Roster Construction Theory

Dave · October 12, 2009 at 10:56 am · Filed Under Mariners 

Before you ask, yes, the off-season plan post is coming. Patience.

Today, I want to spend a few minutes talking about a little bit of theory of roster construction that kind of gets lost when we begin to discuss the positives and negatives of certain moves. The concept is that of the escalating value of wins – is a four-win player more valuable than a pair of two-win players? Most people would say yes, and they base their conclusions on transactions around a paradigm that reflects that belief.

There is some real logic behind the theory. Regardless of resources, every team has to fill 25 roster spots. If you have a significant financial advantage, you don’t get to buy more players – you still only get to go into each game with 25 guys. So, in order to maximize competitive advantage, rich teams try to consolidate as much value as possible into the spots on a team that mean the most – the nine everyday players, three or four starting pitchers, and a couple of late-inning relievers. These 14 or 15 roster spots take up a bulk of the payroll, and the quality of a team is generally decided based on how much value you can cram into those 14 or 15 spots.

With a limit on how many good players you can have, the way for one team to beat another is to maximize value from those players. That constraint then serves to inflate the value of premium players. If you have a six win player at the same position your opponent has a three win player, he can’t make the two teams equal again by just adding a three win player, because those six wins cost him two roster spots. Your six win player and open roster spot is still a better arrangement for win maximization than his pair of three win players.

There’s also a scarcity of real premium players in baseball. There are perhaps a dozen or so guys who could be expected to put up six win seasons on a regular basis, so if you have one of them, the odds that your opponent also has one go down as well. Without a true superstar, you simply can’t build a roster that has as much upper-end potential. Even if you were able to fill out all 15 important roster spots with above average players and avoid any glaring holes in the 10 role player spots, you’d still be looking at a ceiling of something like 95 wins.

So, based on that, it should be fairly obvious that premium players are exponentially more valuable than combinations of lesser players, even with equal total win values, right?

Not necessarily. The potential upside is certainly higher, but value is not just derived by best case outcomes. In reality, value comes from the expected return and the inherent risk of not receiving that value. If you’ve ever invested in the market, you’ve certainly been told about the value of diversification, and how the best way to build wealth is to spread your investments around in order to avoid losing too much due to the failure of any one asset. The value of diversification is true in baseball as well, and works against the increasing value of wins for premium players.

Just as the upside of a team that has consolidated value into premium players is higher, so is the downside. If you have one six win player who hits the disabled list or suffers a significant drop in performance, you can lose the entirety of that value in a single blow. Ask the New York Mets about how quickly a few injuries can destroy a season – they lost double digit wins off their roster when Jose Reyes and Carlos Beltran missed huge chunks of the season this year. The fact that the Mets had two players of that quality was one of the main reasons the team was expected to be a force in the NL, despite some glaring problems with the rest of the roster. However, when they lost those guys, they took a massive hit from which they weren’t able to recover.

Premium players offer both more reward and more risk. For many people, they look at the increased reward and conclude that star players are more valuable, but they do not also account for the increased risk as well. Major League teams do, however.

The linearity of the value of wins is one that has been studied quite a bit by folks like Tom Tango (Mariner employee, noted statistical wizard). Over time, we simply do not see evidence that teams have been more willing to pay premium prices for premium players. Instead, teams have paid for wins on a linear scale at the current market rate. If you are a one win player in free agency, you’ve generally received the market rate for wins. If you’re a two win player, you’ve received double the market rate. Three wins, triple the market rate, etc…

Major League teams have looked the increased risks and rewards of consolidating value into a single player and decided that they’re basically an even trade-off. When faced with a choice of how much to pay for one six win player or two three win players, the dollar values are generally equal. The premium players get one tangible payoff that non-premium players do not, and that’s contract length.

The best players get really long deals – anywhere from six to ten years. Good but not great players get paid on the same linear scale, just for three or four years instead. In order to secure the rights to a premium player, teams are willing to give out longer guaranteed contracts, but not go significantly over the market rate in terms of wins added. If you’re a five win player, you’re going to get $20-$25 million per season, and all you’re really fighting for in negotiations is guaranteed years and perhaps some perks like a no-trade clause.

That teams see the risk/reward trade-off as mostly equal is interesting. I don’t think most fans do. You’ll remember that the people who liked the Erik Bedard trade from the M’s perspective were also those who were critical of the J.J. Putz deal – in both situations, they sided with the team that consolidated their value into the “best player” in the deal. There’s an old fantasy baseball cliche that you’ll still hear repeated from time to time – the team that gets the best guy wins the deal.

