Microeconomics And Offense (Part 2)

Last week’s post examined offense in the American League through the prism of capital and labor. This week, it’s the National League’s turn.

To review, Quadrant I contains teams which are above average at getting on base, but below average in driving runs in. Quadrant II consists of teams which are below average at both skills. Quadrant III contains teams which are above average in terms of strand percentage, but below average at on-base percentage. Quadrant IV consists of the league’s best offenses, which are above average in both variables.

In the AL, Seattle was by itself in the northwest corner of the graph as the league unquestioned worst offense. In the NL, this spot is occupied by two teams, Houston and Pittsburgh. These teams had almost identical offenses, separated by hundredths of a percentage in both strand percentage and OBP.

The other teams in Quadrant II are New York, Washington, Chicago and Los Angeles. The Nationals hope signing Jayson Werth will improve their offense, which was third-worst in the NL last year. All things being equal, Werth is a good example of a player who can move his team both South and East on this graph. He has a good OBP and above average power, which should improve his team’s strand percentage. However, the Nats lost Adam Dunn, so the net change will likely be a slight improvement in OBP and a slight decrease in strand percentage.

Quadrant IV contains the teams with the best offenses in the league: Arizona, Philadelphia, Milwaukee, St. Louis, Colorado, and Cincinnati. The only surprise team here is the Diamondbacks, who were eighth in the league in runs per game. However, ignoring the quadrants and looking at the graph as a whole, Arizona lies closer to the pack of average offenses than in the Southeast corner with the top dogs.

The Brewers had a good offensive year in 2010, but better power production from Prince Fielder and Ryan Braun next year could lead to a better strand rate and a big improvement in runs per game.

Quadrant I contains only Atlanta, and it is an interesting case. Although they were fifth in the league in runs per game, the Braves had the potential to be a league-leading offense. They led the league in OBP, but a sub-par strand percentage prevented it from reaching its scoring potential. The Braves had eight players, with at least 50 plate appearances, who had a .350 or better OBP. However, there was just not enough power on the team to drive these runs in. Brian McCann led the team with 21 home runs, and the leader in isolated power was Brooks Conrad. With a injury-free season from a maturing Jason Heyward, and the addition of Dan Uggla, Atlanta figures to have a better strand percentage and perhaps a big offensive improvement in 2011.

Quadrant III contains San Diego, Florida and San Francisco. These three teams are close to the origin and not very different than the lump of teams just to the North and Arizona to the East. It is surprising to see the Padres in the mix of average offenses, although much of the offense came from Adrian Gonzalez. Without their best slugger next season, it will not be surprising to see San Diego make significant jumps North and West on this graph and join the Astros and Pirates in no-run land.

In Part 3, we’ll look further into microeconomic theory and use the data estimate a Cobb-Douglas Production Function for offense.

Print This Post

Jesse has been writing for FanGraphs since 2010. He is the director of Consumer Insights at GroupM Next, the innovation unit of GroupM, the world’s largest global media investment management operation. Follow him on Twitter @jesseberger.

38 Responses to “Microeconomics And Offense (Part 2)”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Mike Savino says:


    Obviously, the Padre offense took a hit with the loss of Adrian.

    But are we ignoring the upgrades with Bartlett, Hudson, Maybin, full year of Ludwick? I think maybe a little…

    Vote -1 Vote +1

    • Pat says:

      Those players represent good defensive upgrades, but there’s not much for power there.

      Vote -1 Vote +1

      • Mike Savino says:

        I have to disagree. Maybin can’t hit worse than Tony Gwynn Jr. and would be hard pressed to be a defensive upgrade (Gwynn was 12.9 runs above average last year according to fangraphs).

        Ludwick hit terribly for San Diego last year and might have a bounce back season…or not.

        Last season, Jerry Hairston Jr. and David Eckstein were the Padres middle infield. Orlando Hudson and Jason Bartlett should easily surpass that production.

        Obviously, Brad Hawpe/Kyle Blanks aren’t going to replace Adrian’s production but the easiest place to find replacement offense is at 1b/DH.

        All I’m saying is that the Padres definitely got worse at 1b but it looks like the team will be better offensively at at least four other positions, 2b, SS, CF, LF.

        Vote -1 Vote +1

    • Pat says:

      They may reach base more, but my point is that there won’t be much of an upgrade in power (except in eckstein’s case of course)

      Vote -1 Vote +1

      • Dirk says:

        In addition to the players Mike noted, Chase Headley, Nick Hundley and Wil Venable may improve a bit more as they enter their primes. In the end, I wouldn’t be surprised if the Padres score more runs in 2011.

