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?

William Burke:
I believe that Jon Garland made $6 million+ last year.

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.

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.

Jay Jaffe:
The guy ain’t gonna win a Cy Young, and really, we’re talking about a hypothetical here, but seven innings per start-over 200 innings in a full year-doesn’t seem like hay.

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.

Eric Seidman:
Didn’t I write about this yesterday?!?

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

B: 2 runs 7 innings
C: 2 runs 7 innings
D: 8 runs 4 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?

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: 6 IP/5 R, 9 IP/0 R, 8 IP/1 R, 5 IP/9 R, 8 IP/3 R
P2: 6 IP/4 R, 7 IP/4 R, 8 IP/4 R, 7 IP/4 R, 7 IP/2 R

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:

  • For an average pitcher, it basically doesn’t matter;
  • For an above-average pitcher, consistency is more valuable than flakiness, and more so as you get farther and farther above average;
  • For a below-average pitcher, flakiness becomes increasingly prized

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.

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.

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.

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.

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.

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 going to be highly team-dependent to figure out who’s going to end up who has the highest value for this dude. In many ways, it’s to the point where listing them would be silly (offense, defense, bullpen, rotation, health, etc.).
  2. The 5.14 ERA is an interesting point. Inflection-tastic. The RA‘s going to be just a bit higher-call it 5.50 for ease-which makes it possible that this guy is right at the tippy-tip of a massive potential drop off in value.
  3. The baseball player market is filled with weirdness. It’s highly discrete, and isn’t really a single market, but rather a bunch of interacting potential transaction potentials. And dynamic to the hilt.
  4. AL or NL?

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.

Two front office people have brought up Garland, and one scout replied “So can I call this robot of yours Joe Blanton?”

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.

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)

Worse than 5.5 RA/9, 30+ starts:
Braden Looper (5.69), Jeff Suppan (5.90)

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.

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.

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.

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

Guy Who Embodies the Robot:
Livan Hernandez

Guy Closer to Average that Doesn’t Embody the Robot:
Jon Garland

Guy Above Average:
Joe Blanton

Let’s not just mix these guys up. Blanton is a 190-200 IP, 4.00-4.30 ERA pitcher in the NL, which is markedly better than this supposed robot.

Marc Normandin:
But not markedly better than, say, Cliff Lee.

Supposed robot? The Robot is real! Love the Robot!

For the record, “supposed robot” should be read as if I’m impersonating an impersonator of Johnny Cochran.

This is where I ask Clay if we know who in history would be the most similar to the robot.

Can we really know without going by game logs? I really think the consistency makes him very unique.

True. We’ve got almost 40 years of complete game logs. I guess my anticipation is that: Robot:Don Sutton::Trabant:Corsica

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.

Five closely packed entries go to Sad Sam Jones, who did it in 1919-20-22-23- and 25. Five spread entries for Andy Pettitte, in 1995, 2000, 2006, 2008, and 2009.

Among the nine guys who did this four times, one did it in four consecutive years-Scott Karl, from 1996-99. Another did it within a five-year span: Allie Reynolds, in ’45, and ’47-49.

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).

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.

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.

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.

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.”

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.

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)?

Then you have guys like Sid Fernandez, who were great most of the time but couldn’t give you more than six innings.

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.

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!”

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?

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.

Bees and Boras can smell fear. The human head weighs eight pounds.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
So here's the reason this robot doesn't get a contract. No one leaves him in for 7 innings. Look at the list of 5.0+ra allowed guys. There are a million relievers with lower era's than that. They don't get those ra's by giving up 4 in seven every time. They do it by giving up three/5 sometimes and 4 in five others. so you take them out after six because for the next marginal inning you're better off with your pen. Now this robot just lost about 30 innings a year.
But isn't the converse of this that if he gives up 4 runs in the 1st inning, you're guaranteed six innings of shutout ball? It's a thought exercise.
Real pitchers don't last all 7 innings if they've already given up 4 runs, but that's because there are no guarantees that they won't let in even more runs if they stay in the game. But robo pitcher is automatic: 7 IP, 4 runs. So as rcmiller3 noted, once robo pitcher has let in 4 runs, he'll put up zeroes the rest of the way.

Of course, there are no parallels to this in real life, but in magic land, this is a pretty valuable thing.
Even robo pitcher though, having given up one run going into inning 4 becomes a liability.
Jay brought it up, but I'm surprised that no one expanded on the idea that such a player would undoubtedly be signed by Colorado or Arizona (in the NL) and maybe Boston/NYY in the AL if they weren't so pitching stacked already.