It’s just not that simple. Even if Erik Bedard had stayed healthy and pitched well over the last two years, the M’s would have still been big losers on that trade, because they paid a price far beyond what he was actually worth. They overestimated the value of having a star on the roster. The reward wasn’t as large as they thought it was, and they undervalued the inherent risks of tying your cart to one injury prone horse.

Premium players are great. But you have to be careful not to ascribe more value to them than they actually have. Building a team around a few really good players is one way to build a winning team, but it also introduces a higher level of risk than diversifying equal value over the entire roster. One way is not necessarily better or worse than the other. You can win with a core group of stars surrounded by competent role players (the 2009 Cardinals, for instance) or you can with a whole lot of good players and few great ones (the 2009 Angels).

Either way can work, as long as you execute it correctly and value players the right way. You can either bet on a few great players or a bunch of good ones, and both roads can lead to success. The key is correctly discerning how good a player is and what kind of value they really add.

The M’s hit home runs by diversifying last winter, turning one overvalued player into a bunch of undervalued ones. This winter, they might do the exact opposite, turning a collection of decent players into one better player. Just remember that, regardless of what kind of move they make, we still need to judge it on the merits of the individuals involved, and not on some big picture theory that doesn’t hold true. The team that gets the best player doesn’t always win the deal. One six-win player may not be more valuable than a pair of three win players. It all depends on the context.


30 Responses to “Roster Construction Theory”

  1. joser on October 12th, 2009 11:47 am

    This is a great foundation for the coming discussions, Dave. And it’s true that Fantasy Baseball has kind of infected people’s thinking about trades (and not just trades — it’s inflated the perceived value of stats like Saves and steals, and thereby distorted perceptions of certain players) but I think the tendency to focus on the best player in a trade, along with the urge to determine who “won” the deal, is just human nature — just another thing, like the dirt on Wille Bloomquist’s jersey, that we have to get past in order to do rational analysis.

    The point about stars getting longer deals (hello, Felix) is one you’ve made before of course, but bears repeating. I think that teams may be willing to do that in part because it offers a form of diversification as well: not diversification across asset classes (different players) but within an asset (the contract to a particular player). It doesn’t save you from the player suddenly getting old or losing his skills, but it does insulate you a little from random injuries. If you assume he’s going to have at least one partial season due to injuries, having more seasons of him at least gives you a shot at having a couple of injury-free years to balance it out. Of course having more seasons means there’s more seasons in which he might get injured. But instead of this being an all-or-nothing bet like you would have in a hypothetical one-year deal, it’s smoothed out into something that hopefully looks more like a normal distribution of risk (and reward), giving you more chances at getting all the other stars to align for that postseason run. Call it temporal diversification.

    Premium players offer both more reward and more risk. For many people, they look at the increased reward and conclude that star players are more valuable, but they do not also account for the increased risk as well. Major League teams do, however.

    As soon as I read this I thought “Good Major League teams do.” And the Bedard trade is a great example of how the M’s, under Bavasi, were not such a team.

  2. coasty141 on October 12th, 2009 11:47 am

    Would it be wrong to assume a team without premium players (6+ win guys), the management of the players would play a more critical role?

    If you have guys who are 6 win players it would seem as though you’d just sit back and let them play (they would have few weaknesses in order to compile those type of win totals). If you had a team full of 3 win players… would the manager need to make sure the players are being used to their max ability? Whether it is via platooning at the plate, being the proper pitcher for the ball park, or having a good outfield with fly ball pitchers on the mound, etc?

  3. Chris_From_Bothell on October 12th, 2009 11:58 am

    The key is correctly discerning how good a player is and what kind of value they really add.

    What’s the best way to do this for minor league players? I.e. how does one determine the WAR or potential WAR of minor league players included in a deal?

    I tried looking this up on and the sections of the user guide about minor league players made it sound like it couldn’t be done. E.g. the data from before the 09 season made Tuiasosopo’s defense and overall value sound horrific, unless I’m not reading it correctly (which is entirely possible).

  4. Rboyle0628 on October 12th, 2009 12:03 pm

    Very good post Dave, got me thinking a lot. Reflecting back, what were the 2001 Mariners? Were they more a bunch of 3 win players? I know we had Edgar, Olerud and Boone just to name a few. But were they a 3 win team with good managing like Coasty said or did they play more of a we know what were doing kinda deal?

