THE BOOK--Playing The Percentages In Baseball

In some ways, this sounds like “I’m not a fan of other methods because they’re not my method”

I don’t understand his gift card analogy.  Can anyone help?

The question to ask is: how much do you pay the free agent that the other 29 teams don’t want to pay the minimum for?

The perfect example of such a player is Jacque Jones.  I have him forecast to be at 0 wins above replacement.  I’m guessing Chone has him around the same.  He was signed to a minor league deal by the Reds.  Had the Reds signed him to a major league deal, would he get one penny above the minimum?

Jones is exactly the player that everyone is being compared against.  He is a replacement player, he’s treated as a replacement player, and he’s being paid as much as a replacement player.

Whether he gets 1 PA or 500 PA in MLB will not take away from the fact that he is a replacement level player who deserves to get paid no more than the minimum.

(Unless of course our estimate of his true talent level went horribly awry.)

***

For people in Fantasy draft, the valuation is no different for you than for MLB.

Thee were the two thing that bothered me the most, perhaps because I do not understand thing as well as I would like to think.

*For one, replacement players aren’t worth league minimum any more than a Starbucks gift card from your grandma makes coffee free.*

The idea that a replacement player is “worth” MLB minimum seams axiomatic to me.  That is, it does not matter what such a player is “worth” - that is how much he will be paid.  How can you assign a replacement player any other value?

*Also, the revenue-wins relationship is non-linear, increasing at an increasing rate.*

This statement bugs me because it seems to make such good sense, but I thought everyone who has looked at it has empirically found the win-revenue ratio to be linear.

It makes a lot of sense, given the 25-player limit, that the difference between a six-win player and a four-win player is greater than the difference between a replacement player and a two-win player.  But, as far as I understand, everyone who has looked at it has found that to be untrue.  If that is the case, then the statement, even in an interview, needs more explanation.

By the way, for people who don’t know, the system that Fangraphs has implemented is the closest thing to what I do that is out there.  Chone would likely be the next closest.

JC is clearly struggling to figure this all out.  He’s acknowledging he was wrong about Frenchy, and is coming around on the issue of relievers/leverage.  That’s progress.  My guess is he eventually settles on a model that builds off the average player, adding/subtracting from the mean, that ends up valuing the Jaque Joneses at about league minimum.  Then he never has to admits he’s using replacement level.

*Also, the revenue-wins relationship is non-linear, increasing at an increasing rate.*

This statement bugs me because it seems to make such good sense, but I thought everyone who has looked at it has empirically found the win-revenue ratio to be linear.

Depends if you’re talking the player-scale or the team-scale.  For player salaries, it’s pretty linear, maybe slightly non-linear.  But for teams, going from 85-90 wins brings in a lot more revenue than going from 60 t 65 wins.  That doesn’t really affect player salaries, though, just which teams sign the free agents.

Back on the player level, if you set the replacement-level too low—like with old WARP or a JC-like analysis—then you will find the $/win converter is non-linear (like with MORP) because you want to make the first few wins worthless and only value the high-end wins.  But what that’s really doing is putting in a fudged, correct replacement level.

Isn’t setting replacement level talent independent from setting the cost of replacement level?  I agree that people can (and in BP’s case, have) set the talent level of replacement level too low, but the “value” of replacement level is defined by MLB minimum salary, no?  Given the salary in the minor leagues, it is obvious that replacement level play *could* be had for much less than MLB minimum, but the rules prevent that.  Given that fact, I don’t see how anyone could define the dollar value of replacement level as anything other than MLB minimum salary.  JC seems to suggest otherwise (and in his Q&A by proxy seems to state so rather clearly, since he purports to agree with FG’s talent evaluation, but not it’s dollar values).

We use MLB minimum because it’s convenient.  The actual value of a MLB replacement level player is probably closer to 75-100K, which is what I think the top minor league free agents get, at least last time I saw an article showing what they can get.  The difference between the best minor leaguer and the worst major leaguer is nothing.

Especially since one just got demoted and the other got called up (that sentence should be true most days of the season).

Think of authors and book sales.  One author sells a million copies.  He gets a lot of money.  Another wrote 400 pages and sold only 12.  Mostly to his family.  He did a lot more work than the guy who never touched pen to paper, but their income is about the same.

Just like Francoeur had more PA against MLB pitching than I did last year, 652 to 1*, but he didn’t do anything more to win ballgames.

* vs Greg Olson, Orioles fantasy camp, F7.

Excellent use of the Pozterisk.

"Just like Francoeur had more PA against MLB pitching than I did last year, 652 to 1*, but he didn’t do anything more to win ballgames.”

That’s the crux of our disagreement with JC.  He is in fact comparing Francouer to you, and saying the Braves would have considerably fewer wins (and thus less revenue) with you in RF.  We don’t disagree, we just think that it matters that there are 100 guys in AAA who could have put up the same production as Frenchy, and would gladly work for league minimum.  JC doesn’t think that matters.

JC is actually, as of his book, comparing everyone to average. (Many Scully-esque MRP models do, in fact, compare everyone to the “no-hit” baseline, with the caveat that all of them are also average fielders at their position.) The problem is when he computes the value of the average baseball player; if he had his average salary value set properly it’s likely he would, in fact, see a salary of $0/minimum at replacement level.

Now, since then it’s very possible that he moved from an average to a zero-level baseline. From his description, that model would explicitly assume an alternative of nine players who hilt like Rally but field their position as well as the league average at that position.

"nine players who hilt like Rally”

Hey! Small sample size!  Give me 651 more to see what I can do.

Rally/13: don’t you already have enough PAs to generate some PECOTA comparables? What’s the MLE for Orioles fantasy camp?