Free agent decisions have been consolidated into a series of algorithms
Coming into the final week of February the baseball world should be slowly emerging from the caves of hibernation hungry, but in good spirits. Already this month there is nearly 40 more minutes of daylight than at the start. The numbers on the “Opening Day Countdown” clock have teasingly dropped into the 30’s. And, after all, there are worse places to be on the planet than Florida and Arizona whilst the Mid-Atlantic absorbs yet another sloppy semi-frozen kiss from winter. Yet this year the onslaught of canned clichés coming out of Spring Training are noticeably muted. Instead there is an undercurrent of unease. There are the occasional bombastic rebukes of the current direction of baseball as business made mostly by established stars. Surely the floating ice floes reflect only a small portion of the bergs out of sight under the surface. Whether the winter’s discontent will be, “…made glorious by spring” is suddenly a very real question.
In almost any sort of upset to the norm there arise phrases or key words that crystallize the situation. These often go on to be rallying cries. One that has emerged this year is “Algorithm.” As often happens sports reflect the happenings in larger society. Algorithms have taken on a nefarious connotation as they are used for all manner of processes out of individual control. Like any other automated process they are the product of the formulator complete with his or her biases. Anyone who has used a decision tree or a decision-making flowchart has used an algorithm. When Brad Brach signed with the Cubs he raised many eyebrows with the following quote:
“We talked to certain teams and they told us that, ‘We have an algorithm and here’s where you fall.’ … It’s just kind of weird that all offers are the same, they come around the same time. Everybody tells you there’s an algorithm.”
It didn’t take long for the “C” word to appear. Collusion among owners to restrict the movement and compensation of players was uncovered during the 1980s. The settlement figure of $280 Million in 1990 does not reflect the staggering cost in real terms. Inflation alone would drive the current value to $580 Million. But, the state of the industry is a key to understanding the real impact. In 1990 MLB annual revenue was approximately $1 Billion. At today’s income of roughly $10 Billion an equivalent settlement would be nearly $3 Billion. If there are teams playing loose with collusion rules they are playing with some serious fire. One team that surely not doing so is the Nats. The Corbin and Dozier deals are testament to that.
There’s a simpler, more likely, and reasonable explanation for the situation Brach experienced. The answer lies in the game’s latest and greatest shiny toy; analytics. Baseball has operated since its inception as a bit of a farmer’s livestock auction. The eyeballs were the predominant, nearly sole source of data. Whether it was Minor League, international, or MLB scouts the visuals fed the decision-making aided by a healthy dose of gut feel. The hits-and-misses by organizations on players were remarkably disparate. Enter the analysts and things changed drastically. Analytics started by finding what is the mathematically most advantageous pitch to throw Player A in a 3-2 count. It doesn’t take a staff of 21 analysts like the Dodgers employ to figure that out. One product they have been able to produce is the noted algorithm for free agents. Actually, it’s a fairly simple process. But, it took a massive investment of time and labor. Start by producing an “Aging Curve” of production versus age and service time. The data to produce an aging curve is available to even casual fans on free databases. It just requires enormous time to do the sorts.
The “Wins above Replacement” figure quantifies something baseball never worried about in the past; Return on Investment measured in $/WAR. Plug in Player X’s previous production, his age, his service time, and then find the spot on the curve when the WAR production will drop below the acceptable ROI figure. Voila! The team knows how much to offer and for how long. There are all manner of variables. Injuries, especially Tommy John surgery are a key consideration. But, there are enough TJ surgeries in the databases now that one can reasonably predict how a dedicated athlete will produce in the future based upon the variable factors such as age when injured, innings since return, and various metrics since return such as Fielding Independent Pitching. It all becomes just another track in the algorithm. The key word is “predict.” Cleveland pitcher Trevor Bauer was ruing the new reality:
“ (Baseball is using) increased reliance on aging curves and projection of future performance. Baseball used to pay for what you’ve done, and baseball is now shifting to projecting that they’re going to do. It’s very similar to how major corporations are being run.”
Well, Skippy, welcome to the new world. At last glance a $300 Million investment qualifies as something a major corporation would do. No publicly-traded organization would invest multiples of $100 Million without ROI projections. It only makes sense that the franchises of a $10 Billion annual revenue corporation would act like it. The days of the eyeball-test livestock auctions are over.
So, why are all of the various teams coming up with similar solutions with their algorithms? It’s because all of the data is universal. With 30 teams of analysts punching in the same data the answers are going to come out fairly close. Each player that has come through the leagues has left a trail of data. When the data is coalesced it gives discrete trends. “A pitcher at age 37 will yield X% of his previous-year’s WAR not to exceed Y% of his age 34 yield times the standard deviation.” That info is there for anyone to grab. If it sounds sterile, it is. Anything based on math tends to be so. Stand by; however, it will get even more antiseptic as time moves on. Technology will add even more hard data to the mix.
One of the “secrets” the Astros used to improve the performance of their pitchers was the “Rapsodo Machine.” As usual baseball is late to the party. The golfing industry developed the “Trackman” nearly a decade ago. No longer reliant on instructor eyeballs alone to determine the quality of ball-striking the Trackman delivered a powerful set of measurements of clubhead path and rotation as well as exhaustive ball flight information for every swing during practice. Current World #1 player Justin Rose and his instructor Sean Foley use the device during every training session. The Rapsodo is now in the hands of nearly every team. The Angels recently boarded the bus. The sole outlier is Baltimore. The initial use will be to improve performance and identify deficiencies. But, like any other source of data it will eventually feed into the performance projection mechanisms. What this will do is tighten the algorithms more into alignment approaching unity.
Where this is headed is towards the more difficult end of the baseball player pipeline, the draft. Rapsodo machines will spread like wildfire through the minor leagues, colleges, and international feeders. The draft boards will become more uniform. It is simply a part of the evolution of the game.
There are several other burrs under the saddles of current players. The “Opener” tactic used by Tampa last year has drawn multiple sets of ire. Buster Posey is venting openly about the rash of teams “Tanking” for a period of years in order to garner top picks in the draft. But, the molasses-like free agency market has drawn the most criticism. The algorithm used by the teams has been dragged out from behind the curtain edge. Far from sinister the sophisticated logic tree represents the product of accelerated investment by the franchises as they increasingly come to grips with the tools available in a new age.
