clock menu more-arrow no yes mobile

Filed under:

A J.T. Realmuto Extension Analysis

MLB: Atlanta Braves at Philadelphia Phillies Bill Streicher-USA TODAY Sports

In my debut for The Good Phight, I will be analyzing the value of a potential contract extension for Phillies catcher J.T. Realmuto. The concept is similar to the one discussed by my colleague schmenkman, who wrote about a Realmuto extension back in March. Check his article out here.

As schmenkman outlined, J.T. Realmuto, by most metrics, is regarded as the best all-around catcher in baseball heading into the 2020 season. It is for this reason that the Phillies would be short-sighted to simply let him hit the open market, especially after giving up so much to acquire him just last February. He has one year of club control remaining before he is eligible for free agency, making this season the final opportunity to reach a long-term agreement.

So, what does a potential J.T. Realmuto extension look like? For reference, here are some of the larger catcher contracts signed in the last decade:

Realmuto is an interesting case because he does not closely align with any of the catchers above in terms of combined age and performance. Buster Posey was the reigning MVP when he set the mark for contract length at 9 years back in 2013. This seems rather unattainable for Realmuto, seeing as he’ll be 30, not 26, when he signs. Mauer’s mark of 8 years is also likely not going to happen, as he was coming off 3 consecutive top 10 MVP finishes (and one win) when he put pen to paper.

However, Realmuto is arguably better than the other four catchers on the list, and that should not change over the duration of this season. It is my estimation that if 30-year-old Brian McCann and 32-year-old Russell Martin can get 5 years, then J.T. Realmuto can get 6. While it is impossible to say for sure what length the two sides will arrive at, I believe 6 years is a pretty decent approximation.

In terms of average annual value, I see Realmuto setting the record. Mauer’s AAV of $23 million per year was set in 2011. Realmuto will be signing his deal almost a decade later, and due to ever-shifting markets, will likely be worth more than $23 million per season by 2021 market standards. Judging by these conditions, a theoretical 6 year/~$150 million contract ($25 million per year) does not sound too far off.

The next question is, how will J.T. Realmuto progress into his thirties? The answer to this one is much less clear. By and large, predicting human performance is a daunting task. Various modeling practices yield promising results, but sometimes it is simply a crapshoot trying to estimate how good a player is going to be across a six-year span. Regardless of inevitable biases and imperfections, here is my attempt to pinpoint exactly how J.T. Realmuto will produce during a potential contract extension period.

The basis of this exercise is to identify similar catchers to J.T. Realmuto from recent history and use their collective performance to model his future production. When I say recent history, I mean catchers who debuted in 1961 (the start of expansion) or later. No disrespect to Yogi Berra, Roy Campanella and other great catchers of previous eras, but the catcher position has evolved considerably over the last half-century. Comparing catchers of today to catchers of 80 years ago is not the most fruitful endeavor.

I will be mapping J.T. Realmuto’s ages 30-35 (expected contract duration) wins above replacement values based off the data from similar catchers in the same age range. Each year’s WAR value is found using a smoothed local regression line (aka loess) which plots expected points using the sample average.

schmenkman started by comparing catchers from their ages 26-28 seasons, which is what I will also be doing. These ages are not arbitrary and are statistically significant according to linear modeling. The data from these years will be used to refine the dataset so that only close approximate statistical neighbors to Realmuto are included. A plot of Realmuto’s ages 26-28 production, compared to similar catchers, looks like this:

For historical context, I included the names of the five catchers whose ages 26-28 WAR values most closely resemble Realmuto’s. Based on this graph, Realmuto’s nearest approximate neighbor is 4x all-star Darrell Porter. For reference, the all-time defensive WAR and offensive WAR leaders for this period are Gary Carter and Mike Piazza, respectively. In order to alleviate the small-sample size bias that would naturally come with only using 5 players, I expanded the data to include all catchers in the following range:

There are pros and cons to choosing such a data window. One downside is that it is right skewed, meaning more than half of the players included are below Realmuto, statistically speaking. You might notice that some of the players left out are geographically close enough to Realmuto that they could theoretically be included. While this is true, introducing a handful of new neighbors, all of whom have lesser ages 26-28 WAR values than Realmuto, would heavily skew the dataset more than it already is. There is no sense in adding more bias. Because of the current skew, I have decided to approach this issue using two different methods.

For now, here is the refined list of comparable catchers:

This list is quite impressive. It contains 5 Hall of Famers (assuming Mauer and Posey get in) and a combined 103 all-star appearances.

With these players in mind, here are the two selected methods of predictive modeling:

Method 1: Take a simple smoothed average of all players. One benefit of this method is it is the purest reflection of the performances of all selected catchers. This means all values in the dataset are weighted equally. The drawback is the final result will favor the players who have lower ages 26-28 WAR values than Realmuto, simply because there are more of them.

