about this section on less than one war players: are you introducing a problem by failing to account for the value of a replacement player? if we create a threshold for a replacement value player, that will take something off the strange results you found.

so, if 30 teams hire 90 replacement level players, the replacement value players still won’t work for free. they’ll cost at least the league minimum. which means that the players will cost $36M but won’t be worth a single win above replacement. and that would give you an infinite dollar value per war, because you’d be dividing by zero. but do we really want to say that it makes no sense to pay league minimum for a backup catcher, LOOGY, etc.? replacement value isn’t nothing — every team has to have a few: brad thompson, larue, etc.

If video killed the radio star, maybe sabermetrics killed the veteran bench player.

I’ve heard in the past (and it makes some sort of sense in my mind) that, because the division of talent in MLB isn’t even (it’s more of a bell curve with the majority of talent clustered around the 0-1WAR range) that there should be some sort of premium on high-win players. That is, because it’s quite easy to find 1-ish WAR players to fill any position (examples this off-season for fairly rare positions – Odalis Perez was signed on a minor league deal, Craig Counsell is still available and could presumably be had for <$1m) that the more valuable players (3-4 win guys, for example) should cost more per win than the scrubs. This should especially be the case when your team is high up the win curve (i.e. the Cubs should be paying more per marginal win than the Nats as their team is harder to make marginal improvements to).

However, it seems your data goes totally against that.

Perhaps the latest “market inefficiency” in MLB is superstar (i.e. 3-4 win+) players being underpaid! Who’d a thunkit?

I’m guessing that part of the reason is the scarcity of those sort of players – in any given offseason, there might only be 1 or 2 stars available in FA at any given position, so teams have been prepared to pay relatively big $ (unwisely) to guys who’ve had previous success but whose values have been depressed by one or two poor years, hoping to catch lightning in a bottle (see: Gagne, Eric, 2008, Renteria, Edgar, 2009). It’d be interesting to see the spread of the players who have been badly overvalued in the FA market in the last few years – I bet that a good % of the contracts will be of this type (although there’ll surely be plenty of plain bad contracts of the Carlos Silva variety out there – presumably these’ll become more scarce as the bad GMs get progressively removed from baseball).

I guess it’s also possible that this disparity somehow represents the fact that, previously, GMs have severely undervalued defense in a lot of cases. Again, it’d be interesting to look at the division of runs-above-replacement for the two groups of position players (the <1WAR group and the >1WAR group) and see if maybe there’s a defensive disparity there. Could be that a higher % of the <1WAR guys are poor defenders being overpaid for their good offense (whilst giving most of those runs back in the field). It’d certainly be one hypothesis to explain the disparity, though I doubt that it’s actually the reason behind most of it…

$4.5MM per marginal win is a huge cost, $2MM more than the league average of $2.5MM and WAY more than cost-controlled players make (nothing for pre-arb, and $1.3MM, $2.7MM and $3.6MM for 1st, 2nd, and 3rd year arb players, respectively).

So, given that you’re spending money on free agent, you should look to spend no more than $4.5MM per WAR. But if at all possible, don’t spend money on free agents. Sure, if the team’s going to compete and has a glarying hole somewhere, it might be worth it. But otherwise, invest the money in drafts, contract extensions for young players, and the occasional underrated free agent.

of the data concerns the likelihood of a player actually performing up to his previous year’s WAR in the first year of his new contract. For both the 2008 seasons and the 2007 seasons players in the first year of new contract had lower WAR values than in their contract walk years 65% of the time.

I don’t know whether that means players tend to play better in their walk years or they tend to struggle in the first year of a new deal, but it is just one more indicator that says caveat emptor. For example, the 2006 WAR value of the 2006-07 class was $3.8M, but it you factor in the actual performance of those same players in 2007 then clubs actually paid an average of $5.3M per WAR. For the succeeding year the values jumped from $3.4M to $6.4M since the aggregate WAR of that year’s class dropped from 105 to 55.9.

Does it really accurately measure, in the real world, what replacement level players do?

Does the statistical definition change year to year, depending on whether offense was up or down league wide?

And also does the presence of major leaguers who perform below replacement level indicate that what is termed replacement level is too high? In other words, does the theoretical “replacement level” player actually exist, or are there not actually not enough replacement level players to fill every major league roster?