Wednesday, February 03, 2010

The Experiment: If you build it, will they win?

As long-time readers will know, I'm a great admirer of the "Moneyball" strategy of acquiring undervalued players. At the time of Moneyball, most of these players (at least the ones successfully targeted by Billy Beane) fit into the high on-base percentage, low batting average category. There still are deals to be had on some of those players, but most major-league teams (except maybe the Kansas City Royals) have realized the importance of OBP, making it harder to exploit that particular market inefficiency. However, there are still inefficiencies out there, and some of them can be spotted with sabermetric research. This suggested an interesting train of thought to me. In the last experimental post, I took a look at how Madden users might fare against NFL coaches in play-calling and clock management. Now, I want to see if a baseball fan with a passing knowledge of sabermetrics like myself could put together a better lineup than experienced MLB general managers.

Imagine this hypothetical; it's the winter meetings of 2008, and Major League Baseball has allowed my team to compete as a National League expansion franchise, known as the Surrey Scorpions. They've allowed this only because I waved a massive stack of cash under Bud Selig's nose and he couldn't resist. We can now go ahead and put a roster together, subject to some conditions we'll discuss later. Constructing a roster is a time-consuming process, though, so we'll work on putting together our eight starting position players for now and look at pitching and depth down the road.

Now, it would be simple to just pick the players with the highest batting averages or home run totals, but looking at those stats alone overlooks other crucial elements of the game, such as walks and defence. At the end of the day, a major league team's goal is to win games, not batting titles or home run derbies, so we have to look a little deeper. Perhaps the best measure of a player's overall contributions to his team comes from FanGraphs' Wins Above Replacement (WAR) stat. For positional players, it looks at how many wins their batting and fielding skills produce relative to a hypothetical "replacement player" (i.e., an easily attainable minor league callup), with position factored in (there's an excellent explanation of how this works by Jeff Aberle at Behind The Boxscore). This is useful for this kind of analysis because it takes almost everything into consideration and melts it all down to a simple stat expressed in terms of wins.

One popular misconception about sabermetrics is that they're firmly against the traditional ways of analyzing the game. Of course, as with most sabermetric stats, the names at the top are still mostly recognizable superstars. Here's a list of the top players at each (non-pitching, non-DH) position by WAR in 2009.*

*Players with higher WARs than those below them, but lower WARs than the top players at their position are included for the sake of completeness and marked with a *. Additionally, I treated the corner outfield positions as equivalent, as many players move between right field and left field over their careers, and who plays in which slot is usually determined by team need. I've ignored designated hitters to simplify this, so all lineups are eight-man. Salaries are taken from the "base salary" category at MLB Secrets.

And here's a list with just the top player at each position:

As you can see from the players included on that list, it would be completely unfeasible to assemble this team. Some of these guys are extremely expensive veterans, such as Matt Holliday, Albert Pujols and Derek Jeter. Others are emerging young stars who aren't making a ton yet, but certainly wouldn't be traded by their teams for anything less than a king's ransom; this category includes Ben Zobrist and Evan Longoria.

That leaves us with two constraints to consider; cost and tradeability. First, let's take a look at cost. Obviously, finances play a huge role in building a successful MLB franchise. A team that generates a lot of revenue, such as the Red Sox or Yankees, has a much easier time going out and getting the top players via trades and free agency without worrying about their salaries. Money doesn't automatically equal success, though. As teams like the Mets and Mariners have shown over the past years, it's still quite possible to spend a lot of money without creating a good team. Tom Verducci did a great analysis of teams' payrolls, regular-season wins and postseason acheivements at Sports Illustrated earlier this month; I want to take that a step further by looking at how much various MLB teams spent on their everyday players and how many more wins that netted them than just fielding a full team of replacement players for an entire season. Aberle estimates a complete team of replacement players would win 49 games and cost $11.2 million, so you can figure that each WAR the starting position players can add would add a win for that team.

Thus, the ideal lineup shown above has a combined WAR of 56.9, which would mean that they would be expected to win 106 games with replacement-level pitching. With good or even average pitching, they'd probably set a wins record. However, even with a few cheap young stars on there, they still have a combined salary of $69.7 million. That's not completely ridiculous on its own, as the Yankees (who we'll discuss later), spent $114.1 million on their starting eight, but the tradability concerns with the young, cheap players mean that this lineup still couldn't be put together in real life, especially with an expansion team like the Scorpions. If it was, though, it would have an efficiency rating (WAR/cost) of 0.81, which is pretty good; we'll get to some existing franchises efficiency ratings below. At the other end of the spectrum, the starting eight for a team composed entirely of replacement players would cost $4.4 million, and by definition, have a combined WAR of zero and an efficiency rating of zero.

