Fantasy Rounders: Statistical Crossroads

The DFS game can be a shark tank, one that becomes more treacherous in deeper (read: more expensive) waters. There are some experienced gamers out there who have made a killing at Draft Kings and continue to put considerable sums on the line each day of play, and some these “pros” have been known to utilize market dynamics and Wall Street strategies to find success, even in cases where sport-specific knowledge might be lacking. In contrast, I'm an admitted Baseballholic who plays on a near daily basis during the baseball season, but aside from the occasional football Sunday, I don't play the games that are outside of my comfort zone.

There is much to be learned from the professional gamers, and when studying their rosters a few common trends emerge:

*Invest heavily in starting pitchers, willing to spend large amounts on the best arms.

*Players with significant platoon splits are auto-rostered when the pendulum swings in their favor.

*Quality of opponents is paramount

*Recent performance is weighed heavily into roster decisions.

Splits and the influence of starting pitchers are major themes here at Fantasy Rounders. Platoon splits are not part of the salary calculations on Draft Kings, and this leaves them available for exploitation, and many of the regular gamers anchor on the extreme players when the opportunity arises (such as when Buster Posey or Josh Donaldson is facing a southpaw). The last two observations are critical aspects in determining DK prices, and the last subject in particular flies in the face of conventional sabermetric wisdom.

The statistically-savvy are patient, allowing for the accumulation of greater samples before trusting the numbers, and often scoffing at theories such as the “hot hand” due to the influence of random variation. Yet a ballplayer who has been on a great run will not only see his salary cost increase on Draft Kings, but his ownership frequency will often rise despite the cost inflation as some experienced gamers flock to utilize his services. So either the Draft Kings cost-adjusters (as well as its top players) are all being duped by the temptations of small samples or there is actually something to the microscopic view of performance.

To be honest, I've stood on the small-sample side of the fence for quite awhile, as I have watched pitchers who regularly toe the rubber with a fraction of their “true talent” available on a certain day (or in a particular inning), and yet have the skills to dominate if everything jells. Anyone who has worked to master a physical skill can attest to the reality of good days and bad days, and though this trait is naturally magnified in MLB by the quality of opponents for any given ballgame, there are often physical explanations that relate to a player has been particularly on or off his game recently. The inability to see or appreciate these minute details does not preclude their existence, and though the general rules of sabermetrics are a boon to our overall understanding of the game, it's no secret that these aspects break down in a small-sample format such as the playoffs or DFS.

I was faced with just this type of dilemma recently, in a trade proposal for a full-season keeper league. In Episode 16 of the Baseballholics Anonymous podcast (recorded May 7), co-host Sammy Reid offered me Andrew McCutchen straight up for Paul Goldschmidt on the spot (we have both played in this particular NL-only keeper league for 14 years). Sammy knows that I favored Cutch coming into the season, as did he at the time (he chose AC at the top of the draft); since draft day, however, Goldschmidt continued to perform like one of the top hitters in the game while McCutchen battled knee issues and poor performance. It was only about a month of subpar work, and in context I had just said that I believed in Cutch's ability to rebound, but the single-month performance was enough to encourage a knee-jerk decline of the offer. Essentially, I had let one month really sway a decision that impacts the next three years (our cap on keeper contracts in that league).

While not a perfect parallel, I think that this example illustrates the point – McCutchen's early-season slump depressed his salary cost to levels that should have made him an auto-start in DFS based on true talent alone, yet players stayed away as Cutch's performance continually fell short of justifying even the lowered cost. He has since rediscovered his stroke and now I feel confident rostering McCutchen on a daily basis, but if I had ridden my expectations of his full-season performance throughout the month of April then I would have been sorely disappointed in the results. In a full-season league, the manager has the luxury of patience to see if a player can turn things around to approach expectations, but DFS does not allow that luxury, and a common tie between the rosters of top gamers is that they are typically not enticed by ballplayers whose price is surprisingly low due to a recent dip in performance.

Random variation has a tendency to wash out over the course of 162 games, but within the confines of a single contest a player's situation can be greatly magnified. The reason for the streak or slump could amount to a number of factors, from adjustments to approach or mechanics to the relative quality of opponents or complete randomness, but while it is simple for managers in season-long fantasy leagues to sweep these peaks and valleys under the rug, they will separate profit from loss for the regular DFS gamer. Studies of this phenomenon have been lacking, often due to the perceived need to fit constraints that will normalize the sample, but the reality is that a player's performance can swing at any time and may not be perceptible until that swing is far in the rearview mirror.

Ergo, concordantly, vis-a-vis, I have enjoyed more success on DFS when integrating a dose of recency into the analysis, though I typically stick to the extremes. I'm not gonna penalize a guy for a .225/.280/.350 week, but I do avoid those players who are in the midst of 2-for-25 slumps while seeking out those who are knocking the horsehide off the baseball (avoid Bryce Harper at your own peril). In the case of pitchers I like to add observational evidence to seek or avoid players based on whether the numbers are supported by the anecdotes. It steers me into the lake from time to time due to the difficulty in predicting the changing tides of “true talent,” but the additional layer of analytical ammo enhances my ability to put a personal stamp on a DFS roster.

Sometimes it works, sometimes it doesn't, and sometimes it blows up in your face – it's a frustrating answer for folks who are obsessed with finality, but those with a scientific bent will hopefully be similarly captivated by the sheer number of variables that are yet to be understood.