Ahead in the Count: If... (05/10)
the archives are now free.
32 comments have been left for this article. BP Comment Quick Links
| |
|
|
May 17, 2010 Ahead in the CountThe Cost of OPP
After remembering the 1981 hit "Should I Stay or Should I Go" by The Clash with last week’s title on the same topic, we move forward a decade to a 1991 Naughty by Nature hit—and we introduce the money to the equation this time (if you’re down with that). In this article, I will show that players who re-sign with their clubs on multi-year deals provide far more bang for their buck than players who sign contracts with new teams. To review what I have done so far, remember that I started by showing that players who signed with their own teams aged better, as evidenced by the fact that they provided more of their value late in their contracts. In efforts to go through the data more finely for this article, I made some minor changes to the set of players for the sake of a consistent definition, but I reproduce these updated tables below. For the sake of space and awesome old school rap references, OPP refers “Other People’s Players.”
Here we saw that players who sign two-year and three-year contracts appear to age much better, and are probably superior bets. This is not true of four-year deals, but they are a small sample worth exploring, though difficult to draw conclusions. Despite the evidence that re-signed players aged better, it was certainly plausible that re-signed players were doing better later in deals, but worse early in deals, making the net effect a wash. I knew that Sean Smith found that those players were underperforming their CHONE projections by 16 percent, so I first checked whether PECOTA had the same problem, and also checked whether there was a difference between re-signed players and OPP. As it turned out, both groups were underperforming their PECOTA projections by equal amounts. This data set was different than the previous one, because I looked only at players who began their multi-year contracts in 2007-09, since this was the data for which I had good PECOTA projections. This analysis compared projected and actual VORP, because WARP3 was recalculated since these projections were generated, and I needed to compare apples to apples.
This topic became even more relevant, after I used the evidence that players who sign extensions with their current clubs as an indication that the Ryan Howard signing may turn out better than most people expect. Many readers asked whether there was evidence that some of the above results may be biased by age, so I checked and found that even within age groups, the effect of extended players aging better remained strong.
Again, we see that with two-year and three-year contracts, players of similar ages continue to age differently. They provide far more production throughout their deals, while players who sign with new teams instead drop in production rather suddenly. Of course, these subgroups are small enough that they do not individually suggest evidence of aging differences, but together they do show that age is not an explanation for this effect. Instead, players who are re-signed by their teams for two-year and three-year deals appear to age better than players signed for similar terms by other clubs. Although the evidence appears to be that these players are no worse at reaching their PECOTA projections in the first year, and clearly better at outperforming their PECOTA projections in subsequent years, it may be that teams are not actually getting better deals. If teams are re-signing their players for more money than their PECOTA projection suggests, than there is no difference in production. In that case, it is an issue of the limits of computers. Thus, I checked if teams were paying more for production from other team’s players than they are for their own players' production. Below I list the players involved to give a sense of the talent that was available, and also allow further work on this pertinent and surprising issue. First I list those players who signed two-year contracts.
Looking through the two-year contracts, we see a number of guys who seem to fit the mold of either risky players whose teams correctly determined that they would age well, and a number of players who seem to fit the mold of teams getting caught fooling around with OPP. Andruw Jones at 31 years old seemed like a good bet to rebound from the outside. After putting up matching TAvs of .301 in 2005 and 2006, Jones had fallen to a .257 TAv in 2007 on the back of a .242 BABIP while still hitting 26 home runs. PECOTA saw him rebounding to .278, but the Braves did not like what they saw and opted to go with Mark Kotsay in center field instead. Jones’ TAv fell all the way to .175 with the Dodgers in 2008. Geoff Jenkins had played his entire career for the Brewers but they did not re-sign him after 2007. His TAv had only gone down slightly from .274 in 2005 to .268 in 2006, and PECOTA saw him splitting the difference at .271 in 2007. Instead, the Phillies saw him fall to .241, as he lost his role as the lefty half of their right-field platoon in 2008 and they were on the hook for $13 million either way. Of course, when the Phillies signed Jamie Moyer to a two-year contract as a 44-year old in 2007, they were criticized far more, but his 3.71 ERA in 2008 helped them win a championship. Relievers Ryan Franklin, Trevor Hoffman, and David Weathers were pretty great deals on the dollar when they re-signed with their teams as well. Looking through these 69 contracts, we get the following data:
Teams who sign other people’s players are paying 2.29 times as much for a win as teams who sign their own players! This effect is similar in each of the three years tested. Moving on to the players who signed three-year deals, we get the following players:
The Braves kept up with the right Jonses—they dropped Andruw, but re-signed Chipper to a three-year deal for 2006-08 in which he produced 21.0 WARP3 at a cost of $37 million. PECOTA had seen him starting the deal as a 34-year old with a .303 TAv, but he posted .334, .347, and .365 TAvs during the contract and provided the Braves a lot of bang for their buck. Brian Giles and Carlos Guillen provided great values in extensions as well. On the other hand, Jason Schmidt was paid $47 million over three years by the Dodgers to sit on the disabled list and not contribute at all. Adam Eaton and Gary Sheffield were poor values as well. This data is summarized by the following table, showing the same effect:
For three-year contracts, teams are paying 2.1 times as much for other people’s players on a per win basis! This is a shocking discrepancy, and should not be taken lightly. It must be that either players are taking extraordinary hometown discounts (and it is hard to imagine that this is true on such a large scale), or there is a severe market inefficiency. This inefficiency does not appear to exist for players who sign four-year contracts. There are only 19 such players, and only six re-signings, so most likely this is just a sample-size issue, but here are those players:
The four-year contract do not really show much of a clear pattern. Of course, with only six extensions in the set, it is difficult to discern much of a pattern. Garret Anderson and Jose Vidro drag the extensions down, while the Johnny Damon and Ivan Rodriguez deals make you want to get down with OPP after all. The summary statistics are as follows:
Clearly, this is the exact same dollar cost of a win and there is no evidence of this effect among players who sign four-year deals. Regardless, there is overwhelming evidence that re-signed players and newly signed players are not aging similarly and are being compensated very differently. Please look through the list of players, and discuss in the comments any issues you would like to see addressed in future research. It may be that the re-signed players have something else in common, but it does not appear to be age or anything that teams are aware of when they award heavy compensation to other teams’ free agents. When you know your own free agents better than other teams' free agents, you can better determine which players will be worth their market rate, and pay them. For their part, players who are safe bets to age well should be asking more from their teams, all the while justifying this by singing “yeah, you know me.”
Matt Swartz is an author of Baseball Prospectus.
|
Great stuff, and thanks for posting all the research. This is just how I like to see this done, and this will be fun to go through.
Two pieces of information that I would like to see:
1. the number of PA or IP in the year prior to signing
2. the WARP in the year prior to signing
The reason is to see if there is a bias there, that perhaps the overpaid players are the scrub players, not the star players. This is critical.
Just for fun, I'll try to approximate that. I took your list of 48 players that signed 3 year deals. If in any of the three following season they had any season with at least 2 WAR, then I will *presume* they were one of the good players in the season prior to resigning (again, I'm ONLY using this as a proxy, until Matt can generate this critical piece of data).
We had 20 players who signed a 3-yr deal at the age of 30 or under. Of those 20, 13 were "good" (based on the above definition) and 7 were not.
Of these 13 good players aged 30 or under, 8 resigned with the same team, and 5 signed with new teams. The 8 same-teamers averaged a 21.5MM$ deal, for a total 3-year WARP of an average of 6.5 for each player. That's 3.3MM$ per win. The 5 new-teamers were paid an average of 25MM$ and generated an average of 6.0 WAR. That's 4.2MM$ per win.
Obviously, this is only 8 and 5 players, so, we've got huge uncertainty bands here. But, this is the way I'd like to see the analysis progressing.
The focus should be on players who are considered good and who are no older than 30 (maybe 31). Under this more focused view, though with limited sample, Matt might be showing something interesting.
More work needs to be done.
Tango, it appears you've selected on the dependent variable. Put it this way-- if you remove the OPP guys who failed to produce after the deal was signed, you're eliminating a lot of data that could be the exact evidence of what we are looking for.
Why is the fact that Adam Eaton and Jason Schmidt were disastrous three-year deals a lack of evidence that the Rangers and Giants knew something that the Phillies and Dodgers didn't get????
The point here is that the money is a proxy for how good they were before the deal. Eventually, I can dig up the WARP in previous years (if anybody has added to the data sets I provided above, please let me know), but the money they were signed for is the real value. I'm in agreement that this work isn't done, but the cost per win is really strong evidence that these groups are dissimilar for the reasons mentioned.