The following graph shows a loess curve that follows the smoothed average values for WAR between the ages of 29 and 35. I added age 29 because it provides a glimpse of how Realmuto could perform in his walk-year. This model does not factor in the shortened 2020 season, so it is projecting Realmuto’s performance based on the idea of playing a full campaign. Here is the projected outline of Realmuto’s next 7 seasons (in blue):

From the graph, you can see a pretty noticeable decline from ~3 WAR at age 29 to ~1 WAR by age 35. This type of decline is normal for players in their thirties, especially catchers.

Method 2: Take two smoothed averages, one for players with higher, and one for players with lower ages 26-28 WAR values than Realmuto. This method gives equal weight to the two subsets of players, meaning Realmuto’s projected performance will not automatically favor one side. The downside is essentially the same as the pro; the side with less players is weighted equally, which introduces bias.

The same procedure is performed as in Method 1, except there are two loess curves (for the two groups). Realmuto’s ages 26-28 total WAR of 13.4 is the cutoff line for group separation. The difference between the two is shown here:

The average of these two lines is the final model prediction. Interestingly enough, the “better” players actually fall below the “worse” players on average by age 34.

Regardless of which method you prefer, there are two players who stand out as accurate comparisons to Realmuto’s projected performance. Here is Chris Hoiles’ line (in red) compared to Realmuto’s fitted model line (in blue):

This is almost a dead match. The reason Hoiles’ trail ends at 33 is because that is when his career ended, although that was due to injuries, not poor performance. The player who actually completed his age 35 season with the nearest corresponding line is none other than Darrell Porter, Realmuto’s closest comparison from ages 26-28.

Moving onto the financial portion of the analysis. As schmenkman mentioned in his article, 1 win above replacement is worth approximately $8 million dollars on the open market, according to a Fangraphs study. I took this one step further by calculating this amount for catchers, then fitting a linear model line in order to project these values for the years in which Realmuto would theoretically be under contract. Using the last 20 years as the basis, this is the model line of dollar/WAR values for free agent catchers:

Disclaimer 1: Free agent catchers, not all catchers, are used because they are the only group whose salary is decided by the open market.

Disclaimer 2: I removed 2018 from the dataset because that year’s value of ~$30 million/WAR is a dramatic outlier and would skew the entire linear model.

While the current catcher value sits at around $7.5 million/WAR, it is projected to rise above the $8 million mark during Realmuto’s extension years. The chart below contains Realmuto’s projected WAR outputs for both methods, as well as the estimated surplus values for both methods.

If you are unfamiliar with surplus value, it is the basic measure of a player’s performance, in financial terms, relative to his salary. It essentially tells you if a player is overpaid or underpaid. This is the equation used to calculate it:

Surplus Value = WAR x ($/WAR Amount) - Salary

The two methods actually yield very similar results. A 12.6 WAR isn’t drastically different than a 12 WAR. While both surplus values are in the negative, that does not inherently mean J.T. Realmuto is going to be a negative value player. According to his projected performance and expected salary, Realmuto theoretically classifies as someone who is underperforming his contract, which is common for players on long-term deals in their thirties.

Surplus value is a useful tool, especially for small-market clubs, to be able to find impact players on a tight budget. For a large-market club like the Phillies, the number does not mean as much. In the grand scheme of things, if Realmuto is producing for a World Series contending team, then his surplus value takes a major backseat, especially for a club with a $200 million payroll. Truthfully, most large contracts result in negative surplus values.

Both methods indicate that Realmuto’s production will fall below starter level (2 WAR) by age 33, which follows a normal path for catchers. Catchers, for the most part do not age well, and it would be bold to assume Realmuto is the exception. Obviously, there are a multitude of factors that play into catcher decline. Both Joe Mauer and Johnny Bench, two of the greatest backstops in baseball history, were out of the league by age 35 due to years of wear-and-tear sustained playing the position. Catching at the highest level of the game for over a decade is not easy. Could a move to first base prolong Realmuto’s career? Maybe, but his defensive value behind the plate is a big reason why the club wants to sign him. Signing a catcher to a long-term deal is a quantifiable risk, always has been and always will be.

Like I said before, predicting human performance is extremely difficult. These estimated WAR values are what their name says they are, estimations. From this dataset, Javy Lopez is a prime example of the difficulties of tracking human performance. The dark gray in the plot below is the model variance, meaning it is totally plausible for him to have a WAR anywhere in that range.

Despite what the models say, one cannot assume Realmuto will play like Chris Hoiles or Darrell Porter in the coming years. He could overperform or underperform these expectations every season. That is the risk every ballclub takes when they sign players to long-term deals.

In closing, if the Phillies do extend Realmuto, odds are his performance will likely not match his salary. He would need to continue to execute at a high level though his mid-thirties in order to quantitatively justify his price tag, which is a tall task. On one hand, catchers seldom age well and are usually risky investments. On the other hand, if Realmuto is behind the plate for a World Series champion, then that is what truly matters the most.

Thank you for reading.

All statistics and information were obtained from Sean Lahman’s public research database and Baseball-Reference.com