Monday, August 03, 2009

Understanding Surplus Value

Play's to firstJoey Votto is not only the Reds' best player, but also one of their cheapest players. Image by phillenium1979 via Flickr

In the Rolen trade analysis, I spent much of my time evaluating "surplus value" in my assessments of the players involved. This is not the first time I've posted about this, though perhaps it's the highest profile case of this sort. It is, however, becoming a more mainstream approach to analyzing trades (see, for example, Sky's Trade Value Calculator). I thought I'd take a stab at explaining this concept, and how it has been extended to prospect valuation.

The core of it is understanding the connection between player value and payroll.

The major league minimum this year is $400k. You need 25 guys on your team (assuming no injuries, etc...), which means the minimum possible payroll in MLB right now is $10 million ($400k * 25). How well would such a team perform? Well, if you use that $10 million exclusively on free agents, you're not going to get very good players--mostly guys who can't get a job anywhere else, minor league free agents, etc. Those guys are essentially the definition of replacement players, and we generally assume a team full of replacement players will play around 0.350 ball, which amounts to 57 wins per season (you'll see other figures around the 'net--which one you choose doesn't change what I'm about to do qualitatively).

On the other side of the coin are the New York Yankees, who began this season with a $201 million payroll. From other work, we know that each win above replacement cost roughly $4.5 million in last offseason's free agent market. Therefore, if the extra $191 million they're spending above the minimum was invested in free agent salaries, they could reasonably expect to win $191/$4.5 = 42 extra wins above the minimum 57, or 97 wins (for what it's worth, the Yankees are on pace for 99 wins). :)

But take the Tampa Bay Rays. They began the year with a $63 million payroll. That's just $53 million over the minimum payroll. If you invested that extra money in free agents, and filled in the rest with replacement players, you'd expect they'd win another $53/$4.5 = 12 extra wins. That would equal 69 total wins. But they're on pace for 91 wins. How?

The answer, of course, is that their team is not made up of free agents. It is made up of a lot of quality, young players making far less than they would had they signed as free agents. For the first three seasons of a player's career in the major leagues, they make at or close to the league minimum. For the next three seasons after that, they are eligible for arbitration and get raises, but for a variety of reasons still get paid well below what they would as a free agent. As a result, for the first six years of a player's career, they will make less than a free agent regardless of their performance on the field.

Most teams operate under a fairly set budget. The median opening day payroll was about $81 million this year. So if a team has that budget, and spend it all on free agents, you'd expect them to win around 57+16 = 73 games. In order to win more games, teams have to have players who produce value above their salaries. For example, this year, Joey Votto has produced ~2.5 wins above replacement (worth $11.3 million on free agent market), but is making just over league minimum. If you fill your team with those kinds of players instead of replacement players, and still spend $71 million on free agents, you'll obviously win a lot more games.

That's what we're talking about when we talk about surplus value. It's the extra value a player provides above what their salary would bring on the free agent market. The keys to winning, at the club level, is 1) to spend money for good players, and 2) to get players who produce value above what their salary would predict. Most teams can't get by on #1 alone, and the easiest way to do #2 is to get young players making "slave" wages.

This is also what makes prospects valuable commodities. Prospects generally have no or minimal MLB playing time, which means that their "owners" get them for six full seasons when they will make below free-agent market rates. So, the short of it is that if you figure out how much surplus value a typical prospect type provides their team during the first six years (including assessments of how often they provide any value to their team at all--many, of course do not), that gives you a good indication of how much value to assign prospects of that type.

In Zach Stewart's case, I pegged him (as many others do) as a class-B pitching prospect. Pitchers of this sort tend to provide the equivalent of $7.3 million in surplus value over their first six seasons. Another way to think of that is that it's a total of ~1.6 wins above replacement over their first six years (only 0.3 WAR/season). Not particularly inspiring, maybe, but it reflects the fact that many class-B pitching prospects do not pan out for one reason or another. But some do, some provide a little value, and some provide a ton of value. The $7.3 million total is the typical amount of surplus value you can hope to get from such a pitcher.

Anyway, this became rather long. But if anyone made it this far, I hope it helped clarify the underpinnings of this kind of trade analysis.

11 comments:

  1. J-

    Thanks for this. My issue isn’t with the surplus value concept. I agree 100% with all of that.

    My issue is with the zero risk cost attached to prospects. As I mentioned in one of my posts, Wang recognizes this, although he gives it short shrift. (“Prospect risk premium: Although the prospect values that I have presented above represent the expected value of prospect performance, they likely overestimate the trade value of prospects. Due to the risk involved with prospects, there is probably a risk premium when it comes to trading prospects.”)

    Wang knows that he overstates the value of prospects. Every analysis that I have seen that cites to Wang’s article never brings up this point. I’m not good enough with that math to properly account for the risk cost, but I know that it’s there.

    A Top 75-100 pitcher has a 43% of never making a meaningful contribution to a major league roster (Ryan Wagner?). A 50% has a chance of being somewhere between replacement and average (Nick Masset?) and a 7% of becoming a regular-pitcher-to-star (Dave Weathers? Maybe Art Rhodes?).

    There’s a cost in running through all of these pitchers before you find your Dave Weathers or even your Nick Massetts (and at that, your Dave Weathers might not materialize until he’s used his 6 controlled six years). The cost is all the “extra” pitchers that I need to pay for while I’m shuffling guys on the Louisville express and trying to separate the Wagners from the Massets from the Weathers. The cost is also all of those innings that I spend on the Ryan Wagners when they produce sub-replacement level pitching (let’s call this “negative value”).

