May 28, 2010
Ahead in the Count
While last week’s article contrasting the cost of re-signees vs. the cost of other people’s players, or “OPP,” made a strong point that there is a difference between these two groups of players, many readers had questions about various issues, including hometown discounts, the performance of the two groups of players before the deals, and whether the decline was a matter of a decrease in playing time or production. In this article, I break down each of these factors and use them to learn more about the cost of other people’s players.
DISCOUNT FOR SIGNING IN ADVANCE?
A number of readers suggested that “hometown discounts” might explain the difference in cost between re-signed and players who were given contract extensions before reaching free agency. There are two ways this could happen. Either the players are giving discounts because they are willing to sign for less to stay where they are, or the players are giving discounts to teams who can sign them in advance to secure their futures. I had removed any players who had arbitration years bought out as part of their free agent deals to try to remove this bias, but I still left in players that had signed before the season had ended.
To check the effect of this, I simply compared the cost of players who re-signed before the end of last season of their contract had ended to players who re-signed after the season had ended and to players who signed with other teams. The answer appears to be that there is some noticeable discount for clubs associated with taking on the risk of signing their own players early, but that there is still a large difference between the cost of a win from re-signed players who had completed their deals and other people’s players. This is especially true for pitchers, so I have broken the deals down by both length and whether the player was a hitter or pitcher.
Looking at all players who signed two-year, three-year, and four-year deals collectively, the cost per win among extensions signed before the season ended was $3.53 million, on re-signings after the season ended was $4.07 million, and on all new signings was $5.71 million. In other words, there is a discount for signing in advance (15 percent), but it hardly explains the entire effect (51 percent). For hitters, it seems more pronounced as re-signing in advance netted a cost of $3.43 million, while re-signings after the season cost $3.81 million, and the new signings cost $4.27 million. Pitchers cost $3.66 million a win if signed before the season ended, while those re-signed after the season coast $4.53 million and other people’s pitchers $ 8.33 million. While there is a hometown discount given to teams who sign pitchers in advance, it fails to explain much of the difference in performance.
The reason that I do not believe that the hometown discount explains much of the effect is that the discount really seems to come in future years of the deal, rather than at the beginning. If we look at the cost per win of each length of deals, and break it down by hitters and pitchers, we get the following costs per win each year (assuming that the cost is considered to be the average annual value of the deal).
With two-year deals, we see that the cost is pretty similar in the first year of the deal, but that the cost really diverges in the second year.
The deals look better for re-signed players in each year, but the real divergence happens in the last year of the deal, where the cost goes from being not quite two times as high for OPP to being almost four times as high.
Since there is only one re-signed pitcher, these sample sizes are too small to deduce much information. However, they are presented below for the basis of comparison.
Looking at the entire set of tables above, it appears that the real cost does generally come towards the end of contracts, and largely because of the decline in performance during deals by other people’s players. Adding in the fact that the cost per win was not especially lower for deals that are signed in advance, I believe that the real difference in costs between re-signed players and OPP is about asymmetric information. Teams know more about their own players for sure, and it appears that this has enabled them to project their performance better.
DIFFERENT TALENT LEVEL?
One thing that concerned some readers was whether the players may have started from a different baseline in talent level. Thus, I present the change in WARP3 of each group relative to its own performance in previous years.
This should assuage some fear that the groups of re-signings and OPP were coming in with different baselines. OPP had a larger decline in WARP3 relative to their previous performance, in addition to evidence of failing to match projections or maintain value through deals.
DECLINING PRODUCTION OR DECLINING PLAYING TIME?
The question we must now ask is if the decline in OPP compared to re-signed players is a product of declining performance or decline in playing time. This is why I have separated hitters and pitchers and will compare the trend of plate appearances and TAv for hitters, and then the trend of innings pitched and ERA for pitchers.
While both groups of position players on two-year deals see a decrease in playing time at similar rates, the two-year re-signees were able to get their TAv nearly as high in the second year of their contracts as they had been before signing. Since players who sign two-year deals are in their mid-30s on average, this is a pretty clear accomplishment, and explains why bringing their TAv up to previous levels is so helpful in creating value on a dollar per win basis. Since seven of the 24 re-signees were catchers, their .247 TAv does represent a decent amount of value. (Other than a lot of the two-year contracts of re-signed players being given out to catchers, the positional breakdown among hitters was pretty similar for re-signed and newly signed players.)
