Money for Nothing, Part Two

I’m too lazy to do this myself, but I’ll propose a simple methodology to evaluate general managers using the data presented above.  If we assume (incorrectly, of course) that all free agents go to the highest bidder, and also that each GM is spending exactly as much as he’s allowed, we can see how each GM did signing free agents the previous winter relative to his budget.  Obviously guys like Loaiza are going to make a GM look good; the point here is that if you’re Brian Cashman, it’s OK to overspend for a WS, but it’s not a good idea if you want to run the Brewers.

I’m happy to elucidate more as long as I don’t have to open Excel.  :D

Good idea, David.  I’ll try and take a stab at this for a few teams soon.  If you (or anyone else) wants to do this yourself, feel free to develop some commentary and e-mail it to me, and I’ll post it as an article here (as long as it passes my stern editorial eye

And because I listed each player’s win share value in the previous post, you don’t even need to open Excel.  Though it would probably help.

I’ll run a regression on the team values vs. team salaries and get teh over or under on the regression on Monday.  I think that would show us how good the team performed vs their payroll scale.

(Continued apologies for not “bringing data.”)

It was stated earlier that the trend is 0.3 WS per $100k.  This can lead us to a few conclusions:

- An average GM hoping to have a .500 record would need a payroll of:
  81 Wins * 3 WS/Win * $333.3k / WS = $81 M.

- If we assume it takes, on average, 95 wins to make the playoffs, the GM would need a $95M payroll.

There are two key assumptions above:
1)  It’ll take 95 wins to make the playoffs.
2)  Expected wins = Previous Year WS / 3

I’ll leave the first part alone for now, but suffice it to say you’ll probably need to spend more to make the playoffs from the AL East than from the AL Central.

However, I’ll further disect the second point.  I’ll take it on faith (and the work of others) that we can predict with some accuracy the number of WS a player will generate over the next year, after accounting for age, park factors, and such.  I’ll further grant that a team’s number of wins is *expected* to be projected WS/3.  The catch is the shape of the probability curve.

Say for the sake of argument you’ve got two players that you project to earn 10 WS next year.  I’ll call them Juan and Jose.

Let’s say Juan is a 25 year old who just had a break-out season, and Jose is a 32 year old veteran.  From our above projections, we think there’s an 80% chance Jose will perform to his “expected” level of 10 WS, and a 95% chance he’ll produce 8 - 12 WS.  However, we’re not sure yet whether Juan was just a fluke, and his 95% range is 2-20 WS.

Now, let’s say you’re Brian Cashman.  You’ve run your models, and you predict your team next year already can produce 285 WS (95 Wins), and you want to add either Jose or Juan.  They’ve got the same asking price.
As a counter point, imagine you’re Billy Beane, and you’re making the same choice, but your projections show your existing team will only produce 255 WS (85 wins.)

In this scenario, I posit that Jose is more valuable to the Yankees, and Juan is more valuable to the A’s.  I’ll skip the explanation, as I imagine this crowd can follow where I’m going.

(Continued)

Now, let us make another assumption: teams in Cashman’s position have more money than those in Beane’s position.

If that assumption is true, and the conclusion regarding Jose and Juan’s value is also true, we would expect Jose (the proven veteran) to get more money than Jose (the rook), *even though we expect the same number of win shares from each.*

To put that into practice, let’s say you’re a GM trying to decide whether to sign Bartolo Colon or Andy Pettite.  They produced 17 and 15 WS in 2003.

Trying to predict the number of WS each will produce in the future is of course very complicated, but I’m guessing that any reasonable analysis would suggest that Pettite’s got a more narrow band of expectation than does Colon.  And thus, Pettite would be of more value to an established team like the Yankees or Red Sox, whereas Colon would be of more value to an up-and-comer like the Angels or (dare I say it?) the Astros.

This brings us back to the list of team $/WS: although the Yankees spent way over $1M/win, I’d contend that their team had a *very* small expectation of missing the playoffs.  A lot of things would have had to go horribly wrong.  So while their players were expensive in terms of $/WS, they also didn’t have any real risks on their core roster (aside from Contreras and perhaps Soriano).

Meanwhile, whereas the Braves look really good at $/WS, I’d argue they were more lucky on guys like Giles than they were brilliant in choosing free agents.

So, here’s where I’d want to go with this:

  - See if we can find someone who likes to project the future.
  - Get them to put some statistical variance in their models.  We’d want to know not just how many WS we *expect* each player to generate, but also what the 95% region of the curve is.
  - Sum all these distributions up to form one distribution for the whole team, and then compare that to the team’s total salary.

After all, I’m all for celebrating GMs who employ great players on the cheap.  However, I wouldn’t want the GM of *my* favorite team (the Giants) to start searching for the next Loaiza if they can nail down a very-likely 10 WS guy with the same dough.

Cheers,
-Dave

Nice work. I just like the fact that the Yankees overpaid.  However, I do have a concern about the methodology.

First, is it valid to use this on all players, particularly those who have not been free agents and thus priced by the open market? Soriano and Pujols should be great values because no one has paid market value for them. Does this reflect well on the GM, or is it luck and a good farm system?

In the end, I think the GM should get the credit for money effectively spent.  I would think that a team’s Net Win Share Value is the primary responsibility of the GM.  Whether he has achieved it through smart free agent signings, a strong farm system or even luck, these all fall onto the GM.  If his team were unlucky, he’d be blamed for that, too.

David, you’ve got some great ideas.  I’m not much into projecting the future, myself.  In fact, I think that Win Shares are not a good tool for projecting future performance.  I’d rather rely on basic ability metrics for that sort of thing.  Still, the general exercise, as you’ve outlined it, seems right on.

John and I have done some more work with the data.  I plan to post it tomorrow.