Notice: Trying to get property 'display_name' of non-object in /var/www/html/wp-content/plugins/wordpress-seo/src/generators/schema/article.php on line 52
keyboard_arrow_uptop

In February, I wrote an article evaluating multi-year dealsgiven out to players with at least six years of service time, and I discovered something interesting. I found that players who re-signed with their current teams aged better than players who signed contracts with new teams, and not by a small margin. This finding gained some extra attention (and extra scrutiny) when I used it to question whether the Phillies might not have erred as terriblyas sabermetricians had suggested when they extended Ryan Howard's contract for five years and $125 million last month.  The primary question that people asked was whether there was any bias in the ages of players who signed multi-year contracts with their current teams versus the ages of players who signed multi-year contracts with new teams.  In fact, there is some difference in the ages of these groups of players.

 

 

#

Avg. Age

Two-Year Deals' WARP3

Year 1

Year 2

39

34.1

Re-Signings

39%

61%

32

35.3

Signed w/New Teams

63%

37%

 

 

#

Avg. Age

Three-Year Deals' WARP3

Year 1

Year 2

Year 3

15

30.7

Re-Signings

36%

34%

30%

35

32.3

Signed w/New Teams

45%

39%

16%

 

 

#

Avg. Age

Four-Year Deals' WARP3

Year 1

Year 2

Year 3

Year 4

6

31.3

Re-Signings

20%

33%

29%

17%

13

30.9

Signed w/New Teams

29%

25%

21%

24%

 

While not a particularly large difference, the re-signed players were younger on average in the two-year and three-year deals, which were the contracts that showed major differences in aging.  Does this explain the effect?

TWO-YEAR DEALS

Let’s first focus on two-year contracts.  This was the group where the aging difference was the most lopsided, with the re-signed players actually improving during their deals.  I split this group of two-year contracts into different age ranges.  Most players who signed two-year deals were older, so it was difficult to get very large sample sizes.  I split them into a group of 14 players who were between 26-31 years old during the first year of their two-year contracts, a group of 23 players who were between 32-34 years old, a group of 20 players who were between 35-37 years old, and a group of 14 players who were between 38-47 years old.

 

#

Avg. Age

Two-Year Deals' WARP3, ages 26-31

Year 1

Year 2

9

29.0

Re-Signings

40%

60%

5

29.8

Signed w/New Teams

65%

35%

 

 

#

Avg. Age

Two-Year Deals' WARP3, ages 32-34

Year 1

Year 2

12

33.1

Re-Signings

41%

59%

11

33.4

Signed w/New Teams

46%

54%

 

 

#

Avg. Age

Two-Year Deals' WARP3, ages 35-37

Year 1

Year 2

13

35.8

Re-Signings

38%

62%

7

35.7

Signed w/New Teams

62%

38%

 

 

#

Avg. Age

Two-Year Deals' WARP3, age 38-47

Year 1

Year 2

5

41.0

Re-Signings

36%

64%

9

40.4

Signed w/New Teams

72%

28%

 

Even in very small samples, we see a very large difference.  Players who were re-signed by their current teams aged better than those players who signed deals with new clubs.  In fact, each re-signed subgroup actually improved during their contracts, while three of four sub-groups of the newly signed players regressed during their contracts.

THREE-YEAR DEALS

 

Does this hold for three-year contracts?  The answer appears to be yes.

 

#

Avg. Age

Three-Year Deals' WARP3, ages 27-30

Year 1

Year 2

Year 3

9

28.9

Re-Signings

35%

37%

28%

12

29.2

Signed w/New Teams

34%

43%

22%

 

 

#

Avg. Age

Three-Year Deals' WARP3, ages 31-34

Year 1

Year 2

Year 3

4

32.8

Re-Signings

36%

32%

32%

15

32.3

Signed w/New Teams

41%

48%

10%

 

 

#

Avg. Age

Three-Year Deals' WARP3, ages 35-38

Year 1

Year 2

Year 3

2

35.0

Re-Signings

40%

31%

29%

8

36.6

Signed w/New Teams

67%

17%

16%

 

Even in a small sample size, we still see the same effects.  Additionally, the older players who received three-year contracts appeared to drop off even more suddenly at the end.

FOUR-YEAR DEALS

Last time, we found that four-year contracts appeared to show little evidence of re-signed players aging better.  Does breaking this down into players in different age groups reveal an affect?

