Players Don’t Become Terrible at 30
One of the oft-mentioned reasons for the slow free-agent market this winter is that teams are thinking on the same wavelength when it comes to evaluating players. One of the tenets of this theory is that free agents are bad bets because of the aging process. As players age, especially after 30, they get worse on the field, and teams don’t want to get stuck with those decline years.
There is a whole lot of reason in that explanation for the offseason’s lack of activity. There’s also a little bit of faulty logic regarding the aging process, particularly when it comes to this year’s free-agent class and the two biggest names out there, Eric Hosmer and J.D. Martinez.
The first flaw in this argument is based on a misunderstanding of how clubs are compensating players. All teams — and especially the “smart” ones — know and understand that the final year or years of a free-agent contract are unlikely to be valuable in terms of strict wins-per-dollar calculus. It’s generally accepted that those “out” years are going to be mostly dead weight. Players are typically signed to deals for which the total guarantee is equally distributed over the course of a deal. The team isn’t paying an equal amount every year expecting metronomic production over the life of a contract. They expect to receive a surplus of value in the early years and a deficit in latter years. The hope is that the early years compensate for the latter ones.
Teams could, in theory, compensate players a greater amount at the beginning of a contract than at the end, but most clubs choose not to do this because, by spreading the payments out, they get to keep more money in the present, which is more valuable to them. If a team doesn’t want to add, say, a seventh year at $25 million to Eric Hosmer’s offer, it isn’t because they believe he isn’t going to be worth $25 million in that seventh year. It’s because they believe he won’t be worth extra $25 million over the first six years. The value in the seventh year is going to be close to zero in terms of expectations. That’s not the main point I’m trying to make in this post, but it does deserve a mention.
The main point of this post is to show how playing time and production has been divided by age group over the past 30 years. I went back to 1984 and looked at five-season rolling averages based on age group and broke position players out in three categories: plate appearances, WAR, and WAR per 600 plate appearances.
First, let’s look at how playing time has been divided up over the years. For every year on the graph below, it encompasses data from that year and the previous four — e.g., the year 2017 denotes the stretch from 2013 to 2017, while 1988 includes 1984-1988.
In terms of playing time, players in the traditional prime years (ages 26-29) feature the highest allotment of those present here. That has remained fairly consistent for a while, in the 35% to 40% range for all seasons. This makes sense: the players who populate this age group are typically at their best relative to their younger and older counterparts — and to the younger and older versions of themselves.
At the bottom of the graph, we see the oldest players (ages 34 and over) in baseball. The playing time for this group has seen pretty significant fluctuation over the years. In the mid-1980s, older players received about 15% of all plate appearances. That figure went down with the addition of four expansion teams 90s, then way up during the PED era. The figure has come crashing back down over the last few years.
What’s interesting here is that players who have just left their prime (age 30 through age 33) haven’t necessarily followed the same pattern as that slightly older population. When expansion came, the players in their early 30s actually got a greater slice of the pie in terms of playing time. That trend continued through the PED era, as well. Over the last few years, that group hasn’t lost much in terms of the gains made over the previous decades. Now, in fact, they’re near a high-water mark. Meanwhile, the youngest group saw playing time go down during expansion and the PED era, but has come back very strong over the last decade.
The chart above tells us about opportunities but doesn’t say much about performance. The graph below shows the share of WAR by age group.
Despite a change to the y-axis, this chart looks pretty similar the to the one above. A brief examination of the two graphs might lead one to believe that, on a rate basis, the performances of the different age groups have remained similar over the years, with playing time representing the major different in terms of on-field value. Please note the ranges of the axes here, though: while the playing-time graph begins at 10%, this second one begins at 5%. The older players are doing a whole lot worse.
We’ve actually witnessed pretty wild fluctuations over the years in terms of the quality of play provided by different age groups. Young players, however, have never been better than they are right now. There are several possible reasons for that. Perhaps teams are waiting until players are truly ready to play and have success in the big leagues. Perhaps clubs are doing a much better job of developing talent in the minors. Or maybe it’s a question of context: a lack of access to PEDs in the minors might have precluded younger players from their benefits, even as older players prolonged their primes by means of their use.
Here’s what the age breakdown just for 2017 looks like:
| Age | % of PA | WAR/600 |
|---|---|---|
| 22 and Under | 3.5% | 1.4 |
| 23 | 5.2% | 2.0 |
| 24 | 9.0% | 2.2 |
| 25 | 9.1% | 1.9 |
| 26 | 10.2% | 1.8 |
| 27 | 11.2% | 2.6 |
| 28 | 7.3% | 1.7 |
| 29 | 7.2% | 2.2 |
| 30 | 8.3% | 2.1 |
| 31 | 8.4% | 1.8 |
| 32 | 3.5% | 1.3 |
| 33 | 6.8% | 1.6 |
| 34 | 5.0% | 1.2 |
| 35 | 1.2% | -0.1 |
| 36 | 2.0% | 1.1 |
| 37 and Over | 2.0% | 0.0 |
While we’ve seen younger players improve relative to the members of other age groups, those working in their prime years (26-29) haven’t experienced much change. The group of players just past their prime has endured only a minor dip.
As for players age 34 and beyond, they have seen their production plummet. Older players have seen their playing time and their production both cut nearly in half since the height of the PED era. From 1999 to 2003, players 34 and over were responsible for 15% of position-player WAR. The last five years, they’ve accounted for just 6%; last season it was under 5%. Per 600 plate appearances, the oldest position players in the league produced only 0.8 WAR per 600 plate appearances.
