« More articles in 1   |   Go Home

What’s Missing From MRP Systems? The Team Factor

posted by Shawn Hoffman, on December 17th, 2008 at 07:42 pm, in category 1

Rob Neyer wrote a piece yesterday that pointed out how differently certain systems value Raul Ibanez’s next three seasons. MGL and J.C. Bradbury both responded in the comments, as each of their valuations were discussed in the original post (J.C. had $46.5M, Lichtman had about $10M).

Here’s the problem: a player doesn’t have one fixed marginal revenue product. Instead, each player has a different MRP for every single Major League team.

This point is absolutely crucial. Raul Ibanez is not statically worth $46.5M over the next three years, or $10M, or any other number you want to throw out there. He is worth $X million to one team, $Y million to another team, $Z million to a third team, and so on.

From what I understand, the existing valuation systems (including J.C.’s, MGL’s, BP’s MORP, and others) take into account a) the player’s projected on-field value in terms of marginal wins, and b) some function of industry revenues and/or existing salaries (i.e. usually either the average amount paid per win on the free agent market, or the sport’s projected total revenues). Here are some of the key factors that they ignore:

  • The team’s positional needs. If a team has Joey Gathright starting in left field, Raul Ibanez will be worth much more to them than if they have Ted Williams starting in left field.
  • The team’s competitive position. Ibanez wouldn’t be of much use to the Pirates, who are nowhere near making the playoffs.
  • The team’s earning potential. Teams in big markets earn more per win than teams in small markets, and are therefore more likely to offer higher salaries to free agents.

So if the Phillies’ next best option to play left field had been, let’s say, Reggie Abercrombie, and they felt they needed only a 2-3 win upgrade to make the playoffs, Ibanez might really be worth $10 million (in fact, in that situation he probably would be worth much more than that). But at the same time, his value to, let’s say, the Red Sox, is basically zero, since they already have better players in left (Jason Bay) and DH (David Ortiz).

Without these team-specific factors, no MRP system can give us any real assessment of what a player is “worth.” That’s not to say the existing ones don’t have any kind of value; they might work well as a predictive tool, I honestly don’t know. But in actually evaluating whether a player was “overpaid” or “underpaid,” these systems are far too simple.

Feedback? Write a comment, or e-mail the author at shawn(AT)squawkingbaseball.com


2 Existing Comments

Add New Comment

  1. on December 18th at 08:01 am
    JC said:

    The single MRP number is for simplicity. I can’t speak for other systems, but my estimates assume the player is signed to .500 ballclub. On better teams the player will be worth more, I have acknowledged that. Individual team factors do affect a player’s worth, but sample size makes it difficult to identify the exact impact by market. I’ve estimated some models trying to tease out team effects, and it just becomes a big mess because of multicollinearity problems. Luckily, the returns are similar across teams, but it’s OK to acknowledge that a player might be worth more to one team than another. For example, Jeff Francoeur is worth more to Atlanta than any other club because he’s from the area. The effect of market size is difficult to quantify. I use population as a control variable in my revenue regressions to dampen the influence of market size, which players are not responsible for. I have explored models that multiply player performance times market size, but the estimates on the coefficients are not statistically significant.

    You really ought to try calculating some MRP estimates of your own. Then you will see that it is much more of an art than you are making it out to be. The modeler is forced to make tradeoffs to settle on a final model. I’m willing to adopt a superior method. This is the second time you have criticized MRP estimates without offering a system of your own. Arguing that other systems are not perfect is not going to convince anyone doing this type of modeling, because we already understand our systems’ imperfections better than you do.

  2. on December 18th at 09:27 am
    Shawn said:

    JC,

    I think we’re actually agreeing on most points, but disagreeing philosophically. I haven’t done any modeling because I don’t think it’s possible to peg a player’s static value. It’s too dynamic, and involves too many factors that we can’t estimate with any degree of certainty. Even if we try to estimate the team’s earning potential, and use the next best option on the roster, we can’t really know who their next best option is (trades and free agency included), or how much each specific team makes per win, or per playoff appearance.

    The modeling can be used as a predictive tool, as I mentioned in the post. I.e. how much would Player X usually get in free agency? There’s value in that, especially if your team is going to bid on the player. But even if we expect Orlando Hudson to get $10M per year in free agency, that doesn’t mean that his MRP for 2009 is $10M. These are obviously very different questions.

Pittsburgh Florist

Blog Design and Development by Veritocracy Consulting