw out another random thought: This guy would probably increase attendance, and so his value to the team might not be captured by his value on the field. All we have to do is assume that if the expected margin of victory between the two teams is smaller, more people will want to buy a ticket.
Our Joe Consistency gives us that smaller margin because he limits the variance on the RA side of the equation. In the normal case, two variables are random, so the distribution is wider. However, with this guy on the mound, RA is fixed (at least for seven innings) close to the league average (probably a little above, but it depends on league factors). I'm not sure exactly how big the effect would be, though you could simulate it, but I think it'd put more people in the seats.
MS: Wow, I picked a bad day to be busy. This is a great discussion. I just read through all of the last 20 e-mails since my last one. I thought a few things were interesting. Specifically, I found it interesting that a team was using the $4.5 million/win bench line, since that number has been pretty much shown to be bogus. It came from FanGraphs looking only at the first year of deals and looking at projections. Sky Andrecheck showed that teams were paying close to $6 million/win over the course of the whole deal, and I gave approximations of WARP values in different stages of multi-year deals. Sean Smith did an article where he basically realized that the first year of deals were close to $5 million/win, escalating each year. I don't remember the details but it matched up with what I would have expected given what I had done and what he had done). Smith also showed that the $4.5 million win coming from projections was flawed because projection systems were too optimistic about the wins part, so they overstated $/win. But it seems teams are using the sabermetric results now, which is actually an error. That's very interesting-the agents are now aware of the model. I think it might correspond just fine with the winner's curse anyway, so that the winning team thinks they're paying $4.5 million/win, but given that they overvalued the player, maybe it's actually $6 million/win.
And also wow to anyone thinking Blanton is Joe Consistency in any way other than his first name. Blanton is a very good pitcher at this stage. He's basically transformed himself since coming to the Phillies, way more than league adjustments would suggest. He's increased hitters' swing-and-miss rate by a factor of like 30-40 percent. It's a huge difference.
JJ: Let's not forget that Blanton has been spotted with what was believed to be pine tar on his cap. It was a big hubbub in the 2008 World Series, though not as big as Kenny Rogers in 2006, and during the 2009 National League Championship Series it was apparent. Given MLB umpires' lax enforcement in this area and the prevalent assumptions about the Dave Duncan effect in St. Louis, I don't think that we can rule out that he's gaining something from dabbling in the black arts. Not that I really have a problem with it. Some of my favorite pitchers went that route in their latter years, and of course Gaylord Perry was a treat to watch.
KG: From a prominent agent: "A lot for an AL club. Guy who takes the ball and gives you innings and keeps long to middle pen away from field is a valuable guy."
From a front office official: "That's a pretty valuable guy in my mind, particularly the part where you guarantee he's not going to absolutely suck at some point during the contract and become a complete albatross. I'd think that your robot gets $5.5-6 million per year. Of course, now you have to tell me what other answers you got so I know if I'm crazy one way or the other.
"I expected there would be an argument that says there's a big problem because, 'this guy's never going to throw a shutout and help you win a 1-0 game even if you're OK with him giving up eight some day since you could still theoretically win 9-8.' Don't underestimate the value of him never getting blown out after 1.1 innings and burning your bullpen for the next three nights. That can kill a team for way more than just the one blowout that everyone sees."
WB: Sorry, but this one seems a little hard to accept. If your starter is wasted in the second inning, you aren't likely to use your best four bullpen guys in that game because you've just been blown out, and with seven bullpen pitchers, that leaves three others to mop up. Additionally, the ease of swapping that seventh bullpen pitcher back to Triple-A, presumably because he is too tired to pitch for two or three days, for the eighth-best reliever on the 40-man seems like very little drop-off. I don't buy it that the bullpen is more "spent" in this situation. I think you could even argue that you don't go to your top four pen pitchers in the second even if your offense has kept you in the game. Most managers would seem to wait to see if they are still in the game around the fifth or sixth. Finally, one could argue that the starter now doesn't need a full four days off to come back and pitch again.