        Really like the chart by the way, and so will my friend who is a Red’s fan.

        Vote -1 Vote +1

  2. Erik says:

    “Quadrant II consists of teams which are below average at both skills.”

    If you’re going to call them skills, can we get some error bars? How “lucky” is the the strand rate skill?

    Vote -1 Vote +1

    • David says:

      What are potential sources of error in the data?

      Vote -1 Vote +1

      • Barkey Walker says:

        Imagine that the actual events were drawn from a distribution of all possible events… Deming anyone?

        Vote -1 Vote +1

      • Erik says:

        Sorry, I mean “error bars” in the statistical sense. I.e., http://en.wikipedia.org/wiki/Observational_error

        I guess what I’m looking for is some more explanation of the effect of strand rate, simply because I’m not used to seeing it as a tool to evaluate offense. So I don’t know if it’s a measure of luck or skill, or how large a, say, 2% difference in strand rate makes on the success of a team. Sticking error bars on the plot would be one way to address this.

        Vote -1 Vote +1

  3. Resolution says:

    I always thought Quadrant 1 was the top left quadrant, and then the numbering proceeded clockwise…

    Vote -1 Vote +1

  4. MDB says:

    I’ve been a little confused on what the strand rate actually tells us. Isn’t it a combination of luck and skill? For instance shouldn’t we expect the Braves and Reds to move closer to the average in strand percentage?

    Vote -1 Vote +1

    • I’m sure it is some combination of luck and skill, but over the course of a season, I’m thinking of it mostly as a skill. I’ll get into some different ways of measuring this next week.

      Vote -1 Vote +1

      • GTStD says:

        I think over the course of a season, strand rate is an ok measure of how well an offense performed overall; however, I don’t think that it represents a good measure of power. It seems to me to be a better measure of balance and consistency in an offense… of how well the 3rd and 4th hitters are able to perform in an inning.

        It may also be a decent way to consider luck… much like ERA gives an overall idea of how the pitching staff performed, but not a good idea of their true talent or context-independent performance. We can look at an ERA and an FIP and see how much luck and defense factored in, and see how well a pitcher’s results lived up to his potential. By the same token, strand rate feels like a good measure of overall performance, and by looking at the other peripherals, we can get an idea of how well an offense lived up to its potential.

        If we think of the strand rate as offensive efficiency, and the OBP as one contribution to that measure, the graph does give us some useful information. It tells us how efficient they were, and it tells us that they didn’t live up to their potential. It doesn’t actually tell us why, though you infer it is due to power. I think it would be more interesting to use several graphs and change the x-axis. Seeing one with OBP, one with ISO, one with K% and BB%. By comparing all of them, we should see which factors are most correlated, and for each team, see what their deficiencies were.

        For example, on the graph, Cincinatti and Atlanta aren’t that different in OBP, but horribly different on strand rate. I would like to see how they differed in ISO vs strand rate as well. If that is significantly different, we could say that power was a strong factor. If they aren’t, it would just be lineup balance and luck/sequencing. That would make the analysis much more interesting I think.

        Vote -1 Vote +1

    • B N says:

      Well, it’s going to have some luck but a lot of it is still going to be skill. Unfortunately, a part of that skill happens to be OBP however… If you’re getting less outs, it’s easier to drive runners in.

      Vote -1 Vote +1

  5. jscott says:

    Why use strand rate instead of iso for power?

    Vote -1 Vote +1

  6. Barkey Walker says:

    I’ve been thinking about your calling the OBP capital. I think this is very Marxist. He insisted that capital was just the dead labor. He spent a long time trying to prove that you could remove capital from the equation by insisting that it was only the labor that had been put into making it so that there was only one input–labor (and dead labor).

    Vote -1 Vote +1

  7. John says:

    estimate a c=d production function? why would you expect strand and obp to exhibit crs?

    Vote -1 Vote +1

    • walt526 says:

      A Cobb-Douglas only exhibits constant returns to scale if the sum of the exponents is equal to 1.

      Vote -1 Vote +1

    • Kevin says:

      I’m more concerned with a couple other qualities of C-D functions.

      Complementarity – With Strand % and OBP, you would expect a competitive result in output, not complementary or independent. This could be fixed by taking the RBI% (or whatever you want to call 100 – Strand%).