Yes, players generally decline in their 30s, and it is certainly possible that the tail end of the aging curve is considerably more steep now than it was a decade-plus ago. But it’s also worth noting that players aren’t suddenly bad at age 30. Recent research on aging curves suggest we might have been too harsh with a player’s decline in the past. I’m not suggesting players aren’t getting worse in their 30s, but good players in their 20s are likely to remain good players in their early 30s. On last season’s leaderboards, for example, six of the top-20 players and 16 of the top-50 players were between 30 and 33 years old. J.D. Martinez is only just turning 30 and likely still has several very good years ahead of him. Eric Hosmer still has two more seasons in his 20s before he turns 30.
There are real concerns with Hosmer and Martinez. At some dollar amounts, the value isn’t going to be there. But both players also likely to play up to expectations for multiple years into a new contract. It’s possible teams are getting smart on long-term contracts, but two seasons ago, it was the Cubs and Red Sox who handed out the biggest contracts to a position player and pitcher, respectively; next season, meanwhile, the smartest teams in baseball are going to line up to pay all sorts of players into their mid-to-late 30s. There are a lot of reasons for the slow free-agent market, and avoidance of long contracts might be one of them, but there are probably other factors at work with this free-agent class that have a bigger impact.
Craig Edwards can be found on twitter @craigjedwards.
Thank you for this.
The trend to crap on anyone’s value when they get into their late 20s has gone overboard imo.
I don’t think it hasn’t gone far enough but is getting there. Long-term contract to aging players have proven over and over again to be a very bad idea. Short-term ones? Not so much. Instead of offering grandpa five or six years on a contract, teams should be offering him two or three. Grandpa still gets paid, he just doesn’t get to ruin a team’s budget for years and years after he’s completely washed up.
Problem is the luxury tax.shorter and higher aav deals would be nice but luxury tax is calculated on annual salaries.
That should only matter for teams trying to squeeze under the luxury tax threshold to max out for a run. Also note that while money now has more value for the club, it also has more value for the player. So I don’t see a theoretical reason why higher $ for shorter years doesn’t make sense. It may just be player preference for stability.
Has anyone ever been a grandpa while still playing in the big leagues? None of my kids that I’m aware of are old enough yet to make me a grandpa.
I am sure there have been active major leaguers who had grandchildren. The Griffeys and Raines’ played together as father-and son. I don’t think the sons had children, but it could have happened. Bobby Bonds was 18 when Barry was born. I wouldn’t be shocked if some of the long career pitchers were grandfathers, like Phil Niekro, Gaylord Perry, Nolan Ryan. Ryan was married at age 20 and still in the bigs 26 years later.
Stan Musial.
Good call on Musial. He was a grandfather for the last few weeks of his final season of 1963. He was only 42.
Not really…if you are talking “value”. An arb 3 player that is 29 is a much greater value than the FA contract he signs the next year for his age 30 season. A team is buying an asset that is more likely to decline, has to pay more for it, AND (most importantly) has to buy not just the age 30 season…but possibly the age 31, 32, 33, 34, and 35 seasons. If that player would sign a 3-year deal, he’d get a crazy high AAV compared to a 6 year deal….but players generally won’t do that.
If Chris Mitchell (or someone else) could make a KATOH for FA’s, this debate would be mostly settled and give us insight into which types of players continue to be productive into their mid-30’s.
Two words: Albert Pujols.
See: Votto, Joseph.
Even Votto, who is definitely producing good value and appears to be an exception to the Pujols Problem, shows the difficulty of signing players to long and expensive contracts. In other words, see: won/lost record, Reds.
How much of the steep decline in 34+ year old players from the early 2000s is solely due to Barry Bonds?
Sure, Bonds had an affect. So did Steve Finley. Significantly smaller for Steve Finley. There is also Moises Alou, Larry Walker, Rafael Palmeiro, Kenny Lofton, Craig Biggio, and several others. I remember during the “steroid era” there were a lot of older players, and it seemed reasonable to expect a player to continue playing well into his 40’s. I remember a 2007 game where the Mets had 4 guys in their 40’s in the starting lineup: Glavine, Conine, Alomar Jr., and Alou. I remember when Tony Gwynn got his 3000th hit at age 39, he said, “I’m talking 4000!” I don’t think anyone would be able to say or hear that with a straight face nowadays.
We aren’t dumping on Eric Hosmer because they think he’s going to decline after 30 (which he will), we are dumping on Eric Hosmer because he hasn’t been that good in his 20s.
Martinez is a slightly different story, but the five-year deal the Red Sox reportedly offered him snaps right in line with the concern that he decline after age 33.
>Martinez is a slightly different story
How so, exactly?
http://www.fangraphs.com/graphsw.aspx?players=3516,6184
Because if you want a good player now, signing Martinez will probably help you with that goal.
I agree that Martinez helps teams as a good player now, but I can also understand why teams believe that he’s a great hitter but are extremely reluctant to stretch beyond 5/$125 given some of the negatives (bad corner OF defense, injury history).
I think it’s also tougher for his market that NL teams have to look ahead and wonder quite what they’ll have in 3 or 4 years if the bat drops off some and the corner OF defense gets even worse, because they can’t just slide him over to DH.
Unfortunately your graph is including JDs 2011-13 seasons when he was a completely different player (hitter), see ISO and the multitude of launch angle stories written about him. Whereas Hosmer has displayed a similar skillset throughout his ups and downs.