Based on current usage trends, I would say that the more damaging pitcher for the bullpen would be the guy that is "adequate, but can never get through six" and always "forces" the manager to burn four relievers "Ozzie-style" to finish the game.
BM: I think maybe that guy's name is Scott Kazmir (the one who is adequate most of the time but never makes it through six)?
KG: Then you have guys like Sid Fernandez, who were great most of the time but couldn't give you more than six innings.
WB: Kazmir or Sid may be good pitchers that are welcome on nearly anyone's staff, but if you are trying to determine their effect on their team's bullpen (and assuming modern bullpen roles of one inning per pitcher), then it takes Kaz/Sid plus four more pitchers to finish a nine-inning game. That should be more difficult to sustain than the pitcher that goes eight innings one night and two inning the next time out. The front office guy in Kevin's post suggested that the occasional two-inning guy harms you in the present game plus two or three more games after that. "That can kill a team for way more than just the one blowout that everyone sees."
Back to Kaz/Sid, the "right" answer is probably to have a guy in the pen who can always go two or three innings after the Kaz/Sid pitcher, almost like paired pitching in the minors. This would "deliver" the game to the closer (again, assuming modern usage of one IP closer). This would be an acknowledgement of a known limitation and providing a framework to handle the situation, as opposed to continually asking Kaz/Sid to do something they don't seem to be able to and then being frustrated by their "failure." I obviously haven't coached professionally, but I have certainly seen several amateur coaches who just cannot see a player for what he is as opposed to what the coach wants him to be. Often it seems as though MLB managers suffer this same limitation.
KG: This is from a general manager: "Great question. Two or three years ago, I'd have said three years at about $8 million per year. The mid-4 ERA guys were getting $10-11 million per year. Today, Blanton signed for three years, $24 million. Hometown discount and one arbitration year factored in, but it seems to be a pretty clear market shift, so I'd say $5-6 million on the open market for your guy. That level of performance probably gets more like $4-5 million for one year, but the guarantee (of health) probably gives him a small bump, and justifies the multi-year deal. A level below Garland. That's quick back-of-the-napkin math. I'm sure I'm missing something."
From a front office official: "Interesting question. Frankly, I'm not sure what he's worth. Obviously, he's not worth much to a team that can't score runs (say the Giants or Mariners last year), since they'd just lose anyway. He's got some value to the 2009 Yankees or Red Sox or Angels, since they'd probably win more than half his starts. Of course, there are a lot of pitchers who can give you a RAA of 5.10 or so-don't suppose it's a coincidence that you chose a pretty good approximation of replacement level for his performance?
"So I guess the question is: What is the value of certainty? Certainty has more value to teams in the playoff hunt than teams who need everything to go right to get into the playoffs. For example, the Pirates won't make the playoffs by signing certain mediocrity. They need some of their mediocrities to get a good roll of the dice. But I could see the high-scoring teams with playoff expectations valuing the certainty fairly highly.
"Some teams would probably not even be interested in the guy while others might value that certainty at several millions dollars per year. I don't know. He might get three years and $21 million or so if he were a free agent. I wouldn't be surprised if he got more or less, really. It's hard to value a machine!"
DP: The problem with this guy being interesting to a playoff team is that he's close to useless when one actually gets to the playoffs, right?
RC: Not to besmirch the front office folk whom you've drafted into this conversation, but this does run into the "merlot problem." The estimates that they are giving are in line with a calm, rational analysis. Not surprisingly, they've all come around to a pretty good consensus on what the market will bear. It's the kind of leisurely question that you can answer while lingering over a glass of merlot.
In reality, what people say they will do and people will actually do under stress/emotion/pressure is very different. The areas of the brain that handle rational thought (front of the brain) and that handle emotion (middle of the brain) are separate places. We have all made bad decisions when we "weren't thinking." (Please, spare us the stories...). I wouldn't go so far as to say that suddenly this guy would be deluged with four-year, $72-million offers, but if an agent smells fear, he might pull an extra million or two out of someone per year. It only takes one guy to buckle and suddenly, you have a really silly contract that "sets the market" for other similar players ... er, robots.
WC: Bees and Boras can smell fear. The human head weighs eight pounds.