      Diminishing Marginal Returns Everywhere – This only applies if the exponents are less than 1, but I wonder if a cubic production function might be more representative of reality.

      Vote -1 Vote +1

  8. Wavaw says:

    I’m surprised how well the Dbacks did in team strand %. Even with their power the team set a record for most Ks in a season.

    Vote -1 Vote +1

  9. Oscar says:

    Yeah, this is fatally flawed.

    Vote -1 Vote +1

  10. In case people don’t know, I have these graphs laid out for every year in baseball history at http://www.baseballgraphs.com. However, I use ISO instead of strand percentage.

    Actually, I prefer the graphical layout I put into this article:


    …I think it better addresses some of the underlying issues you’re raising here. I used this format last year in THT’s Graphical Report.

    Vote -1 Vote +1

    • GTStD says:

      Take what I said before (I hadn’t read all the comments yet), and read that link and then be happy. :)

      Those measures layout more specific result with more drawable conclusions… hooray!

      Vote -1 Vote +1

    • walt526 says:

      Yeah, I’ve been mulling it over since Part 1 was posted a few days ago and for the life of me I can’t understand why the author is using Strand Rate rather than ISO as a factor of run production. I’m also not quite sure why he’s laying out a four-quadrant graph with league average at the origin as I’m not sure that the unconventional layout really adds anything to the analysis. Maybe his reasons for doing both will be made clearer in a forthcoming post.

      Vote -1 Vote +1

  11. Nick says:

    Both factors used are also directly affected by the park’s effect on HR rate. Notice the grouping

    If there were adjustments for neutral environment, might show a bit more ‘true’ skill? But I understand that teams play their games in the real world….anyway, just a thought

    Vote -1 Vote +1

  12. John says:

    Couldn’t “getting on base” easily be defined as labor?

    Vote -1 Vote +1

  13. DIVISION says:

    So, basically, my D-backs were the BEST of the “average” offenses in the NL.

    Not sure if that’s a backhanded compliment or something to actually applaud.

    Vote -1 Vote +1

  14. Mitchello says:

    Would it be possible to create indifference curves using this type of data? Maybe compile data from the past several years and use contour lines or something? The only thing is that I would make the strand rate descending instead of ascending so the IC would hold its traditional shape.

    Vote -1 Vote +1

  15. Phil says:

    Same post I put on the AL article:

    Let me start by saying I absolutely love economics. I majored in economics in college, and am now employed by the Foundation for Economic Education (yes I love the Austrian stuff).

    With that said this post is comical. You used microeconomic theory to show that teams who get the most runners on base and strand the least of those runners will score the most runs? DUH! You can replace “OBP” with labor and “strand %” with capital but really you just stated the obvious.

    Economics has its place in baseball. For example, micro-theory can accurately predict contract length/salaries for free agents. But don’t bring economics into an area of the game where it’s not needed just to try and look smart.

    Vote -1 Vote +1

  16. Ed says:

    Phil…you hit the nail on the head. Nice try, but economists have been doing much better work for decades.

    In the the academic sports economics literature they have been estimating production functions for about 30 years, but they are more sophisticated than what we see here. Generally speaking, the “outputs” are wins and the “inputs” vary but generally include things like hits, SB, HR’s, etc. The value of coaching is often measured by comparing actual wins to predicted wins for a given amount of inputs.

    If you are estimating a discrete dependent variable Jesse, you should use a Poisson model btw.

    Vote -1 Vote +1

  17. Adam says:

    I feel like strand rate might not be the best variable to use in these graphs. If you’re trying to show what certain teams lack, and how they can improve- ie. on-base ability or power- strand rate is a little bit obscure.

    I would like to know what a graph of on-base% vs. isolated power would look like compared to this graph. We know that power helps to drive in runs (and therefore lower strand rate), but so do singles. There’s simply no way we can say that “If you add a power hitter to your lineup, your strand rate will go down.”

    The on-base part of this article works completely: you can logically assume that if you get a guy that walks a lot and gets on base a lot, he’s going to improve your overall team on-base%. It just doesn’t work the same way with strand rate. Graph ISO vs. OBP and see where the teams lie. The teams in Q1 could drive in more runs by adding a power hitter. Teams in Q2 need both power hitters and on base guys. Teams in Q3 would benefit by adding a high OBP guy, and Q4 still represents the best offensives.

    Vote -1 Vote +1

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

Current ye@r *