Then again, I suppose using park/league factors seems like it's cheating the meat of the question.
TB - your final point could be figured by finding out how many Freddy Fitzsimmons jerseys were sold during those six years, and what that translates to in today's dollars (allowing for percentage increases due to the overall amount of merchandise sold now compared to then, which perhaps can be correlated to attendance figures?). I would say that going down the road to prove that is probably way more trouble than worth, hmm?

All - great article, thanks for sharing this with us. A fun read.
Now I really want a 1927 Freddie Fitzsimmons jersey.
I bet RoboPunter could kick Joe Consistency's butt ...
1) Really fun stuff, thanks for posting.

2) Gary and Rany, good to see you chiming in. If these roundtable discussions are the only way we see you on BP, I hope the BP roundtable becomes a weekly feature.

3) The only teams that should even consider signing the Robot are the ones to whom the $5MM is an insignificant drop in the bucket. The marginal value of 4 RA/7 IP at $5MM/year over 4.5 RA/7 IP at $450k/year is infinitesimal.
My nomination for RoboPitcher .... Walt Terrell
Can you guys continue doing stuff like this, haha.
If I'm the Cubs, I'm a buyer... Start Robo pitcher on days the wind is obviously blowing out. Also, maybe all games at coors with that HUGE outfield. Makes the rest of the staff all that much better. 30 years $300M easy.
If I'm the rest of the NL Central, I'm hoping the Cubs buy too. Signing a consistently replacement-level player would make your rivals giddy at the thought of lining up to face him.

Weird how you can make an argument for why you'd want this guy on YOUR team, but also that you'd want him to be on your OPPONENT's team.

I'd much rather face Joe Consistency in 2009 than Ricky Nolasco. At least Joe doesn't have the possibility of making me look foolish.
The point is that Colorado pitchers allowed 4.68 Runs/game at home last year, so Joe Consistency's 5.15 Runs/game (while absorbing 7 innings a start in normally bullpen-chewing Coors Field) would be a plus if he were the 4th or 5th starter.

Moreso than if he'd been surrendering 5.15 Runs/game in San Diego or SF.
Oh, I understood the point. I was being tongue-in-cheek. I was just joking that if I were an opposing manager in the same division, I would be drooling over facing RoboPitcher several times a year. I'd line my ace up against him, field my best defense, and it's an almost-automatic win.
This was great.

Also, who is Gary Huckaby?
Good eye, Brian. Could he be related to BP founder Gary Huckabay?
The answer to the marble question, of course is depends on how big is the urn. If the urn can hold more than 70 marbles (maximum black > 50), then choose 20/20 when want white makes sense. Otherwise the 20/n would offer better %.

We had a similar conversation about Jeff Suppan and whether he is even worth rostering over on the NFBC message boards. Basically, it depends on the quality of the team and what they are hoping to achieve.

If you are going no place fast, but need a 5th starter and want to get all your guys in the minors more seasoning, this guy is valuable in that he fills the void and keeps your bullpen from getting overworked.

On the other end of the spectrum, if you are a contending team that needs a fifth starter, then you look at how many runs you might give up and allow (with and without him) and see if he makes sense. Let's say, for example, you have an 85 win team without this guy (not counting the fifth starter's spot). Then he makes no sense to add and you'd rather some volatile high-upside minor leaguer who might actually help you or better yet, spend the money on someone better.

But, if you have a team that would win 95 games, even with this guy included, then he is invaluable because economically you don'yt need to spend more and at the same time he gives you much more certainty than a minor leaguer and keeps your bullpen fresh.
Just a comment about the general format. This roundtable plays to BP's strengths; in particular, it shows what strengths Russell brings to the group. Count me as a reader who wants to read more such roundtables in the future.
Yup. Really good read. Thanks for this.
Merci beaucoup.
Shouldn't that be "oshen horoshow?"
I couldn't find any mention to the robot's intangibles or how well he gets along with the guys in the clubhouse. Surely as a robot, these things would be programmed in, correct?
The average AL team in 2009 scored 4.82 runs per game. If our machine gives up 4 over 7.0, and the bullpen can handle a 4.50 RA, this team will give up 5 runs per start. Multiply by 162 and this team scores 781 runs and allows 810 (essentially, the Rangers offense and somewhere between Toronto and KC on the hill), for a Pythagorean win % of .483. I'd bet that just about every average or better offense would take that as their 5th starter, and would pay a nice price for the privilege (without even getting to bullpen issues).