 

 

#

Avg. Age

Four-Year Deals' WARP3, ages 27-30

Year 1

Year 2

Year 3

Year 4

3

30.0

Re-Signings

15%

44%

30%

11%

6

29.2

Signed w/New Teams

22%

28%

26%

23%

 

 

#

Avg. Age

Four-Year Deals' WARP3, ages 31-34

Year 1

Year 2

Year 3

Year 4

3

32.7

Re-Signings

25%

23%

28%

24%

7

32.4

Signed w/New Teams

33%

24%

19%

24%

 

This does not appear to provide any evidence either way, as we are now dealing with such small sample sizes that we cannot find any effect beyond the noise. Of course, this does not mean that there is no such effect on four-year deals, but it is certainly does not prove that there is.

CONCLUSION

Recently, well-respected sabermetrician Mitchel “MGL” Lichtman has stated that if this effect were true, it would be the “one of the most interesting and significant things to come out of sabermetrics since DIPS.”  Of course, Lichtman also professed a profound skepticism of whether the results could be interpreted as true. 

To me, this is not a matter of true and false.  The results above are reflective of past events.  They happened.  Inasmuch as the contracts’ terms are reported to the media truthfully and WARP3 is computed in the way it is (and the results were the same for FanGraphs’ WAR), these results happened.  The teams that re-signed their players did think that their knowledge of the players indicated they would be good bets, and they turned out to be just that. Thus, it is true that re-signed players aged well, as their teams believed.  That does not imply that the future will necessarily show this same effect.  In fact, the very publication of this series of articles might change this outcome, as other teams might target players more aggressively if their know those players' current teams are bidding for their services.  All I can say conclusively is that this happened on average for the 140 contracts that ranged from two-four years long and concluded after the 2007-09 seasons. 

Whether it will continue is not clear, because it is a market outcome, not a baseball outcome.  When Lichtman alludes to DIPS, he is referring to a baseball outcome.  If pitchers are unable to exert much control over BABIP, that will remain true as long as pitchers are trying to get hitters out (of course, if pitchers minimized their FIPs, this might not remain true).  Once a market effect is discovered, it can be changed.  While this market inefficiency is at least based on inside information and therefore stands a larger chance to stand the test of time, general managers can react to the market and make different decisions about signing other teams’ free agents.

In fact, this effect may not have existed in earlier years as re-signings may have been based on factors that no longer control clubs’ decision-making process.  All we can say for sure is that these outcomes in the tables above are real for the years tested, and the effect appears to be large enough to suggest it is not random.

I will continue looking at this topic, and invite others to do the same.  Determining this effect is no small issue, and may be a huge step in understanding a side of baseball economics that is not based on information you can find on Retrosheet.  Sabermetrics has largely focused on aiding teams to make better decisions based on data generated on the field, but has remained unenlightened about data generated off the field.  This could be a significant bridge to that information.

Thank you for reading

This is a free article. If you enjoyed it, consider subscribing to Baseball Prospectus. Subscriptions support ongoing public baseball research and analysis in an increasingly proprietary environment.

Subscribe now
You need to be logged in to comment. Login or Subscribe
Richie
5/10
More good stuff, and very-well reasoned. Thanks.
acmcdowell
5/10
I may be being dense, but I'm having trouble understanding these charts. I can't figure out what these percentages are (WARP3 over the life of the contract?) and why we should care about percentages rather than absolute numbers.
bmmcmahon
5/10
I agree. It looks like the percentages are showing the % of the player's total contribution for each year of the deal. But what if, for example, the total contribution of the re-signed players is much lower than that of the players who left? It doesn't help the signing team much if most of the players' contribution came later in the deal if they're starting from a low baseline.
xenolith
5/10
I'm pretty sure this is showing what percentage of the total WARP for the life of the contract was amassed in each year. Say a guy had 15 WARP over a two year contract, but had a 10 WARP year and a 5 WARP year, then this would have shown 66% in the first year, and 33% in the second year.
mbodell
5/10
Totally, the percentage argument is not that interesting without the context of raw WARP both before resigning and after.

If each group had WARP of 4 on average before resigning and the group the team resigned put up WARP of 0.39 and 0.61 in years 1 and 2 while the group that signed with new teams put up WARP of 6.3 and 3.7 in years 1 and 2 I wouldn't concluded that the group that resigned with the team aged better, but your study would.

The split over the lifetime of the contract is not the most important thing by a long shot.
metty5
5/10
Matt,

As always, great stuff. I still question if the results would vary if you excluded players who signed extensions and only used free agents. Do you think you could have that variation come down the pipe?