There was a major drop in playing time in the second and especially the third year of OPP position players, while re-signed players played regularly throughout their contracts. Juan Encarnacion and Dave Roberts losing their jobs played an obvious role in this case. Additionally, three-year re-signees had TAvs that were at least as high in each year as they had been in the previous years. The last year of three-year contracts to OPP saw a sudden drop in TAv. David Dellucci and Jacque Jones were the primary culprits there. Carlos Guillen and Chipper Jones were the hometown heroes whose production turned their deals into major hometown discounts.
As always, breaking down four-year deals into such small sample sizes does not yield much information. Both groups did a good job of getting playing time throughout their deals, obviously staying healthy in the process, while neither group saw much of a decline in production either.
Overall for hitters, a lot of the difference in performance comes from decreased production rather than decreased wellness, though there is some evidence of both.
Moving onto pitchers, we first analyze two-year deals.
The most shocking difference here is the second-year ERA of two-year deals. After performing similarly before signing their contracts and even under performing a smidge in the first year of contracts, re-signed players vastly improved with a 3.44 ERA in their second year of two-year deals, while other people’s pitchers fell off a cliff. Guys like Joel Pineiro, Ryan Franklin, Dennys Reyes, and Jamie Moyer stepped up their games in the second year.
A decline in innings characterized much of the difference between three-year deals. Pitchers like Adam Eaton, Matt Morris, and Eric Milton lost their jobs in the third year of their contracts, ironically making the collective ERA of the three-year OPP group stay down towards the end of the deals.
Four-year deals again fail to provide enough of a sample size, as the only four-year extension among pitchers belongs to Tim Hudson, and he is only being compared to a group of five OPP. His Tommy John surgery is something that the Braves probably did not foresee, but a strong performance early in his deal did mitigate some of the value loss late in the contract.
Gathering all of this information, it does seem like a mixture of performance decline and health decline played a role in explaining the difference in performance. However, since these issues are so related, it is difficult to say anything too emphatically.
Taking all of these tables together with what we have learned in the last few articles, we can get a pretty clear picture of what is going on. Teams are getting far more production on a per dollar basis from re-signed or extending the contract of their own players than by signing other people’s players. The real difference in production seems to come at the end of deals, indicating that the primary cause of this appears to be asymmetric information, whereby teams are using intelligence they have about their own players and using it to determine who is likely to perform better. This seems to be a function of being able to detect both declining performance as well as health.
Recent discussion at The Book Blog about the most recent articles has led to some other research on the topic. Tom Tango has noted that the real difference in cost per win in my sample comes from pitchers, which makes intuitive sense as pitchers are more vulnerable to injury and sudden performance declines, making asymmetric information a larger issue. In the comments section, Shawn Hoffman noted that this effect carries over a much larger sample. He looked at multi-year deals that covered 1990-2009 and found that hitters cost 40 percent more per win when signing with other teams, as did starting pitchers, while relievers cost 70 percent more per win when signing with other teams. Thus, it appears that small sample size is unlikely to be plaguing the more detailed analysis of the 134 deals looked at above. I include a Google Doc of my data for those interested in extending this work, which I think would certainly be useful.
The main takeaway for teams is simple. When you are considering a player on the free agent market, be careful in determining why he is a free agent in the first place. Which team let him go? If a team that generally spends money and is competitive fails to sign a player who appears productive, teams should check his health and look for reasons why team expects him to decline. Similarly, when considering one’s own player’s impending free agency, be careful not to expect the grass to be greener on the other side. An attractive replacement may appear available, but perhaps it is better to play with the hand you’re dealt.
Ultimately, the hand one is dealt appears even more important in building winning baseball team than we previously thought. We know that players with less than six years of service time contribute about two-thirds of production in the first place, so drafting and signing amateur talent certainly must be a primary determinant in winning. Of course, as my recent research on Contract Service Time and Wins has shown, the correlation between production by players with at least six years of service time and overall winning is particularly high—in fact, it is markedly at higher at .63 than the correlation between production by players with less than six years of service time and overall winning at .45. However, this work would suggest that the most efficient way to get that extra production from players with more than six years of service time appears is to sign one’s own players who are likeliest to age well. Therefore, identifying and developing amateur talent yet again appears to be the key to winning baseball. The only catch is that you have to start paying that talent, and if necessary, pay even more than you have been paying.