    On the other hand, there is approx. a 75% chance that we get Scott Rolen’s projection correct simply because we have many more Rolen samples on which to base our projection. And because of there is a much smaller chance that we get Rolen’s projection wrong, there is less cost. Even if he produces at only 80% of expectation, that’s still better than replacement value, and I don’t need to (a) have the replacement cost of Jerry Hairston; and (b) have the negative value of Adam Rosales.

    CTM

    ReplyDelete
  2. I have no theoretical basis for this and could be way off, but it seems to me the median surplus value would be much more appropriate to use rather than the average surplus value. The average surplus value of prospects is going to be severely skewed by the superstars. It seems that evaluating prospects on what you'd typically get from them is better than what you'd hope to get.

    ReplyDelete
  3. In three sentences Brad said what I've said poorly in 5 posts.

    CTM

    ReplyDelete
  4. Brad,

    I see why you'd say that. But median value, at least for pitchers of Stewart's caliber, is almost certainly zero.

    CTM,

    Ok, I think you're finally homing in on something substanative with the risk issue. But, fwiw, I still disagree with you on it.

    Risk management is somewhat dependent on your risk tolerence, which means it's hard to put explicit numbers on it that work for everyone. But my take on it is this.

    If you are only looking at one player, and are operating in a vacuum, then risk becomes a big deal. For example, a few offseasons ago we started to see individual young players signing long-term contract extensions for well below what even a conservative projection would indicate they were likely to make if they went year-to-year through arbitration (Tulowitzki, Granderson, Longoria, Shields, etc). The reason they took those contracts was risk--the first $10 million or so that they made was the most important of their lives, and by locking it in they effectively guaranteed themselves a lifetime of financial security. They gave up the opportunity to earn much more in exchange for abolishing the risk of a disastrous injury preventing them from earning that first $10 million. Individuals are risk-averse, because they often have no second chance.

    Their teams assumed the risk of a botched contract by signing the players early, of course. Why? Because teams don't operate in a vacuum. The Rays, for example, "insured" the Shields contract by extending other players, like Longoria, Pena, etc. One of those contracts might not work out, but if you do enough transactions of that sort, and all are favorable on average, you can absorb a deal that ends up as the worst case scenario and still come out ahead.

    Similar arguments are behind why my feeling is that the risk premium you mentioned is likely to not be that big of a deal for teams. If everything rode on one player, then risk becomes important. But teams have lots of players and make lots of deals over time. If you compare one team that discounts all of their prospects due to risk vs another that goes on average value, what you'll probably see is that the discounting team will be less likely to have a disastrous trade for prospects. But the non-discounting team is likely to do better over time, all else being equal.

    To put it another way, by discounting due to risk, you're intentionally working in a loss to each deal to prevent a possible disaster. That makes sense when you only make one or two player decisions. But when you make as many as a MLB team makes, I think you need to start accepting a bit more risk so you don't shoot yourself in the foot with every deal.

    That's my take, anyway. I'm not sure that a lot of teams follow that advice, but my feeling is that the smartest ones--Rays, Twins, etc--do.
    -j

    ReplyDelete
  5. In a vacuum without consideration of your current farm system, it would be better to use the median surplus value of a prospect when considering a trade. As your system depth goes up at the relevant position it obviously becomes more appropriate to use the average value because you are pooling your risk among several players. Obviously trades aren't made in such a vacuum, but system depth at each position should probably be taken into account when valuing prospects. I have no idea how you'd do that, but still...

    ReplyDelete
  6. Brad,

    So, again, median surplus value of most minor league prospects is zero. Does that mean that you treat them as valueless? And if so, how will you ever build up depth? If you treat them as valueless, then you will not hesitate to discard them in trades for veterans with narrower (and more reliable) projections. Doesn't make sense to me...
    -j

    ReplyDelete
  7. J,

    Is it true to say that the median surplus value of a B pitcher is zero? What if you use the median value of only B pitchers?

    Also, I think that some smart teams are recognizing that B / C pitchers may now be overvalued. While the Red Sox were hesitant to deal Bard and Bucholtz, they were willing to flip Hagadone and Knapp in the Martinez trade. On an average surplus value standpoint, that trade didn’t make sense. On a median surplus standpoint, it might.

    CTM

    ReplyDelete
  8. I guess I'm confusing myself. What I'm saying is that in a vacuum, the value of any single prospect (or, at least, any B-level or below prospect) IS near zero because it's unlikely they will contribute real value in the future. The value comes from pooling risk in your farm system because the odds are one/some of them will contribute it. But you're right, that value can just be attributed to that prospect.

    I'm a near idiot when it comes to probabilities and statistics though...

    ReplyDelete
  9. CTM, not sure what you mean. If a slight majority of type-B prospect pitchers are busts--which should be right--then yes, their median value is zero. I don't think that means we should assume they have zero value, though, because some do turn out to be excellent, valuable pitchers putting up good numbers while making next to zero money. Again, the number I'm using is about 1.6 WAR over the first six years. That's one below-average season in the first six. Not a particularly aggressive projection, perhaps, but it seems appropriate to me.

    Brad, I think what you're saying is very much in agreement with what I wrote above. Because teams are working with numerous players, they can take on at least some of the risk inherent in any one player because they insure (pool) that risk with all of their other players. Overall, they will come out ahead of a strategy that avoids assuming prospect risk, because some of their prospects will pan out--even though some of their prospects will bust.
    -j

    ReplyDelete
  10. "If a slight majority of type-B prospect pitchers are busts--which should be right--then yes, their median value is zero."

    Of course that's right. I see it now. Thanks.

    CTM

    ReplyDelete
  11. Just wanted to say to CTM, Brad, and everyone else in this and the last thread, thanks for the spirited discussion. It's been a while since I've had that occur on my blog--my fault, as I haven't been very active. But I enjoyed it, even if we didn't necessarily come to agreement on this stuff. :)
    -j

    ReplyDelete