Thanks.
J
TangoTiger1
5/10
Matt, interesting stuff. Can you publish your data, something like:

playerid,ageSigned,yearSigned,DollarsSigned,NumYears,WARP0,WARP1,WARP2,WARP3,WARP4
(for each year they were signed, with WARP0 being the WARP prior to signing)
joelefkowitz
5/10
Rather than attributing the results to the clubs having extra knowledge on how players will age, isn't it more likely that resigned players are more highly coveted players to begin with, and thus end up aging better? Isn't it well noted that elite players age better?

Instead of, or maybe along with, controlling for age, you could control for WARP in the years leading up to free agency.

Great work though.
DavidDSG
5/10
Matt,

It's hard to understand how meaningful these results are without seeing the actual WARP numbers. I'd like to echo the readers who are asking to see WARP for year 1, 2, etc. for each group.

- David Gassko
swartzm
5/11
Thanks to all those asking for WARP by year totals. I had forgotten about this. I'll do them in the comments, but I can post an unfiltered post if it's not clear.

AVERAGE WARP EACH YEAR FOR EACH DATA SET (N = SAMPLE SIZE IN PARENTHESIS):

TWO YEARS
26-31 YEARS OLD
Re-signed: 1.64, 2.48 (N=9)
Newly signed: -0.34, -0.18 (N=5)

32-34 YEARS OLD
Re-signed: 0.37, 0.53 (N=12)
Newly signed: 0.28, 0.33 (N=11)

35-37 YEARS OLD
Re-signed: 0.80, 1.31 (N=13)
Newly signed: 0.73, 0.44 (N=7)

38-47 YEARS OLD
Re-signed: 1.00, 1.80 (N=5)
Newly signed: 1.16, 0.44 (N=9)


THREE YEARS DEALS

27-30 YEARS OLD
Re-signed: 1.99, 2.11, 1.59 (N=9)
Newly signed: 1.13, 1.43, 0.73 (N=12)

31-34 YEARS OLD
Re-signed: 4.13, 3.73, 3.68 (N=4)
Newly signed: 1.17, 1.35, 0.29 (N=15)

35-38 YEARS OLD
Re-signed: 0.40, 0.31, 0.29 (N=2)
Newly signed: 1.96, 0.49, 0.48 (N=3)


FOUR YEAR DEALS

27-30 YEARS OLD
Re-signed: 1.47, 4.17, 2.87, 1.03 (N=3)
Newly signed: 1.50, 1.85, 1.77, 1.57 (N=6)

31-34 YEARS OLD
Re-signed: 2.27, 2.03, 2.53, 2.17 (N=3)
Newly signed: 3.89, 2.87, 2.23, 2.86 (N=7)



So, all in all, it doesn't seem like there is any real pattern to the overall quality of the players signed. The sample sizes obviously get pretty small, but on the aggregate, they were statistically significant, and nearly every one of the individual samples seems to point in the same direction. I'll continue to look for different ways to splice this data set and please make suggestions if you have any. Thanks for these comments.
TangoTiger1
5/11
Matt, thanks for the data.

Looking at the data, it becomes imperative to find matching pairs to avoid selection bias. For example:

26-31 YEARS OLD
Re-signed: 1.64, 2.48 (N=9)
Newly signed: -0.34, -0.18 (N=5)

Those newly signed are pretty much useless. I don't see what their WARP was in the year prior to being signed (which I hope you can also include), but I'm going to guess it's going to be also low. They were probably injured. Something about them explains how they can sign a two-year contract and become replacement level.

Now, if you tell me that BOTH groups were a WARP of say 1.80 in the year prior to signing, then that would be VERY interesting. (Again, presuming no health issues.)

Look at this data:
THREE YEARS DEALS

27-30 YEARS OLD
Re-signed: 1.99, 2.11, 1.59 (N=9)
Newly signed: 1.13, 1.43, 0.73 (N=12)

Again, there's no way these are two similar groups of players. The newly signed must have been alot of platoon players, while the re-signs are regulars. Unless of course you tell me that their WARP in the preceding seasons were both 1.80 or something, which I doubt would be the case.

And here:
FOUR YEAR DEALS

27-30 YEARS OLD
Re-signed: 1.47, 4.17, 2.87, 1.03 (N=3)

Clearly, 3 players, with that kind of jump, that means that there was one superstar season in there that skews it all.