Now, let's discuss the tradability concerns. One of the problems with many attempts at this sort of hypothetical analysis is they involve taking over an existing team and making a few changes that the author would like to see. These simulations can be useful, but they're limited, because part of the team's success will naturally come from the pieces that are always in place. Also, many of these situations involve proposed trades, and the problem with these is that you can't guarantee that any trade will be accepted by the opposing general manager (no matter how reasonable it may seem). Controlling for these effects from pieces already in place is the main reason why our hypothetical involves starting a completely new team with Major League players. Sound familiar? That's right; this will be the Springfield Nuclear Power Plant Team 2.0!


However, unlike Mr. Burns, we can't just go and grab all the best players out there (which is what the lineup above would be). There is a limit to Selig's generosity; he's refused to allow the Scorpions to partake in the draft for this year or sign anyone under contract to another franchise. Thus, we're limited to free agents. Here's a list of all the 2008-2009 free agents that we'll be working from. It's not particularly long or particularly good, but it's the best we've got.

Using this list, I've produced two potential batting lineups for the Scorpions. Here's the first one, which assumes that the team can afford a $100-110 million payroll for for the entire team (counting pitchers). Thus, this is the best available lineup in terms of WAR that can be constructed with solely free agents.

This lineup produces 24.5 wins above replacement, which is quite good, as we'll see later when we look at actual major league lineups. However, as you can see from the salary of $57.3 million and the Efficiency Rating of 42.8 per cent, it's not particularly good value for money. Most of that's thanks to the large salaries of Mark Teixeira ($20.6 million) and Kosuke Fukudome ($11.5 million). If we replace them with cheaper but less productive guys, we have a more efficient roster (and a more realistic one) that still produces very well. Here's what this looks like:

This lineup produces 18.9 wins above replacement for a cost of $26.2 million, giving it a much better efficiency rating of 72.1 per cent. Its cost is also quite reasonable, especially when you consider that it's generally more expensive to assemble teams out of free agents than in-house talent. As mentioned earlier, a complete team of replacement players would be estimated to produce 49 wins and cost $11.2 million. Thus, if we bring in these guys and keep our replacement pitchers and replacement bench players, we have a payroll of an incredibly cheap less than $37.4 million and a team that would be expected to win a respectable 68 games. For reference, that's similar to the Marlins' payroll last year and well below the 2009 payroll of every other team.

How do real batting lineups compare? I took a look at the WAR ratings and salaries of the (non-DH) positional starting lineups on 12 MLB teams to compare, including very successful teams like the Yankees, notedly efficient teams like the A's, big spenders like the Mets and noted sabermetrics disdainers like the Royals. WAR numbers are from Fangraphs, salaries are from MLB Secrets or Baseball Reference and the starting player at each position is from Baseball Reference's data, based on who started the most games for them last season. Here are the results.

And here are those ratings in a table:

Of the 12 teams analyzed, only three lineups (Yankees, Tigers and Dodgers) had higher total WARs than the efficient free agent lineup I proposed above, and all of those teams had a much lower efficiency rating than my lineup. Only the Yankees' lineup produced more total WAR than the best free agent lineup out there. The Marlins, Indians and Rangers were all more efficient than my efficient lineup, but produced fewer total WAR. When you consider the limited number of free agents available, the artificial restriction of not allowing trades or draft picks and the (generally) higher cost of signing, To me, that shows there are still undervalued players out there, which means that the central theme of Moneyball is still very much alive. However, the trick is still figuring out which players are going to do well each season. That's a subject for a future post.

(As always, feedback is welcomed, either in the comments below, via Twitter or Facebook, or by e-mail.)


  1. Great post, Andrew! Really interesting thought exercise, and as a past player of Baseball Mogul, I had always noticed how fungible the performance difference is between a lot of the minimum and low-price veterans and the overpriced mediocrities. As Bill Veeck said, "It's the price of mediocrity that kills you."

  2. Anonymous3:28 PM

    Ranked by difference between Total WAR+Replacement Wins & Actual 2009 Wins

    Remember that DHs and Pitching excluded.

    Team Difference
    Dodgers 22
    Yankees 20.6
    Marlins 19.7
    Rangers 19.5
    Reds 17.7
    Tigers 17.4
    Mets 11
    B Jays 10.5
    A's 9.1
    Royals 8.3
    Indians 1.4

    Teams at the top either out performed expectations by a great deal or had extra good pitching/DH production. Opposite for the bottom.

  3. Very interesting. Thanks for the info!

  4. Wow, this was amazing Andrew! Great work - cool to see how a replacement level team would fare against the rest of the league. Maybe in the next 5-10 years, the Scorpions could be a reality!

  5. Thanks, Ian! Glad you enjoyed it. It amazed me just how inefficient a lot of these teams actually were in terms of WAR, especially when cheap free agent alternatives were available. I'm betting that's going to improve as teams become more knowledgeable about stats, though.