Also, maybe I missed something that addresses this, but the analysis here screams “selection effect”. The only players who last until their 30s are ones were are still good, likely the best ones in their 20s.
If you want to look at this, you need to look at within-person change.
Agreed.
The Joe Mauer’s of the world continue to play even after they have declined %60+. The 0-1 WAR players who decline at 29 are simply out of the game.
Using a rate stat to measure performance compounds this misleading bias- of course WAR/600 averages for large samples, in any age group, are likely to hover around the average performance for all players who are selected to play.
For example, Jim Abbott’s average season was (not surprisingly) around average. This is partly because he was selected to play when young, when he approached average ability, and he stopped being selected to play as a veteran when he couldn’t meet the standard. That doesn’t mean that one-armed pitchers don’t struggle to perform- they do. It’s just that players selected to play often perform around the level of the typical players selected to play. The real difficulty one-armed pitchers have, with respect to MLB performance, is shown in their lack of playing time.
But to[i]sign[/I] a player in their 30s implies they are attractive to sign. The selection bias is acceptable and representative of the skills sought after in this article.
This article isn’t implying “sign anyone in their 30’s because they were good and will be good.” It’s implying “if you think you should sign someone in their 30’s, don’t worry about the decline as much as before. If he’s worth considering signing, he’s worth signing.”
Huh?
What Craig has done has showed that the WAR of players aged 30-33 doesn’t fall off much, if at all, by not taking into account sample attrition of people who weren’t good enough to hack it before.
This is like if you said “there is no meaningful decline in self-rated health between age 50 and 60” because you decided to ignore the fact that the sickest people died. The people who survived likely have had a fairly substantial drop in self-rated health, but the average for that age bracket doesn’t move because you’re essentially deleting the sickest people out of the pool.
The underlying point Craig is making with showing that WAR for those remaining in the majors from 30-33 is that if a player is good enough to stay in the majors, he’s good enough to sign. The selection bias actually benefits the study.
Of course, those who aren’t good enough don’t play, which isn’t seen here. But that’s besides the point – if your only reservation to signing a 30+ year old is age, this article says ignore that (through the WAR analysis).
Shouting selection bias because there is a nonrandom sample without understanding it’s impact is foolish. But by the downvotes I have and your upvotes, we have at least 9 people doing that.
“if a player is good enough to stay in the majors, he’s good enough to sign.”
Disagree, there is absolutely nothing in the article competently showing that a 30-33 year old player is worth signing.
Among many strange limitations, the author threw out catchers entirely from his analysis. He also did not consider fielding or defending.
He didn’t have to use a tiny sample heavily skewed by PED-users Giambi and Ortiz. There are endless ways to approach this research without a needlessly tiny sample. The author could loosen the comparisons to players within 50% of Martinez’s career PAs, or the author could keep catchers in the sample, or the author could use a larger RC+ range than just 120-140. For a serious analysis of whether such players are worth signing, the author could include fielding and defense.
Throwing catchers out of the sample and not using fielding or defense is ESPN level poor.
Which is a completely different selection bias issue we weren’t discussing.
We were debating the merits of a sample issue due to survivor bias as selection bias. I argue that it isn’t relevant here, because the survivors are the relevant sample to analyze.
Essentially, like I said already, if a player is worth signing, and would be signed for their 30-33 aged years, then their age shouldn’t prevent the signing, based on this article’s conclusions. I never said that a player should be signed because he will be 30-33.
We can discuss the legitimacy of the population Craig used, that’s a different discussion than we’re having, but if we assume that the population is legitimate, then the survivor bias has no influence on the results (which is my argument).
Absolutely none of that is true. That is not Craig’s point, that is not what this shows, and that is the exact opposite effect of what this selection bias would entail. Let’s walk through this.
In Craig’s post, this is what happens:
1) Players are generally best around age 27ish. This is not controversial.
2) Players who aren’t good at earlier ages drop out of the sample as they age. There are reasons for this, from the idea that they cost more to that they don’t have as much room to growth.
3) This means that players who were worse from age 26-29 are much less likely to be in the sample during age 30-33. This essentially deletes the worst performers from age 26-29 from the pool, raising the overall average from age 30-33.
4) Craig uses this to argue that the dropoff from age 26-29 to age 30-33 isn’t that steep, ignoring that fact that the pool of players from age 30-33 was a fundamentally better group of players from age 26-29 than those who dropped out, overestimating the value based on “age”.
Let’s work through an empirical example, with real life hitters (minimum 120 PAs in both cases). Here are age 26-29 hitters in 2010:
http://www.fangraphs.com/leaders.aspx?pos=all&stats=bat&lg=all&qual=120&type=8&season=2010&month=0&season1=2010&ind=0&team=0&rost=0&age=26,29&filter=&players=0
These players averaged 1.62 WAR
And here are age 30-33 hitters in 2014:
http://www.fangraphs.com/leaders.aspx?pos=all&stats=bat&lg=all&qual=120&type=8&season=2014&month=0&season1=2014&ind=0&team=0&rost=0&age=30,33&filter=&players=0
These players averaged 1.42 WAR.
Overall, the average player at age 30-33 has similar value to the average player at age 26-29, both here and in Craig’s graph. It is about a 0.2 decline, which in Craig’s graph is about 10% of the value and here is more like 15% of the value.
Now, let’s see what happens if we take difference scores for all players with more than 2 WAR in 2010 (aged 26-29), since those are presumably the players you think are “worth signing”. We’ll calculate difference scores for those players. And we’ll be extra charitable and not count players who didn’t reach the playing time cutoff because of injury or ineffectiveness in 2014 (we should really just give them zeros, but I’m lazy).