I think Christina is right where she takes the simplistic view and comes up with about $5m.
I completely agree, wonderful format. Keep them coming.
I would say that the Robot is valuable for the following reasons:

1) The manager can "play to win" on a fair number of outings. For example: If the team scored 4/5 runs by the 5th or 6th inning, he can start "manufacturing runs" to get the lead/buffer for his team. If the manager KNEW that he could go into the 8th with a 5-4 lead by bunting, pinch-running, etc etc in the middle parts of a game, he would do it and let his relievers decide the outcome, rather than hope for more scoring opportunities in the 8th/9th against the opponents better relievers (and then start employing the “small-ball” tactics as is typically done now. Granted, if his 8th/9th guys blew the lead, they would be out of bullets. However, a manager has to like his odds of manufacturing runs against a tired starter/middle-relief type, more than against the opponents best relievers and hoping his relievers close the deal, than hoping to score against the other team’s best relievers. It allows the manager to choose when he makes his bets and, more likely than not, lets him make these bets against worse pitching. This gives them an in-game advantage over their opponent.

2) This gives the front-office an advantage in team construction. Simply put, relievers are worth more to this team than other teams because of the certainty that the Robot provides. Well, I should clarify; “dominant” relievers are worth more. Here’s why: The team knows that in 20% of its games it will likely (could have extra innings) require exactly 2 innings of relief pitching and thus can leverage their best relievers for these games. It is safe to assume a team will score 4 or more runs in the majority of their games because only 9 teams (iterations of teams, not organizations over the 5 years, though this is another aspect to consider because of the 3-year fictional contract) in the past 5 seasons have averaged fewer than 4 runs. In these close games, good relief pitching is crucial, this teams knows that it will have a disproportionate number of close games compared to other teams. (sidebar: does someone with more advanced statistics at their disposal [Mr. Davenport?] have the actual data distribution of how many runs teams score per game, rather than just the season average?) I assume that teams scoring tallies per game are somewhat normally distributed around 3-6 with fat tails of 7+ and 0-2 RS. Assuming that this distribution is indeed true, a team with the Robot can gain more from dominant relievers than a team without, giving them a competitive advantage in team construction.

3) The Robot allows a manager to play games preceding the Robot’s start more aggressively during the regular season. A manager can burn through relievers at a higher rate because they know that only 2 innings will be required the following game. The only down-side to this strategy (which I believe to be an acceptable one, given the gain of more heavily utilizing relievers in the game preceding Robot’s start) is that if you utilize relievers whom you would desire to use in Robot’s start, they may not be available. However, a team would almost certainly have better overall bullpen usage and most likely lower injury rates, with our Robot manning the 5th spot (probably about right given that 5.14 RA) in our rotation.

4) Finally, I would posit that there is a psychological advantage. A team, regardless of offensive prowess, would start with a definite hurdle in mind for what it takes to win and this is probably a good thing to have in mind for teammates who are made of flesh and blood and have emotions and inconsistencies. Five runs at the onset of a game isn’t asking too much if it is known that given decent relief pitching (from your shiny Dominant relievers) you are going to win. Though it certainly isn’t something easily quantifiable, the knowledge that scoring 5 runs will guarantee you a lead in the 8th inning is probably not a thing to thumb your nose at either, regardless of relief pitching.

No, the Robot is not independently valuable because you can replace him/her/it easily at the league minimum or close to it. But the fixed nature of the performance allows the front-office, the manager, and the players to leverage their performance, managing style and decisions around a fixed point. Granted, that fixed point isn’t great, but its steady and that’s something.
The last point is the most important: the certainty that robo pitcher brings, even if it's mediocre, lets managers and the FO make decisions involving this player with much more confidence. It seems to me that this is more or less exactly what the focus on stats and metrics for real players is chasing.
Assuming the same mediocraty in the bullpen, the number of runs allowed is 833 over 162 games, and let's assume 8 3/4 innings / game or 1418 innings for the year / 7 , is 6.32 pitchers required on your staff for total replacement-level pitching. So using 7 robo-pitchers costs you $35M (at the suggested $5M per robot).

No arm trouble? One robot costs $150M for 30 years?