  5. bat guano on October 12th, 2009 12:27 pm

    Great posts of late Dave. I’ll be interested to see how you put this one to practical effect in your offseason plan. I’m kinda hoping you tipped your hand with the “turning a collection of decent players into one better player” comment in the last paragraph. Prince Fielder or Adrian Gonzalez, please (we can dream can’t we?)……

  6. Dave Clapper on October 12th, 2009 12:38 pm

    Ugh. 2001, it appears, is the last year in which FanGraphs does not have WAR listings. Pulling up 2002, though, here’s what we get (and assuming that defense remains fairly consistent (which, yeah, I know, isn’t a completely valid assumption, but we’ll take what we can get)), I’ll throw in what the OPS was in ’01 vs. the OPS that got them the offensive part of WAR listed here (’02). And even though I’m sure LF was better than nothing, trying to piece together Al Martin/Stan Javier/Mark McLemore is more than I wanna take on.

    Mike Cameron: 4.8 (’01: .832, ’02: .782)
    John Olerud: 4.6 (.873, .893)
    Ichiro Suzuki: 3.9 (.838, .813)
    Bret Boone: 3.8 (.950, .801)
    David Bell: 3.3 (.718, .762)
    Edgar Martinez: 2.4 (.966, .888)
    Dan Wilson: 2.0 (.708, .721)
    Carlos Guillen: 1.4 (.689, .719)

    So, assuming defense was about the same and only offense changed, Cameron, Ichiro, Boone, and Edgar had higher WARs in ’01 than ’02; Olerud, Bell, Wilson, and Guillen had lower. All the lowers are by not much, looks like (except maybe Bell), and some of the higher ones were much higher (especially Boone).

    Looks like about six of the everyday offensive players were higher than 3, and some were significantly higher. Pitching? I dunno. But just this suggests that they were much more than a bunch of 3-win players (although they had plenty of those, too).

  7. bat guano on October 12th, 2009 12:53 pm

    Makes me wonder if there’s a significant difference between a team constructed for a 162 game season with lots of depth (i.e. a bunch of 3+ win players like the 2001 M’s) and one constructed for a long playoff run. My insticts tell me that having a couple of superstars on the pitching staff and in the middle of the order may be more important in the playoffs than in the regular season, although I guess the 2009 Cardinals might disprove that. Any thoughts Dave (or anyone else)?

  8. Dave Clapper on October 12th, 2009 12:57 pm

    And, looking at Ichiro, 2002 appears to have been a bit of an outlier for him defensively, posting only a 4.7 fielding value. Assuming 2001 was in line with career averages, his fielding would have been 10.7. That difference alone is worth almost a full point of WAR.

  9. Mike Snow on October 12th, 2009 1:06 pm

    A bunch of 3-win players is not lots of depth, it’s just a bunch of good players. It’s not depth unless some of them play the same position. Also, superstars in the lineup are not more valuable in the playoffs than they are in the regular season. The one thing that can make for better playoff construction is superstar pitchers, you just need a team that can get them there and hope they don’t break down on you. Other than that, there’s no difference between building a team for a 162-game season or a 7-game series, it’s just that the latter is subject to a lot more random variation.

  10. hejuk on October 12th, 2009 1:11 pm

    Some thoughts on context:

    1) It seems like the riskier route is better for teams close to, but not favored to reach, the playoffs. The greater chance of a 6 win bump from a superstar is worth more to such a team than trying to avoid a zero win bump by acquiring two 3-win players.

    2) Availability of value at different win levels: if there are more bargains among below average players (0-2 wins), then it makes sense to acquire a superstar and surround him with 1 and 2 win players. If you have a 6 win player at market value ($30 million, say, taking 1 win = $5 mil) and two 1.5 win players below it ($5 million a piece, say), then you are paying less ($40 million) than you would pay for three 3-win players ($15 million x 3 = $45 million). The natural question, then: isn’t it in fact easier to fill out a roster with below average players making less than they’re worth (through trades, the minors, etc.)? And so wouldn’t getting a superstar be the right strategy in general?

  11. Dave Clapper on October 12th, 2009 1:22 pm

    Btw, we essentially had two 6-WAR players this year: Gutierrez was a 5.8 and Felix was a 6.9. Ichiro was a 5.1, and nobody else cracked 3.

  12. Shizane on October 12th, 2009 1:49 pm

    Boyle/Dave Clapper –

    That was my exact thought when reading this: what about 2001? I’m interested to see Dave’s answer.

  13. natebracy on October 12th, 2009 2:32 pm

    According to Rally’s historic WAR database:
    Boone = 9.3
    Ichiro = 7.6
    Cameron = 6.4
    Edgar = 5.5
    Olerud = 5.3
    F. Garcia = 4
    McLemore = 3.6
    C. Guillen = 3.2
    Bell = 3.1
    Moyer = 3
    4 in the 2’s, 3 in the 1’s, 9 other positives, 4 zeros, and 5 negatives.