Name 2010 2014 Diff
Will Venable 2.3 0.8 -1.5
B. Phillips 3.8 1.6 -2.2
Carl Crawford 7.7 2.6 -5.1
Omar Infante 2 0.4 -1.6
Shane Victorino 3.2 0 -3.2
Jose Reyes 2.5 3.5 1
Justin Morneau 4.9 2.6 -2.3
Miguel Cabrera 6 5.1 -0.9
Cody Ross 2.1 -1.2 -3.3
R Weeks 6 1.1 -4.9
Joe Mauer 5 1.6 -3.4
Josh Hamilton 8.4 1.2 -7.2
Jose Bautista 6.4 6.3 -0.1
Adrian Gonzalez 4.4 3.6 -0.8
Corey Hart 3 -1.1 -4.1
John Buck 2.8 N/A N/A
Alex Rios 3.2 0.2 -3
Ryan Raburn 2.3 -1.3 -3.6
Kelly Johnson 5.4 0.6 -4.8
Nelson Cruz 4.9 3.7 -1.2
Angel Pagan 5 1.7 -3.3
Mike Napoli 2.5 2.4 -0.1
Shin-Soo Choo 6 0.2 -5.8
F Gutierrez 2.1 N/A N/A
Robinson Cano 6.3 5.2 -1.1
Martin Prado 3.9 2.3 -1.6
C Pennington 3.5 1.6 -1.9
Ryan Braun 3.8 0.8 -3
Geovany Soto 3.3 N/A N/A
David Wright 3.5 1.7 -1.8
J.J. Hardy 2.1 3.3 1.2
Chris Young 4.2 0.3 -3.9
Stephen Drew 4.8 -1 -5.8
Joey Votto 7 1 -6
Jed Lowrie 2 2 0
Nick Swisher 4.3 -1.8 -6.1
Prince Fielder 2.7 -0.3 -3
Russell Martin 2 4.9 2.9
Chase Headley 4.3 4.2 -0.1
C Granderson 3.9 1.2 -2.7
Brian McCann 4.9 2.4 -2.5
Ryan Hanigan 2.1 1.1 -1
Alexei Ramirez 4.1 3.2 -0.9
John Jaso 2.3 1.7 -0.6
Nick Markakis 2.4 2.5 0.1
K Kouzmanoff 2.5 N/A N/A
Casey McGehee 2.9 1.6 -1.3
Tony Gwynn 2 -1 -3
Ian Kinsler 3.3 5.2 1.9
Andre Ethier 2 0.8 -1.2
Michael Bourn 4.4 0.3 -4.1
Yadier Molina 2.3 2.9 0.6
Ben Zobrist 3.8 5.5 1.7
Hanley Ramirez 4.2 3.4 -0.8
Hunter Pence 2.7 4.5 1.8
Denard Span 2.6 4 1.4
Dustin Pedroia 3.2 4.1 0.9
Brett Gardner 6.1 3.5 -2.6
Even when we throw out the guys who are out of baseball later, these players averaged a decline of 1.92 WAR, which is enormous considering they averaged 3.8 WAR to start with. They lost half their value. The intra-individual analysis, which is not affected for the changing population, basically shows an extra 35% decline. The population level analysis, that Craig is showing here, is severely underestimating the decline of 26-29 year olds to 30-33 by dropping the bad players from the 26-29 when they get to 30-33.
… but JD isn’t currently a FA because teams don’t want to pay him “at 30”. They don’t want to pay him at 35+… and he is demanding a contract that covers his age 36 and 37 seasons… the most likely seasons to start his decline.
Teams are starting to play hardball with those later years, they simply do not want to pay for them and they finally seem to understand that they might not actually have to. Cano got a late-career contract, was that the last one? Maybe Ellsbury before him? Has anyone else been paid for their age 36+ season in a 4+ year deal?
Hosmer just seems to be pricing himself out of a contract by asking for too much money, I don’t think anyone is actually afraid of his age.
I was thinking the same thing, but the article addressed that point. The key is the total money spent vs the total value earned. If Martinez is going to be worth 10 WAR for the rest of his career, and a WAR is worth $9M on the FA market, he should be paid about $90M. Whether a team pays that as 4/$22.5 or 6/$15 matters a little from a cash flow and dead money perspective, but the total value is the most important thing.
But we know teams are willing to pay $100M or even $125M (Red Sox), the problem is JD wants way more than that, well over $150M.
The problem with this reasoning is that it ignores the effect of the luxury tax and associated penalties – it’s not the money, it’s the LT and penalties that causes teams to not want to carry dead money in years 6 and 7 when the player is providing 0 WAR.Continuing to pay big money at the back end of the deal gives teams no room to replace their production if the luxury tax limits are in play.
Except his 10 fWAR career value might extend beyond 4 or 6 years, so you wouldn’t necessarily want to pay for the full 10 wins… and the $/win goes up, so his 2021 wins will be probably be worth more.
STEAMER has him at 2.3 fWAR for 2018– if you are figuring 2018 at $9M and add 5% per year and assume -0.5 per year in aging, then:
2018 2.3 fWAR @ $9.0M = $20.7M
2019 1.8 fWAR @ $9.5M = $17.1M
2020 1.3 fWAR @ $9.9M = $12.9M
2021 0.8 fWAR @ $10.4M = $8.3M
2022 0.3 fWAR @ $10.9M = $3.3M
TTL 6.4 fWAR @ $62.3M
So, for the rest of his career, this would project him at $9.7M/win on average.
