Historical WARP and OFP

Now that the trade deadline is heating up, baseball’s best fan past time is pricing out trades and dreaming up returns for their favorite clubs. Analysts and writers have a tougher line to follow. First and foremost, not only do clubs hide their proprietary player evaluation and analytics systems, they also hide their risk assessment and pricing strategies. Given that discussing the trade deadline requires discussing what a player might do in the future, and what a team might have to surrender to acquire that player’s services, trade season is essentially one gigantic opportunity to try to determine strategies for pricing risk and therefore making transaction. Insofar as baseball teams operate as businesses, even throughout the player development side of things, they are determining their aversion to the risk associated with each particular upside (or lack thereof), and then finding a suitable partner to meet that upside.

Grading Trades: Surplus
Translating OFP
Cashing Out OFP
Organizational Logic and Playoff Trades

Over the course of the past year, I have used a WARP-based system to assess transactions. I use a harsh depreciation system to demonstrate the assumption that a player’s value will only decrease in time, which basically attempts to price trades closer to their worst-case scenario rather than their best-case scenario. I have priced prospects by assessing the distribution of 18,848 careers and using that distribution to approximate prospect value. This model employs Baseball Reference Play Index WAR data.

Assessing the careers across the history of baseball produces a clear distribution of talent, and also helps to clarify what a player’s ceiling looks like on the field. For example, by the time a batter reaches 1.1 WAR, they are within the top third of all batters in the game. This is helpful to temper expectations of how prospects should produce, and also to understand whether an MLB player is truly elite. Using this career wide scale to assess transactions ensures that analysts can quickly translate the distribution of talent to assess the likelihood of future player production (and therefore the risk of acquiring a player or prospect).

This scheme works across individual seasons, as well, which can be drawn from Baseball Prospectus CSV functions (for example, I found approximately 29,428 individual pitching seasons with recorded WARP, and thousands more with unrecorded WARP, and 95,790 individual batting seasons with recorded WARP, which can be assembled according to mean and standard deviation). Once the mean WARP for pitchers and batters is identified, one can easily scale nearly every player in baseball history according to their percentile on a season-by-season basis:

These scales can be used to approximate Overall Future Potential (OFP), as well, as the distribution between prospect classes can be compared to the distribution between historical seasons. For example, according to the 2013 Baseball Prospectus Top 10 organizational lists, those 300 prospects (and approximately 150 “just interesting” guys) are distributed as follows: 6.7 percent 70 OFP, 27.8 percent 60 OFP, 32.7 percent 50 OFP, and 33.3 percent 45-50 OFP (“just interesting”). In this scenario, 60 and 70 OFP prospects neatly align with the 1+ and 2+ standard deviation historical WARP seasons, while the 50 OFP prospects wind down to the mean WARP or fall just below replacement level on a single season basis. This should align with what one would expect a prospect to produce once they reach the MLB (for example, it would not be surprising if Mauricio Dubon was a player that accumulated between 0.0 and 0.7 WARP on a seasonal basis, while Josh Hader produced 4+ WARP at his best; we could certainly draw such estimates from their tools and scouting profiles).

By identifying mean and standard deviation for individual WARP seasons, one can assess player value in monetary terms based on the progression of each standard deviation:

I believe this is a useful, if crude, system because it seeks to provide meaning to a statement such as, “if Lewis Brinson is a star prospect, he will be likely to produce at least 28.0 WARP in his career;” alternately, one could reasonably expect Brinson to have a 4.0-to-6.0 WARP ceiling should he reach his optimal OFP. I depreciate this historical value in order to express the risk of Brinson reaching that level. Obviously, GM David Stearns and President Jon Daniels did not price out Brinson as a $196 million player (using one free market assessment of the value of WARP); however, depreciating Brinson’s ceiling to accommodate the risk that (1, at that time) Brinson failed to reach the majors and (2, perhaps more plausibly) Brinson plays closer to his floor than his ceiling in the MLB gets Brinson close to Jonathan Lucroy’s value. Placing Overall Future Potential (OFP), Wins Above Replacement (WAR or WARP), and contracts ($$$) on the same scale produces a solid at-a-glance pricing system that allows fans, analysts, and writers to quickly consider risk and reward. A similar price emerges if one moves from a historical career evaluation model to a model that assesses players based on their likely ceiling of seasonal WARP.

One benefit of assessing more than 18,000 baseball careers and scaling those seasons to prospect expectations is that the different parts of these systems speak to each other easily and clearly. We can literally test our assumption that the Lucroy trade was in fact a pretty good deal for both sides on the day of the trade. Obviously, post hoc analysis is necessary each and every year following a trade to test those assumptions. As in Benefit-Cost Analysis, it’s not simply enough to drop things the day of the trade, and adding analysis on an annual basis can help to fine tune assumptions about value, as well (or solidify trade deadline trends). In the case of the Rangers trade for Jonathan Lucroy and Jeremy Jeffress, depreciation analysis shows the rapid decline in surplus value that follows poor production:

Using WARP, OFP, and $$$ to assess trades is inherently problematic insofar as it (a) incorporates biases involving the Replacement Player Model, (b) only assesses players according to marginal value, and (c) assumes that player value can be expressed in one particular figure (be it cash, future potential, or current production). Yet, pushing back on (c), I don’t think it’s entirely problematic to say that an analyst can express player value at one point in time while also understanding how that value can change very quickly, on a seasonal basis, or over the course of a career. Jimmy Nelson is a fine example of this type of issue; the Brewers’ righty struggled with command and mechanical adjustments throughout his first couple of seasons, but working through adjustments has helped him produce notably above average runs prevention in 2017. It’s not wrong to assess Nelson in such a manner now (notably better than average), nor was it wrong to previously assess Nelson (struggling rotation depth). The narrative can connect to the statistics, and one can use a transactional model to assess risk and value in order to judge trades and perhaps understand how value is allocated within a given organization; one could even use such a system to analyze how an organization acquires risk (whether they are risk averse, or neutral, or aggressive).

It should also be clear that players can produce well beyond their OFP. Nolan Arenado is an example of such a player, a 50 OFP top prospect in 2013 who nevertheless entered 2017 with 21.9 WARP over four MLB seasons under his belt. But fans and analysts should be wary of the lesson of Arenado; for one Arenado, the 2013 Top 10 organizational prospects included 35 players with 0.0 or lower career WARP within the class of 50 OFP prospects (the same class as Arenado), and this is prior to considering the forty-five 50 OFP prospects from that class that had yet to reach the MLB (like Tyrone Taylor, for example). Arenado is a valuable lesson about how players can exceed their OFP, but one should understand that developing a single Arenado cost 80 players who have yet to reach the majors or are producing replacement level careers. Incidentally, the 2013 Top 10 prospects rated 50 OFP entered the 2017 season with 190.8 WARP over 287 seasons, which corresponds quite well to the mean seasonal 0.46 WARP produced above.

Both WARP and OFP have their respective imperfections as measurement systems, but their benefits also allow them to serve as solid transactional assessment tools despite their shortcomings. In the case of the Brewers, one can literally price out the value of the club’s extra cash, surplus of prospects, and the depreciated (or maximum) surplus of any intended trade target in order to understand whether a trade is worth the risk. Absent databases full of proprietary scouting, mechanical, and health information, this type of at-a-glance measurement system can approximate transaction prices and help one understand whether teams made an advantageous trade, or simply a good baseball deal.