How many wins would the Yankees have had in 2009?

How many wins would the Yankees have had if they had an additional $70M to spend on hitting in 2009?

Based on this I believe the $5M number is low low low.
KG - how many of the front office officials who thought RoboPitcher were worth lots of money were Dayton Moore?
Every team in baseball allowed 4 runs per game or more on average. Wouldn't the fate of RoboPitcher depend mostly on the quality of the bullpen?
Give a team 4 of these robots, plus several 4 era robot relievers. What would you have? A team that wouldn't have to worry at all about defence! Fill your roster with good hit/no field players and move your fences in. I think that team would win a lot of games.
I'm seeing a lineup with Branyan at 3B, Uggla at 2B, Youkilis in LF, Posada at catcher, Jeter at short...
Forget Jeter at short, why not another 1st basemen. It's interesting to think about how odd the strategies for the two teams would be, Roboteam would field all hitters without even caring about what positions they play, since RA is fielding independent in this magical case, and the other team would field the best fielders available completely independent of offense since they score 4 either way.
This whole time I'm thinking, why hasn't anybody mentioned that he would have amazing value because you can stack your lineup with your best 8 or 9 hitters because you don't care at all about defense - and then the last poster touches on this. Of course, Gordon went too far because you can't field your whole roster this way unless you had 4 starting robots.
Don't forget that it goes both ways. While the robot's team can't sign a bunch of all-hit players for 35 games a year, the opponent can just call up their best fielders from throughout the organization and try to hold the robot's team to three runs. Every time the robot starts, he'll face a team with three center-fielders and the two or three best shortstops in the organization. While the robot's team can't constantly cycle their roster, an opponent to plays him once or twice can just swap a few players with options for better defenders on the 40-man.
Blernsball anyone?
Dudes, who cares about what you'd do with a robot pitcher? The real question addressed is what's the value of mediocre innings from a starting pitcher? 'RoboPitcher' has to do with 'ideal type' intellectual constructs and all. Since Suppans and Livans are out there, your team can probably sign one if they want. So should they? How much if any value do they actually have? That's what's being asked and looked at here.
The question is being a proxy. By signing guys like Suppan (or Wakefield, or whoever your comp du jour happens to be) you are gaining some level of certainty in performance and whiel you can't be 100% certain you are getting 7ip/4er EVERY time out, you can be significantly more certain youll get something close to that with the signing of these guys. This is why the "philosophical" (ivory tower, whatever your degrading/complimenting adjective attribution may be) argument is worthwhile. We are not saying "this guy exists and should we give him a contract?" we are saying "guys similar to this exist and are they being paid appropriately?"

The absoluteness of the proposed Robot allows us to filter out some of the noise in the conversation pertaining to this ilk of pitchers. It nullifies the tangents that people get caught up in when they talk about Suppan (He's old! His Fastball is flat! And not that fast!) or Wakefield (How do we handle Knuckleballers! He's got a goatee!) and allows us to more appropriately address the value approximation of these guys and whether or not they are worth the sometimes gaudy, and other times thrifty contracts.
I am wondering about a few points that weren't clarified:

1)Does Joe Consistency let in those four runs at the same rate or could he let all four runs in one inning? If the latter is true and teams know of his consistent tendencies, on some rarer starts couldn't they just pull him out after 5 innings after giving up zero or one runs? Couldn't teams interfere with the experiment and get more value out of him in that sense?

2)Could this guy get a better contract midway through the season? If a team (like the Red Sox last year) had a bunch of starters get injured/played awfully, they would be glad to scoop up a pitcher who could pitch seven innings for them as a filler. They wouldn't have found him necessary during the offseason though since they already signed Brad Penny and John Smoltz.

3)Is his stuff also consistent every time? Would a below average hitter get just as good pitches as the above average one? If so, then he becomes a valuable situational pitcher on a roster. He could get the bottom part of the lineup out probably as often as a great pitcher if his stuff is consistent as well.
Great idea: pitch him only against the high-octane offenses, like how the Yankees used to save Whitey Ford to pitch against their contenders. More bang for your 4 run buck.
also, this was a great read. Could you guys possibly make a weekly column where a philosophical question is raised and the great thinkers all assemble together and analyze? I know it's a lot to ask for, but I think it would be a great addition to the site.
The innings by this robot are valuable as well as the ease on the bullpen. It's interesting that when I see this line in the box score, I think "bleh."