–
Use the depth charts as a base for 2018 [2.7 fWAR] and I get 8.7 fWAR over six years for $84M [same $9.7M/win average].
Use a 5-4-3 projection for 2018 and the same otherwise, I get 10.1 fWAR over six years for $99M [$9.8M/win average]
Use the Fans projection of 3.7 fWAR for 2018 and I get 14.7 fWAR over 8 years, $146M total [$9.9M/win average]
I picked the numbers out of the air. The actual numbers don’t matter, it’s the concept.
7 years takes him to age 36. If he follows Jose Bautista age curve he is great through 34 and has a couple of down years at the end. With 70% of players making below market salaries teams can afford that. I remember people predicting David Ortiz was done at 34. He had some of his best years in his late 30’s
Citing one caught-PED user as a counter example isn’t great evidence.
I agree with most of your post except specifically “age 36 and 37 seasons… the most likely seasons to start his decline”. This implies “post peak” which the data show is a decreasing number (post steroids era?) and most likely in the early thirties.
According to the last graph, 30 year old players are not suddenly bad at age 30, but their decline starts at age 31 season before entering their least valuable year of their careers at 34, then its basically over. For everyday players.
That bump in 1990 is driven by Schmidt, Brett, Downing and Fisk, among a few others HOFs/Very good players (hello, Dwight Evans). All we have now is…Cano. Miguel I guess as well, if he steps it up. I guess the older players now don’t have to play for a contract, so they don’t work on it. I remember Schmidt back int he day working on 1 or 2 year contracts max into his mid-late 30’s.
There’s also Winfield and Yount. Mike Schmidt actually retired midway in the 1989 season.
THANK YOU
The article is fine, but the available evidence as far as contract offers to players like JD Martinez and Hosmer is already consistent with this article. A more accurate title might be “Teams Already Know Players Don’t Become Terrible at 30” – that’s why, e.g., the Red Sox have reportedly offered $125MM/5yrs to 30 y/o Martinez (https://www.mlbtraderumors.com/2018/01/red-sox-offer-to-j-d-martinez-reportedly-in-range-of-100mm.html).
The Red Sox offer is fully consistent with your prior article on JD’s expected contract given his age and hitting prowess, where you concluded “If you want to take the contract down to six years, then the number comes back at $150 million.” And, as you mentioned in that article, you only looked at hitting, and ignored fielding and baserunning value, which makes JD look “better” as far as an expected contract value.
That’s a “market value” deal for Martinez compared to close comps like Justin Upton (also a 30 y/o RHH, corner OF who has put up ~14 WAR over the last 4 years – projected values are similar), who basically signed a 5yr/$106MM deal this offseason.
Exactly. Came here to say the same thing.
Any analysis on Martinez’s expected contract value that leaves out fielding and baserunning is pretty poor.
https://www.fangraphs.com/blogs/j-d-martinezs-potential-for-falling-off-a-cliff/
Infact, the author projects Martinez for 3.8 WAR. Steamer and Depth Chart projects range from 2.3 to 2.7.
J.D. Martinez’s Contract Estimate — 7 yr / $159.8 M
Year Age WAR $/WAR Est. Contract
2018 30 3.8 $9.0 M $34.2 M
2019 31 3.3 $9.5 M $31.2 M
2020 32 2.8 $9.9 M $27.8 M
2021 33 2.3 $10.4 M $24.0 M
2022 34 1.8 $10.9 M $19.7 M
2023 35 1.3 $10.9 M $14.2 M
2024 36 0.8 $10.9 M $8.8 M
Totals 16.1 $159.8 M
Assumptions
Value: $9M/WAR with 5.0% inflation (for first 5 years)
Aging Curve: +0.25 WAR/yr (18-27), 0 WAR/yr (28-30),-0.5 WAR/yr (31-37),-0.75 WAR/yr (> 37)
The authors’ projections for Martinez ignore more accurate methods, and instead project zero decline from age 29 to age 30- based on a sample of 10 players (including caught PED users Jason Giambi and David Ortiz) . The article admits that certain players in the sample lost entire seasons, or were quickly out of baseball- but their unplayable decline removes their poor performance from the sample.
Boras himself would struggle to present more misleading assessments to a serious audience. IMO there is too much counterfactual methodology in the Travis/Craig series on how teams need to spend more on FAs.
Good piece. I’d be interested in seeing a breakout of players say, 33 and older, on multiyear contracts. Part of the problem with the idea that total value of the contract is spread over the duration, rather than AAV and production being linear, is that it becomes harder to remind yourself of that as the player ages. And, if the contract is very expensive, harder to sit the underperforming older player. Pujols had -1.7 bWAR last year, with over 600 PAs, and a contract with 4 more years on it. Maybe he can recover to the 1.4BWAR player he was in 2016. But that’s just 6 WAR over the next 2500 PA’s plus a roster spot–if he can manage that.
Craig, you wrote:
“It’s generally accepted that those “out” years are going to be mostly dead weight.”
Is it possible that teams want to get away from this idea, and limit the money and years they pay to players they have to release at the end of their contracts?
Without the Luxury Tax, it makes total sense to pay a higher AAV for a shorter length contract. The Tax complicates things, making it better to reduce the AAV. But maybe there is some movement towards the higher AAV with the benefit of limiting dead money at the ends of contracts.