What we have here is evidence that it cost more to get wins from OPP than your own re-signed guys. Regardless of how good or bad the players were beforehand, the players allowed to reach public bidding were real bad contracts on average.
"Tango, it appears you've selected on the dependent variable. "
Matt, I obviously didn't make myself clear. Of course you CANNOT do what I did. I said this:
"again, I'm ONLY using this as a proxy, until Matt can generate this critical piece of data"
I shouted that out! Therefore, there's no reason for your first two paragraphs, since I already conceded that you can't do what I did.
***
I'm trying to show if there's bias in the data. I would like to have the previous season (or PECOTA actually) WARP and PA or IP. You are suggesting that a good proxy of that is the salary signed. That's a good enough suggestion (though that is still a proxy for what we want).
If we break them up the 48 3-yr players in thirds, we can put the 16 guys who signed for more than 24MM in the "great player" pile, the 16 at under 16MM in the "average player" pile, and the other 16 in the "good player" pile.
Here's what we have:
GREAT players
8 same-teamers, 29MM$, 8.2 WARP, 3.5MM$ per win
8 new-teamers, 31MM$, 2.1 WARP, 14.8MM per win (!!!)
Six of the 8 new-teamers were pitchers, only 2 were position players.
GOOD players
4 same-teamers, 23MM$, 6.4 WARP, 3.6MM$ per win
12 new-teamers, 19MM$, 2.7 WARP, 7.0MM per win
Eight of the 12 new-teamers were pitchers. Three of the 4 new same-teamers were pitchers.
AVERAGE players
3 same-teamers, 14MM$, 7.0 WARP, 2.0MM$ per win
13 new-teamers, 12MM$, 3.2 WARP, 3.8MM per win
Yes, very damning, if the salary paid can be used as a proxy for talent.
***
It seems to me that pitcher v nonpitcher deserves its own breakdown. I count 28 pitchers, 20 non-pitchers.
PITCHERS
8 same-teamers, 25MM$, 5.5 WARP, 4.5MM$ per win
20 new-teamers, 20MM$, 1.8 WARP, 11.1MM per win (!!!!)
NON-PITCHERS
7 same-teamers, 23MM$, 9.7 WARP, 2.4MM$ per win
13 new-teamers, 18MM$, 4.4 WARP, 4.1MM per win
First thing we see is: pitchers are way overpaid. So, that's going to be a huge source of bias, and therefore, if you have any subsample that doesn't have the same pitcher to non-pitcher split, that's a bias.
Even so, we still show a huge split. So, this still goes toward Matt's point.
Chipper Jones is one of the 48 players, generating 21 wins and being paid only 37MM. Now, it's certainly not like the Braves knew something that everyone else didn't here. But, any sample that Chipper is a part of will end up looking really really good for that sample.
That said, the most striking thing in the sample, and this goes toward Matt's point is this: of the 8 pitchers who generated the most wins, 6 did it with the same team, and 2 with new teams. Of the 20 pitchers who generated the fewest wins, 18 did it with new teams, and 2 with the same team. Here's the list of pitchers, with an "x" meaning same-teamers:
x Jason Isringhausen
x Livan Hernandez
x Brad Penny
Jason Marquis
A.J. Burnett
x Ryan Dempster
x Kelvim Escobar
x Freddy Garcia
Braden Looper
Bob Howry
Chad Bradford
Jon Lieber
x Vicente Padilla
Tom Gordon
Kyle Farnsworth
Scott Eyre
Danys Baez
Jamie Walker
Miguel Batista
Scott Schoeneweis
Armando Benitez
Matt Morris
Esteban Loaiza
x Odalis Perez
Eric Milton
Jaret Wright
Jason Schmidt
Adam Eaton
Notice the clumping.
The non-pitcher list certainly looks much more random than this, though a bit of clumping at the bottom:
x Chipper Jones
Rafael Furcal
x Carlos Guillen
Omar Vizquel
x Randy Winn
Mark DeRosa
Bengie Molina
x Brian Giles
x Geoff Jenkins
x Melvin Mora
Jacque Jones
Aubrey Huff
x Michael Barrett
Shawn Green
Alex Gonzalez
Frank Catalanotto
Dave Roberts
Adam Kennedy
Juan Encarnacion
David Dellucci
So, I think Matt has a great point, at least as it pertains to pitchers. And, I think, it's easy to explain that. Pitchers get injured far more, their talent hinges on their arm, and the teams know far more on their own pitchers.