Ultimately though the value to a team is dependent upon all of the factors mentioned, mainly offense and bullpen. It also depends on the predictability of the offense and bullpen. Even if you have an offense that averages more than 4 runs per game (the minimum to make this a winning pitcher), you're still going to lose too many games if the distribution of runs/game is highly skewed.

The bullpen make-up is also a tricky question because the team's expected winning percentage is dependent on the quality and dependability of its best relievers. If you assume that the team always has 2 innings of quality relief in it when robot pitches, then it simplifies the calculation of the value of the offense because basically, you would only need to determine the % of the time that your team scores 5 or more runs to determine the expected winning %. It doesn't matter as much about the quality of the poorer relievers on the team because they would presumably get work in robot's games only if the team was losing 4-0 or 4-1 when robot left after the 7th.

I really enjoy this thought exercise and hope you do more of it.
I guess I looked at this a bit differently. I took it from the idea that I don't have to use this guy as a starter necessarily, but basically as a mopup pitcher for 32 games that are blownout in the first or second inning. The way I look at it, if a game is blown out in the first or second inning, you're probably using 4-5 pitchers to cover the rest of the distance which makes them unavailable for the next day's game that might count. There's a reason why Tim Wakefield, even on his high ERA days, has value in that he can soak up a lot of innings.

Now, I might also use him as a fifth starter or emergency starter if one of my four regular starters needed a day off, or against a high offensive team or high offensive environment. Either way, I think there is some value in there.
Fantastic read, and thank you for sharing!

Just as an FYI, the question can also be phrased to the GMs slightly differently... "You're not going to get a chance to sign this robopitcher because his agent won't negotiate with you. But, what do you think the other GMs would offer him for a 3-year deal?"

There was a marketing team for a car company back in the day that asked people what they wanted in a car. They learned that car owners value consistency, craftmanship, and engineering in their cars. So, the car company built and marketed cars for those characteristics. They couldn't sell them.

The marketing people went out and asked folks what their neighbors wanted in a car, and they learned that the neigbors wanted flashy, sexy, sporty cars - the kind of cars that were actually selling off the lots.

Moral: people are too embarassed to answer the marketers' question honestly - they give the answer that they think will make them look good.

Some GM is going to offer Robopitcher 3 years at $7.5m per.

All the way through the roundtable, I was hoping someone would ask if the Robot had a good face.
I agree with a number of the other commentators; this is a absolutely great format that plays to your strengths, and I would love to see more of it. Thanks.
This was fantastic. Thanks.
a hopelessly ivory-towered academic, I don't exempt myself from what I'm about to say:

I'm amused by how loosely very smart people can be in their thinking. Carleton's basically referencing (without citing any sources unfortunately) a set of psychological experiments that identify a tendency. Then he gives a startlingly simplistic analysis of what motivates that tendency.

Then this Pease fellow totally misunderstands Carleton just so he can show how much "smarter" he is than "most people." (Remembering, of course, that most people quite rationally take an "irrational" abstract thought experiment (I mean, somebody should tell Pease that urns aren't indefinitely large) and construe it in such a way that it becomes "rational" enough to answer. Which is to say, given what most people know about prototypically-sized marbles and given what most people know about a prototypically-sized urn, forty marbles is near the upper limit of its capacity.)

And THEN, and this is the real kicker, Schwartz expends a bunch of cogent sentences explaining how and why most people could very rationally "irrationally" perform the experiment (and, in general, make decisions). But then, he goes on to to call that performance (and, in general, those decisions) "irrational" precisely because he can't seem to think around the comically narrow definition of rationality of rational choice theory.

In effect, it seems to me a really good example of people averring a priori a definition of rationality that can then be tested only by setting up comically artificial problems. And when the subjects of these experiments "fail" to be rational precisely because of the comically artificial nature of the problems, the experimenters say, "well, people are irrational" when at best one can say nothing more than "well, people don't know probability theory."
I don't know who this Schwartz fellow is, but he should ask the Swartz fellow why he said he didn't believe that the behavior in this experiment was irrational ;-)

In all seriousness, the point I was trying to make but maybe wasn't clear is that I'm not sure GMs are irrationally risk averse, but the economic assumptions required to generate a rational perfectly competitive market do not hold in baseball due to lack of free entry of competing businesses, so they may be. Thanks for your thoughts.
Okay, smartie. Explain it better for our edification, please.
At the risk of repeating you....what's your point?
"I'm amused by how loosely very smart people can be in their thinking. Carleton's basically referencing (without citing any sources unfortunately) a set of psychological experiments that identify a tendency. Then he gives a startlingly simplistic analysis of what motivates that tendency. "