Smart teams will need to balance, treading a line to stay under or close to the Tax threshold, while also trying to remain flexible by limiting dead money for their aging free agents.
On average there should be no difference between spreading it out and going short-term. The lack of older dead contracts offsets the higher dollars on the new contracts. But if you want to max out and go for a championship while staying under the tax threshold, then it makes sense to do lower AAV deals, which will come back to bite later on when maybe your contention window will have closed.
Big, long contracts may have been miscalculations about the player or about salary trends, or they may have been a byproduct of very competitive bidding or strategic behavior. This winter lacks the competitive balance among teams that would promote very competitive bidding league-wide, and perhaps this lack of competitive balance often existed previously but was not as big a factor to some front offices who may have wanted to maximize wins or sell more tickets. Also, it is quite possible that GMs are a lot less secure in their jobs now, and making a bad big signing is a good way to get fired, especially if the value proposition is based on less “concrete” factors like blocking a rival (who doesn’t do this in fantasy?) or the possibly debunked “showing players and fans you are serious about winning” (a la Jason Werth).
A unique talent can swing a race, put a team over the top, or even deter competitors from trying. Several teams may have lost out on a World Series title by not claiming/ blocking Verlander on waivers. That is a gigantic loss. Stars can have a disproportionate effect, and “overbidding” is sometimes not only correct, but the only reasonable action.
It is hard to see many of the current free agents as big stars. There is also little room for strategic effects when there are only two contested division races, and thus fewer motivated bidders for any one player – so few that teams are smartly waiting to see if other moves will leave them as the only serious bidder for a specific player. Perhaps teams on the fence are also waiting to see if a team ahead of them will make a huge splash and ruin the benefit of their own potential moves.
There is one exception – JD Martinez is an exceptional hitter, and one of the two competitive divisions is in the AL where his value is maximized. I expect JD will end up with a huge payday, and possibly even get it from the Yankees to block the Red Sox if the Diamondbacks back off. Saving luxury tax money is great, but it is not worth a division title.
In short, what is new is a general lack of very competitive bidding. This could be because teams are strategically waiting this year to see how other moves affect their status, or in hopes of being left the only serious bidder for a player they want. This winter, the “first-mover advantage” was exercised through trades and reliever signings, leaving most other free agents pieces on the board for the end game.
Given that the Yankees seem well stacked in the corner outfield spots (Judge, Stanton, Gardner), I am not sure how JD Martinez would fit on the roster. I suppose he could be the DH, but that’s where Sanchez is going to be when he is not catching, and when he is catching the DH would normally be the odd man out in the corner outfield. I suppose they could give up on Bird and try to convert someone to first base, but it seems like a lot of money to pay out on an experiment, especially when Bird is actually good when he’s healthy.
The amount that JD’s demands will have to drop in order to get the Yankees in on him is unfathomably large.
I agree with Jon. There are plenty of Yankees-specific roster issues, including that they already have 5 OF’s under contract for both 2018 and 2019 if they exercise Gardner’s 2019 option at an incremental $10.5 million over the buyout amount.
The broader point, though, is that I think a of people ignore the dual advantages for the Yankees and Dodgers of dropping under the luxury tax for one year. Part of it is re-setting the penalty rates: though even Jeff Passan, for example, understated even that benefit in his column last week by looking at only hypothetical 2018 luxury tax for the Dodgers and ignoring that the 2019 rate is also lower.
The other part, though, is that any big-time free agent signing is going to be a multi-year contract, and these teams almost certainly expect to be over the tax line in 2018 and subsequent seasons. So these incremental commitments not only imply higher luxury tax rates in 2018 and 2019, but they also imply probably being over the luxury tax by greater amounts in 2019 and subsequent seasons if these teams also add big dollar commitments by signing players from the post-2018 free agent class. The differences get even bigger if those incremental dollars make the difference for these teams blowing through the surtax levels at $20 million and then $40 million over the tax line.
Side discussion; who else is actually starting to get worried about the continued lack of activity on this year’s free agent market? It’s January 22nd and STILL nothing. Over/Under 60% of current free agent’s anxieties are starting to rise?
Nearly all the major unsigned FA will know that they have rejected deals, already, that overpay them. They will know that their agent Boras is doing a PR campaign, etc. It takes two parties to make a deal, often three given agent conflict-of-interests, it is rare that one party is in the dark feeling anxious or surprised if a deal has not happened.
It’s just a PR campaign aimed at baseball fans to suggest otherwise: ‘oh no, poor pablo alvarez’ and ‘think of David Freese!’
Worried? Not at all. But it’s more like disbelief that every day I check the news and it’s just crickets.
this is great, thanks! have you done any purely longitudinal analysis? i’m assuming this is all cross-sectional, which is still informative, though we obviously would have some selection bias inherent in the results. e.g., only the top guys come up under age-22 and presumably only the most talented guys can stick around into their late 30s.
oops, sadtrombone already said essentially the same thing.
“If a team doesn’t want to add, say, a seventh year at $25 million to Eric Hosmer’s offer, it isn’t because they believe he isn’t going to be worth $25 million in that seventh year.”
This is not the complete answer i believe. Yes, a lot people don’t understand that a team buys X years of a player for Y dollars spread in W yearly payments. But there is still something missing from the valuation models for contracts we employ.
Example: A player is going to be worth 100M in 5 years, and 110 in 6 years. The way we do things, by adding the WAR price, inflation, yearly WAR drop of production.. gives the answer that both contracts are “fair”/”neutral” between team and player.