  14. Dave Clapper on October 12th, 2009 2:59 pm

    Ooh, didn’t find that. Nice get. Here’s the link, for those who want it:

  15. Breadbaker on October 12th, 2009 3:49 pm

    That’s a great site. One interesting note is if you look up A-Rod, I think you’ll be surprised at what WAR considers to be his best season.

  16. Mike Snow on October 12th, 2009 4:39 pm

    Keep in mind that those WAR numbers are calulated differently, and can vary significantly from the WAR on Fangraphs. For example, Ichiro’s 2004 goes from 6.5 WAR to 8.1 WAR.

  17. rmac1973 on October 12th, 2009 4:45 pm

    Is it OK if I just wait for the English translation of this post?

    ::: blink, blink :::

  18. KaminaAyato on October 12th, 2009 5:22 pm

    rmac, it doesn’t seem too hard to understand, but here are the Cliff Notes.


    Question: One 6-win player vs 2 3-win players?

    Argument for obtaining a 6-win player – getting the most bang for your buck. Argument against – falling off a cliff if the player can’t play for some reason (i.e. injury).

    Argument for getting 2 3-win players – diversifying your “assets” in case one of them gets injured. Argument against – taking up an extra spot to get the same effect.


    Myth: Teams that trade for the big-name player and give up prospects are getting the better end of the deal.

    Actual: It depends. Sometimes teams are building their teams in different ways and the trade works out for both sides. Sometimes it fails because a team overpaid (see Bedard & Putz)


    Myth: Players with higher win values are paid a premium because they add so much to a team and take up only one spot.

    Actual: Players are paid for their win-values, nothing more. But instead of being paid a premium on each win, they’re given longer contracts.


    Overall Summary: There’s more than one way to skin a cat. You can build a team around a bunch of good players and some stars, or a team with quite a bit of stars and role-players. Remember to keep that in mind as we head into the off-season wheeling and dealing.

  19. Rboyle0628 on October 12th, 2009 5:57 pm

    Dave Clapper, interesting info about Felix, Guti and Ichiro! This leaves me slightly happy yet slightly disappointed. I’m not quite sure how to approach this one. How do you construct the roster to build that serious contender we all desire?

    I would hate to see the team part ways with Felix, but realistically he could bring enough prospects back to seriously kick us into gear in 2011. Or, most of those prospects turn out busts and we are left in a similar position without Felix.

    I sadly just don’t see a viable option for building around them. Ichiro is getting older, Felix is well, Felix, and Guti can be one of the best CF’s in the game. I said in another post, how about going after J.J. Hardy, and resigning Wilson to a cheaper deal if he takes it. Moving Lopez to 1st, seeing if Tui can play 3rd decently and letting Branyan DH, and use Griffey in a reserve designated hugger position. Hardy is a risk, can he return to previous form. If he even provides a 3 WAR player that is an upgrade. Opinions? Just a thought, not really looking at specifics just kind of a base thought if anything could work like that or could make sense.

  20. justme on October 12th, 2009 6:20 pm

    Sign me up for some diversified portfolio. At least until this team develops a strong farm system.

  21. Pete Livengood on October 12th, 2009 6:34 pm

    Mike (or Dave or anybody who knows) – How are the historical WAR values calculated differently on Fangraphs and

  22. Mike Snow on October 12th, 2009 9:49 pm

    Pete, I’m not sure of all the differences, so I’ll just highlight those I can identify. Fangraphs bases its fielding numbers on UZR, whereas simply estimates range from play-by-play data (which is what allows it to go back before we had zone-based hit location data). They also appear to be using different replacement thresholds, as Fangraphs gives Ichiro a pretty consistent 2 runs credit per season extra for the gap between replacement and average. I’m not sure what the basis is, because the theory for where they set replacement level sounds the same, but that translates into 0.2 WAR worth of difference. There are also big variations in the batting values. Both sites are a kind of park-adjusted linear weights, it seems, but I don’t see much information on how the weights are determined, so that’s about as much as I can explain (it would probably tax my math skills to explain more than this even if the information was provided).

  23. bookbook on October 12th, 2009 10:20 pm

    “how the best way to build wealth is to spread your investments around in order to avoid losing too much due to the failure of any one asset.”