I'm asking that he provide a sophisticated analysis that cites sources, since he's saying that the summary above is wrong or misleading. He seems to believe that he can educate on this topic, which, seemingly, is now baseball-related. So...
not you krissbeth, I was replying to his original (lengthy) post that didn't really seem to have a conclusion.
His point is he is the smartest commentator on the boards because he stashed a Latin dictionary in his ivory tower, which is why he could not spell Swartz correctly :)
Andywest, that is.
Related question (or suggestion for another discussion): What's the ERA of the best robot humans can make, which is not permitted to throw over 90mph?
Assuming you let Greg Maddux decide the pitch selection on the fly...I say in the 1-2.00 range. Should be able to get superhuman amount of spin and better location than a human. I don't know if a robot can throw a changeup, though. And it can't field. Still, I say you teach it throw a forkball plus like 6 other pitches (4-seam, 2-seam, cutter, curve, slider, screwball, sinker), play the infield in, and reap a Cy Young winner.

Here's a spinoff of your spinoff question: if all this superhuman robot pitcher could throw was a perfect sinker (or any other single pitch--sinker's what came to mind for me, but I guess Rivera's made a strong argument for cutter as one pitch you can live on), how would it fare?
Tversky-style question for Russell:

How different is the answer you get from those front-office types if, instead of talking about a robo-pitcher who always gives up 4 over 7, you instead talk about a rule change that would allow you (but not your opponent) to skip the defensive half of the first 7 innings, instead spotting your opponent 4 runs...

Would you be willing to reduce your roster size by 1, in order to have this option? Would you be willing to pay money, in addition to giving up a roster spot?

Clay probably remembers that we had this conversation back in the Dark Ages, on, when the "Flake" stat was first invented. That came out of my pestering Michael Wolverton for some stats on the variability of support-neutral game performances. The consensus back then was exactly what Clay said above -- that in general you want your above-average starting pitchers to be consistent, and your below-average starting pitchers to be flaky. The farther below average the mean performance is, the more valuable flakiness becomes. You would much rather have Joe Twoface, who gives up exactly 6 over 7 half the time, and 2 over 7 the other half of the time.

(Of course, if you pull Joe Twoface after he's given up 3, the universe implodes.)
Tversky indeed! I have the feeling that you wouldn't get any takers from the GMs. I think that the appeal of RoboPitcher here is that he looks human, but instead of having to suffer through the inconsistency that goes with humans, he's a guarantee. I think that's what the GMs would pay for. And they'd pay for it out-of-sync with your proposal, which of course is exactly the same thing.

Conclusions: humans are irrational.
Great discussion. I think it's important to remember that a consistent human pitcher is a lot closer to an inconsistent human pitcher than a robot.
I think people are underrating the value of the Robopitcher's performance.

Let's actual assume we have a team with 5 Robopitchers, and give them the 2009 Yankees offensive performance. Assuming a bullpen with a 4.50 RA, they would on average give up 1 run during their 2 innings, and going with the consistency theme, let's assume that this is exactly what happens. In 75 games, they scored 6+ runs, so they would win all of those. In another 25, they scored 5 runs, which would take them to extra innings. If they win half of those, that takes them to 87.5 wins.

Not bad for a team that gives up 5 runs a game, and that's before you consider other strategies that could be employed as result of this consistency. As Tommy Bennett stated, Robopitcher would be more valuable as a reliever. The team could start their bullpen, and have Robopitcher be a 7 inning closer. Obviously, if they found themselves tied in the middle of the game, they could already start playing it like extra innings (while essentially always enjoying home field advantage.) You could focus your payroll dollars on a good offense. You would need fewer relievers on the roster. You could rest your best hitters as soon as you had enough runs. Good management should be able to improve that win total into the mid 90s.

I saw some pretty low values for the Robopitcher, but I think you could take 5 Robopitchers, a good offense, and be on the fringe of playoff contention before you account for any strategic advantages. That's certainly worth something.

Of course, what takes this hypothetical so far from reality is that no actual pitcher gives us anywhere near this level of guarantee. All of the real life examples of Robopitcher performance also carry the risk of deteriorating performance or injury, which is something I think most people tend to underestimate.