But that’s not really the case, 100M/5y is most of the times better for the team than 110M/6y because it has the added advantage of preserving opportunities. The team could in year 6 offer 10M/1y for the player or not. We don’t account in dollars for the cost of opportunity in each contract. Discussing if a player should receive that 7th year for 25M of dollars is a way to talk about something missing.
max 30 you must play in the minor leagues
Nice article. I’d be interested to see playing time percentages by service time over the last 5 – 10 years. I am wondering if more teams are opting for substantially cheaper, though fewer expected, WAR at a given position. So if a team has the choice of either (a) paying $12mm for an anticipated 1.5 WAR 3B (or $8mm per WAR), or (b) paying $545k for an anticipated .5 WAR 3B from within its system (or $1mm per WAR), are more teams opting for the latter. Also raises interesting question of how teams are balancing maximizing expected WAR output per position (or per plate appearance) versus expected WAR output per dollar spent.
Much of the decline from aging manifests in not playing, rather than playing poorly.
Age related decline can manifest in a few ways (1) injuries limiting playing time; (2) decline in ability to point of being unplayable; and (3) decline in performance. The analysis above only focuses on the third type of decline- lower WAR/600.
Older player brackets tend to put up decent rate stats if they play- because there is a selection bias where only the high performing older players continue to be chosen (unless a Pujols-contract situation or another superstar legacy situation exists). Players who quickly decline in their 30s often stop playing- this is particularly true for players who were never established as above-average regulars.
The big tell in the data above is that the 30-33 group gets half the playing time of the ‘prime’ group 26-30. That’s a huge amount of decline, and does tend to show that many players get terrible at 30- and stop playing. Martinez and Hosmer are examples of guys who will probably avoid getting terrible at 30, because they are established as above average regulars and can decline by 35% or so and still be roughly average. But paying 22m+ for roughly average players is not very efficient, so clubs are rightly avoiding 6+ year deals for guys who will be 30+ and approaching roughly average 2-3 years into the deal.
Seems like an answer looking for a question, and as mentioned in comments above, major selection bias. Plus there is opportunity cost and sunk costs involved – older players are likely on guaranteed contracts and can’t be optioned.
Also, people who think 30 is a magic number are missing the point. It’s not a cutoff line, more like the mid-point of a range, around 29-31 (give or take) where certain skills tend to decline sharply, and that there are certain player profiles that we believe are particularly vulnerable to this decline, which the article doesn’t address at all.
WAR not a good measure since older players get moved to less valuable positions on the defensive spectrum. 1B and DH for example getbdocked 1-1.5 WAR while a SS gets almost 1 WAR added for fielding their position poorly. So if your looking at offense its better to look at the offensive stats. The days where a Jeter can remain at SS until he is 40 over.
The few players at age 33 in 2017 are telling (3.5% of total PA)
I’d like to see a summary of players performance in their last year over time to see if players careers are ending earlier despite better performance in their last year
It occurs to me that if you drop the talent pool by forcing older expensive players out then younger cheaper players can be ready earlier and play better.
I’d kind of like to see some of those charts extended back to the 60’s. A lot of great players who played into their 40’s in the 60’s to early 80’s
One of the author’s first articles ‘How Much are Last Year’s FA Worth Now,’ was very on point.
https://www.fangraphs.com/blogs/how-much-are-last-years-free-agents-worth-now/
“No team can afford to skip free agency entirely, but those players are the most expensive, have the greatest risk and produce the least value compared to their cost. The winner’s curse is real.”
Using some examples from his article- David Price fell off at 32, pitching 74 IP and racking up 1.5 WAR- his lowest total WAR in the past 8 seasons. The data above wouldn’t show decline in WAR/600, but there is a huge falloff in WAR/$M because the real falloff is playing time.
Price is a great example of the winner’s curse: he posted 4.0+ and 200IPs+ for 7 straight years, and even performed at a 4.5 WAR/200IP rate last season. Even so, Price’s 30m/y contract is hugely underwater- and he is now only projected to have 2-3 above average seasons left on his 5 year deal.
The obvious problem with veterans is that they become increasingly likely to suffer injury or wear/tear problems after as they approach and pass 30- and unlike young players, veterans coming back from injury don’t tend to get stronger during the time off, but instead an end-of-season injury for a 32 year old means attrition both from the injury and from the year of aging until he’s next seen.
eh, one injury doesn’t kill Price’s value. He actually had surplus value in year 1 of the deal, and needed to miss 2/3 of year 2 to fall under value.
Yes, but that’s like being up 10 on the Patriots after 3 quarters- you have to do better than that, because you’re expecting to lose that much in the 4th.
Disagree.
One injury in the early portion of a 6+ year deal for a 30+ year old generally does destroy the surplus value, as it did for Price.
Having surplus value in year 1 is not rare: Price did well (4.4 WAR) and posted about $8.6m in surplus value year 1 [$9m/WAR, $31 salary], indicating in part what a great player he was at signing and throughout his career.
Then he had a moderately serious injury: an ongoing problem with elbow inflammation.
He then posted “only” 1.5 WAR in year 2, while making $30m, puts him around roughly -$17.5m for year 2 [taking cumulative contract surplus value to -$8.9m]. He was quality but only pitched 70+ IP.
If a 30+ player on a big money long contract isn’t well into surplus value by the end of year 2, the contract is very likely underwater.