    I’m not going to argue against your main point, Dave, but your example didn’t work for me:
    The best way to build wealth actually is to put all your eggs in a very few baskets if your goal is to be wealthy, and you’re willing to accept the consequences of poor performance. Arguably, this is exactly the situation in Major league baseball (especially before the playoffs were expanded). Winning 88 games doesn’t actually get you much closer to a world series pennant than winning 58. As the Indians have proven, sustaining long, gradual rebuilding phases is about as easy as maintaining a world beating offense of exclusively singles hitters.

    The M’s braintrust is better equipped than most to execute a stars-and-scrubs strategy (they recognize the hidden value needed to get almost-average production for peanuts). As you posted weeks ago, the team needs stars.

  24. bookbook on October 12th, 2009 10:40 pm

    “The premium players get one tangible payoff that non-premium players do not, and that’s contract length.”

    I hope this doesn’t seem like nitpicking, but the premium players absolutely are getting paid more per win added via the contract length mechanism. When you give A-Rod a contract that guarantees to pay him like a 6-win player for his age 39 season ten years in the future, it’s not because you think he’s going to defy aging and/or the vagaries of injury. It’s because you can’t just pay him $4 million per win for the total wins you expect him to produce and get the deal done. Heck, when you give any pitcher a 5-year contract, year 5 can be seen largely as a deferred signing bonus and applied to the $/win of years 1 to 4. Ok, that’s an overstatement, but clearly the projected wins in the out years for a 6-win player decay from the baseline. Yet, the contracts usually guarantee payment at a fairly steady level. The difference is deferred premium for the added value of the concentration of wins in the early years.

  25. 6-4-3 on October 13th, 2009 7:17 am

    I admit this is tangentially related at best to the topic, but [tangentially related at best to the topic]

  26. Chris_From_Bothell on October 13th, 2009 9:37 am

    Still can’t find best way to estimate WAR or projected WAR for minor league players, to estimate total value of package you get back in your average trade.

  27. Alex on October 13th, 2009 9:47 am

    It has definitely been shown, by Tango and others, that teams pay about $X per win for free agents, paying linearly per win. (Meaning that it isnt $X * W^1.5 or $X * W^2 or something like that). A team pays the same for a player that is expected to be a 6 win player, as they do for two 3 win players.

    My question is: Is this the correct strategy, or are the baseball general managers collectively making a mistake?

    If I had the option of a 6 win player and 0 win player, or for the same cost I could have two 3 win players, I would prefer to have the 6 win player. This is because I have a good chance of finding an undervalued 1 win player to replace the 0 win player with, who I can get for essentially free. (For example, see the Mariners acquisition of Ryan Langerhans). The club with a few stars and a bunch of replacement level players is easy to upgrade. The club with a lot of 3 win players is hard to upgrade. I would rather pay a lot for my stars, and then try to fill in other spots with undervalued bargains.

  28. Adam Guttridge on October 13th, 2009 10:56 am


    Well written post.

    However, the point about linearity in terms of $/win…. I’m not sure you’re correct there. Have you got a link for Tango’s work? I did some work (on my CPU in storage… yikes) that came to the conclusion that the inflation of 2004-2007 was driven largely by mid market guys…. as in, the Pujols’ and Manny’s and Jeter’s of the world did not experience nearly the amount of inflation during that period as did the Carlos Lee and Paul Konerko types. It’s an important point.

    Also, have you seen the Guttridge-Wang trade model?

    Eminently relatable stuff.


  29. mkale on October 13th, 2009 3:14 pm

    If you’re giving a greater number of years to a higher-WAR player, isn’t that taking on more risk? (That they get injured in year 2 out of 8 and their WAR gets cut in half or goes down to zero after that.)

    You’re committing ($/Win * Wins * #years) in both cases, so the two three win players are likely getting less total money than the one six win player.

  30. Mariners2620 on October 13th, 2009 10:30 pm

    The Mariners organization has already talked about the want to get faster. The Mariners 2001 lineup had alot of speed which obviously led to alot of success. I don’t know how they intend on getting faster this off season, but can’t wait to see what they pull off. It is easy to say just go out and grab Chone Figgins and Carl Crawford, because those are both areas in which we need to improve, with Beltre departing and Saunders not quite proven at the plate yet. Figgins and Crawford both offer exceptional defense along with a consistent bat, but they will both be looking for lots of money which the Mariners do not have much of in which they can just go out and throw around to a different variety of players. Get a trade for Escobar and Fielder and we are set until Ackley is ready to play in the bigs out in LF. I wish it was that easy, because that sounds like a solid team. Looking forward to your off-season play Dave.

Leave a Reply

You must be logged in to post a comment.