I will disagree with Matt with regards to Ryan Howard, for three reasons:
1. Howard is not a pitcher
2. Howard is young
3. Howard was signed two-years out
Therefore, his discovery here, which looks great, won't apply very much, if at all, to Ryan Howard. Basically, Howard's representative group is so extremely limited, that it's hard to look at these results to try to infer something about the Phillies and Howard.
Matt's research stands well so far, on its own, and it doesn't need any tie-in to Ryan Howard to carry weight.
I like the idea to look into hitters and pitchers separately! I forgot about that. I had started with a few splits into this issue in the first article on this topic a couple months ago, but I should have checked it again when it came to money. Thanks for breaking it down.
I do agree that Howard isn't the best example of a player who is likely to age differently than others. I think the only reason that he would be an example of the "information asymmetry" issue is because of the supposed workout plan. However, Howard was the big extension contract du jour a couple weeks ago, and I mentioned it in the article as a reason. I hate hitching my analyst wagon to the "Howard deal is good" simply because I said "it might not really be that bad." I'm still not thrilled about it as a fan of the team. Still, the comment about not needing any tie-in to Howard isn't really relevant. The Howard article was a great way to re-broadcast this research which got burried without much comment until Howard was signed. I'm a practical guy, and I knew that had a good strong enough relationship to my result that it was worth re-stating clearly.
Chipper, on the other hand, is an injury-ridden superstar. Clearly that makes him a better candidate to be a bargain based on information asymmetry. This is all the more relevant because it's the Braves, who are by all reports a tremendous scouting team who does the due diligence on these type of issues. If the Braves think Chipper's unhealthy after his deal is up and don't re-sign him, I sure don't want him.
Focusing on the 19 guys with the 4-yr deals, 7 were pitchers, and 12 non-pitchers.
The 7 pitchers had only one pitcher resign (Hudson, 47MM$, 11 WARP, 4.3MM$ per win). The 6 pitcher who were new-teamers averaged 39MM$, 7 WARP, 5.6MM$ per win. Again, pitcher inefficiency once more.
The 12 nonpitchers had 5 same-teamers (38MM$, 8.9 WARP, 4.3MM$/win) and 7 new-teamers (40MM$, 11.6 WARP, 3.4MM$/win).
This last one goes against what Matt is saying.
***
Merging the three year and four year players, this is what we get:
35 pitchers, 25.5MM$, 3.8 WARP, 6.7MM$ per win
32 nonpits, 27.0MM$, 7.8 WARP, 3.5MM$ per win
So, a huge source of bias is how much pitchers are getting paid. This should come as no surprise to anyone.
The breakdown of pitchers:
9 same-teamers, 27.6MM$, 6.2 WARP, 4.5MM$ per win
26 new-teamers, 24.9MM$, 3.0 WARP, 8.3MM$ per win
The breakdown on non-pitchers:
12 same-teamers, 29.4MM$, 9.4 WARP, 3.1MM$ per win
20 new-teamers, 25.7MM$, 6.9 WARP, 3.7MM$ per win
11 same-teamers, 28.7MM$, 8.3 WARP, 3.5MM$ per win (excludes Chipper Jones)
***
The strongest conclusion is: pitchers are severely overpaid relative to non-pitchers.
The second strongest conclusion: teams signing someone other than their own pitchers to long-term deals are taking a huge risk.
Thanks Matt for the data. This second one really surprised me to the extent it's happened.
Interesting. I didn't think about the four-year deals being biased that much, but the breakdown of pitchers and non-pitchers is perfect. It's the same result after all!
I think this only says that pitchers are overpaid if the replacement level is perfect for both. If the replacement level for pitchers is overestimated or if the replacement level for hitters is underestimated, it may be fine.
One thing that this makes me think is that maybe there is a rational basis for teams over-drafting pitchers. They seem to be such bad bets based on Sky Andrecheck's draft pick analysis, consistently producing less than hitters drafted in similar rounds. Hitters on the free agent market being better bets than pitchers might make reevaluate this. I guess it would depend on vacancy availability to sign hitters. i.e. how many lineup spots are replacement level for a given team vs. how many rotation spots are?