***

If I do a weighted average of all the 27-30 year olds in your list, this is what I get:

21 players re-signed, with first year WARP of 1.77 and 2nd year WARP of 2.56. That's pretty interesting, for sure. The question is why. Is it because of insider knowledge? Was it that they were 2.10 players who DROPPED to 1.77 and then rose to 2.56? We desperately need more context, specifically their WARP in the preceding season. And PA (plate appearances) would be good too.

23 players signed with new teams, with a first year WARP of 0.91 and 2nd year WARP of 1.20. It looks to me that players who signed multi-year deals with new teams are alot of platoon players.

Given the small number of players (44 players 30 and under) who signed multi-year deals, a couple of big seasons could skew things.

So, more information please, more context, as noted above, and I think we can move forward better.

Thanks, and I love it when others do all the hard work and roll up their sleeves. Everyone benefits.
swartzm
5/11
In this article: http://www.baseballprospectus.com/article.php?articleid=10505, I looked for this. I got all the PECOTA projections for each group of players. The re-signed players underperformed their PECOTAs by 13% and the newly signed players underperformed their PECOTAs by 15%. That's basically the same. Sean Smith did that study without doing it by subgroup that showed CHONE over-projected free agents by about 15% too. So it seems like they are performing similarly on aggregate the first year. I think that's selection bias-- I discuss this in that article. I agree that I don't have perfect matched pairs. I think that's why it's important to aggregate to get some sense of these things.
TangoTiger1
5/11
Matt, I think it would be helpful if you can provide a data dump as I requested, so we can look for possible bias, and perhaps some of us can do our own studies based on the same dataset. Is this a reasonable request?
swartzm
5/11
The majority of this was from a proprietary data set that somebody gave me, and a lot of it was pieced together from that and other sources. I'm not sure I'm even allowed to give it away like that.
TangoTiger1
5/11
Matt, I presume you mean that the list of free agent players was the proprietary list that someone compiled for you? The actual salary they signed is possibly proprietary or possibly from Cot's Contracts? The WARP data is obviously from BPro. Did I get that right?

If so, then the list of players and salaries from 2006-2010 can be obtained here: http://sports.espn.go.com/mlb/features/freeagents?season=2006

The pain in the butt is always in the linking of the ESPN ID to the BPro ID. Obviously, you linked your source's ID to BPro's ID.

Anyway, I'm trying to understand how far you can give out the data within whatever arrangement you agreed to. It seems that for us to replicate and extend your work, someone would need to simply link ESPN to BPro IDs.
swartzm
5/11
Okay, it seems like giving out my spreadsheet won't be an issue. I don't have all the data you ask for but I do have names, ages, deal lengths, and WARPs. I'll try to put something together soon, maybe a google doc.
TangoTiger1
5/11
Wonderful, thank you.
studes
5/11
I think you're misinterpreting the 26-31 year olds who signed with new teams. Their WARP went from -.34 to -.18, but you have them providing more value in their first year in your percentage table in the article.

BTW, I also have a problem with this methodology, using the percentages to argue that players that re-signed with their teams "aged" better. Unfortunately, I can't really articulate why--I'll see if I can come up with something constructive.

OTOH, I find your general conclusion (teams that re-sign their own players get better results cause they know their players best) intuitively correct. It would be great to really "prove" it.
swartzm
5/11
I agree the 26-31 two-years are a useless sample in isolation actually. I should have said something about that, but I forgot when I was forming the tables. I decided to include them because they are part of the whole sample and I didn't want to be removing players from the sample. But Andruw Jones -2.0 WARP the first year of his Dodgers' deal is such a matzoh ball hanging out there that any way I did the ages, some group was going to be stuck with 31-year olds and they were going to age badly. And Abe Nunez was pretty ugly at 30 years old with -1.6 WARP the first of his two years w/ the Phillies too. In future articles, I'm going to make a point to avoid using percentages alone. I've gone back and forth using percentages and total WARPs, but I think I should use both.
TangoTiger1
5/11
I think what would be interesting is if you break it down further by WARP class in the preceding year. So, anyone with a WARP of 3+, WARP of 1.5 to 3.0, and WARP under 1.5. Something to make sure that the two classes you are comparing are differentiated only on the basis of being re-signed or not. As it stands, the bias in talent level might be fairly strong in explaining the discrepancies.

Or, include dummy parameters (for WARP class and age class), run a regression, and let's see what you get. I'd like to see the regression equation, especially if it shows a 0.5 or higher for "re-signed with same team".
rraymo1
5/11
How does the "hometeam discount" play into this? Did some teams try to resign their players but got outbid?
hotstatrat
5/11
Perhaps another way to approach this would be to compare the WARP3s to the average per annum salary of the contract.