Price now projects for 3.3 WAR next season, which is worth roughly the $30m he will make. The problem is that he is projected to decline 0.5 WAR for each season following: ’19 (2.8), ’20 (2.3), ’21 (1.8) and ’22 (1.3). Price projects to be a roughly league average starter in his age 36 season (2021), and still useful in 2022 at age 37. Even so, his projection for the rest of the contract is underwater
Already -$8.9m, adjusted by-
2018: 3.3 WAR($33.0m value) $31m salary, -6.9m
2019: 2.8 WAR($28.5m value) $31m salary, -9.4m
2020: 2.3 WAR($24.0m value) $31m salary, -13.9m
2021: 1.8 WAR($19.5m value) $31m salary, -25.4m
2022: 1.3 WAR($15.0m value) $31m salary, -41.4m
I should have, but didn’t, include 5% inflation for the value of WAR, but the difference would be moderate- but for that and perhaps other reasons, I’ll estimate a -$35m value remaining on Price’s contract.
And this is for a great, great ace SP, who blended both consistency and top tier performance, with very few red flags when he hit FA. He didn’t become a terrible player at 30, but he did become a roughly terrible asset: few players on a team’s 40 are worth -$35m.
He could rebound like Greinke and rebuild his surplus value, but the vast majority of stars with this trajectory, while dealing with injury, go the wrong way.
The issue isn’t just the injury that took out part of his year 2, it’s the likelihood also that it takes time for him to get up to his best 33 year old self in year 3. Young players return from injury generally worse- from the injury, but with some countervailing dynamic of having perhaps grown more into their bodies, or tackled a steep learning curve on this or that. Vets return from injury with a double loss- more battle scars on the body, and -crucially- vital time having gone buy while recovering and then getting back up to speed.
Could there be some selection bias here? I’m thinking that teams could require that players in their 30s be more “proven.” In other words, younger but ineffective players would still be given a chance to play, based on their potential. But once a player is older, his playing time would be based on his actual production, since he’s presumably already reached his potential.
This is of course mathematically vague, but it seems that therefore, the aggregate value (i.e. WAR) of younger players is lower than it could be while the value of older players is more “accurate,” indicating that there is a bigger drop-off than it appears.
How about survivor bias?
The analysis in this article includes all active players, but neglects to include the players who fell off a cliff before they got old.
See: Fielder, Prince.
The trend I want to see go is comparing player A to player B to try and predict how they will age. Certain players may have similar builds or skillsets, but that is only part of the equation. You have to factor in genetics, adjustments, and just plain luck/circumstance. I remember last off-season one author predicted that Jose Bautista would age gracefully, and we see how that went. IMO, player ageing is a complete crapshoot, and using a one-size-fits-all Win Curve Model is a poor way to predict something that is frankly unpredictable.
Why do people who object to empiricism generally sign up at fangraphs and 538?
All of the variables you mention are best assessed via systemic, comprehensive, weighted and contextualized observation- i.e. empiricism. Aging isn’t at all a crapshoot or unpredictable, highly predictive models (adj r sq > .6) for age related decline can be built for nearly all large samples of players.
I question the basic method here. You need to track individuals as they age, not simply age groups over time. The latter is prone to selection bias.
Other quibbles:
— I assume it’s just a typo in the second chart, changing the second age grouping to “26-30” instead of “26-29”?
— Why these particular age groupings? For instance, I find age 25 getting more PAs than age 29, thus more deserving to be grouped with 26-28.
— Exactly what is meant by calling 26-29 the “traditional prime”?
I know the question is to the author, but I believe 26-27 had long been the peak year for an average player except that the past 8-10 years has seen the peak of the performance/age curve shift to 24.
The data in this article, plus some data that was omitted, show why older players, even good FAs around 30, are a worse deal than younger players.
A 23 year-old who produces 1.5 WAR on a $550,000 one-year contract is much more valuable than a 30 year-old producing 3.0 WAR on an expensive, multi-year contract. The younger player will theoretically improve, while still being affordable, while the 30 year-old will theoretically decline while being an expensive fixed cost. Further, if the younger player declines, he can be cut with little cost. It’s a much better investment in the future. The value is even less favorable for the older player if the performance is similar.
Of course, this theory will likely not apply to teams trying to win this year, if the additional 1.5 in projected WAR makes a difference in this year’s playoff position and competition. But most teams have adopted a strategy of trying to maximize their windows, instead of trying to take long shots each year, to the disappointment of Scott Boras.
It seems there was more of a “learning curve” for many players entering the league in years past. That’s disappeared to some extent as evidenced by the Gary Sanchez’s, Clay Bellinger’s and Rhys Hoskins’ types who enter the majors and surpass anything they showed in the minors. Is it a case that rookies are more prepared, or is it more related to the style of play now, coupled with the decrease in more seasoned players who in past years would be more likely to exploit their weaknesses early on? I’m not sure I believe that myself, but I do believe something has changed that makes it “easier” for rookies to enter the league with a bang, and that’s a good thing for MLB teams looking for cheap and impactful talent. I suspect there’s a connection.
Regarding this statement discussing playing time: “What’s interesting here is that players who have just left their prime (age 30 through age 33) haven’t necessarily followed the same pattern as that slightly older population.” Perhaps this age group doesn’t have as much of a drop in playing time as some may think could be because at the time, nobody knows for sure if the player is really declining, or perhaps decreased production could be due to an injury or a long slump, and the team figures that based on past performance, they will progress back to the mean of their careers. It might take a couple of years for the team to figure out that the other potential factors are not the cause of the decreased performance, and then the pronounced drop in playing time begins after that.