GeneralUncategorized

The Case For The Franken-player

Fantasy championships are often won on the margins. In a competitive league, getting every last at-bat and maximizing every position on your roster can be the difference between finishing in the money or missing out. In my dynasty league, the perennial winner is able to maintain his crown with a dominant pitching staff and a deep bench full of good-to-acceptable hitters that he mixes and matches with. One way to get a little extra out of your hitting categories is through playing splits, just like a major-league manager does.

For the purpose of this exercise, I will focus on batters who have significant positive splits against right-handed pitchers. I am not looking at positive splits against left-handed pitchers due to the relative infrequent number of starts they make. It would be difficult to justify a roster spot to a player who you might only use once or twice a week, no matter how good they are. I compiled three-year statistics for all left- and right-handed hitters to attempt to smooth out random variance.

For right-handed hitters, there was really no one that stood out as over-performing against right-handed pitchers. Actually, Jonathan Schoop and Mark Canha did, but their sample size is so small I have no faith it indicates anything except statistical noise. It wasn’t really surprising, as opposite-side batters are assumed to have the edge, but I had thought that at least a handful of righties would show a split advantage.

Among left-handed hitters, there were some very pronounced split advantages. To my surprise, the poster child for a pronounced split, Curtis Granderson did not appear. Neither did Mitch Moreland, the future fantasy part-timer who became the impetus for this article after I acquired him last week. Allow me to paraphrase something my econometrics professor said, that has always stuck with me: Have some sort of an idea of what to expect. If the results are completely unexpected or counterintuitive, consider the possibility that you have made a mistake and not in fact made a groundbreaking discovery.

To my relief, when I took a closer look at Granderson’s splits over the past three years, I found that while 2015 showed the extreme split that you’d expect with Granderson, the previous two seasons were simply so poor that he was completely impotent against all pitchers. Similarly, while Moreland also had an extreme split last season, it was really his first successful season in the majors.

Here’s your basic strategy. Once you’ve identified a player that has dominated righties but is brutal against lefties, you start him only against righties and never ever play him against lefties. Easy enough, right? Here’s how valuable it can be:

Consider the table below, which compares a Mitch Moreland/Desmond Jennings hybrid compared to an elite hitter such as Jose Bautista.

 

Player AB R HR RBI SB AVG OPS
Mitch Moreland 420 50 23 60 2 .290 .880
Desmond Jennings 200 30 5 15 8 .250 .700
Mitch Jennings 620 80 28 75 10 .280 .835
Jose Bautista 500 85 32 91 5 .258 .870

The Moreland/Jennings hybrid is better than Jose Bautista is projected to be this year, a player I calculate him to be worth 6.7 standard deviations above replacement (without positional adjustment). The combined player was worth 6.8 standard deviations above replacement. Admittedly, the 120 plate appearance difference is meaningful but I believe you get the point. I’ll leave it to you to guess which player would be harder to acquire. For what it’s worth, I traded Dylan Bundy to a very astute owner for the privilege of owning Moreland.

Now let me introduce the concept of a “super platoon.” Consider the hybrid of Matt Adams and Adam Lind playing only against right handers. If you can simply get 300 at-bats out of each one you could have the following player. Ideally, you could probably mix and match to get around 900 at bats out of the two, filling the roles of 1.5 positions on your roster. The bottom row refers to standard deviations above the mean for a typical owned player in my 7×7 dynasty league. My league uses OPS and offensive strikeouts in addition to the traditional five hitting categories.

Player AB R HR RBI SB AVG OPS
Adam Lind 300 47 13 52 1 .313 .912
Matt Adams 300 37 12 44 1 .298 .829
Adam Adams 600 84 25 96 2 .308 .870
Standard Deviations 1.5 1.35 2.22 -.7 3.0 2.24

This player would be the eighth-best overall fantasy player in between Mookie Betts and Andrew McCutchen. The closest category by category match to this player is Prince Fielder, if he had better ratios. Is this realistic? I hope so, as I intend to utilize splits in my own fantasy strategy. At the very least the two players used would be much cheaper than the one-player alternative.

If you’re content with taking last year’s splits or an average of some number of years then your task is done. However, students of baseball statistics know that one of the overriding forces in the field is regression to the mean. It simply takes a lot of at bats to establish a baseline. According to research done by Tom Tango in “The Book,” it takes approximately 1000 at-bats against lefties to establish a baseline for left handed hitters with confidence. For the sake of research, I regressed all left handed hitters in my sample against 1000 at-bats of league average split performance against left handers to obtain the following data.

Player AVG OPS
3-yr avg/regressed 3-yr avg/regressed
Adam Lind .313/.299 .912/.871
Matt Adams .298/.286 .829/.802
Gerardo Parra .291/.286 .778/.761
Andre Ethier .286/.279 .828/.806
Denard Span .311/.303 .800/.785
Pedro Alvarez .245/.241 .807/.787
Ender Inciarte .309/.301 .768/.752
Carl Crawford .294/.292 .760/.764
Scooter Gennett .307/.291 .796/.759
Lucas Duda .250/.249 .863/.848

If you’re interested in the math, look below, otherwise feel free to skip the next paragraph.

  1. Determine the “observed split difference”
  2. Regress it against 1000 at bats of league average lefty vs. lefty – this is the new platoon split
  3. Take the projected stat value and add the vs. RHP portion of the split
  4. The result is the minimum split value that you would expect

An example using Adam Lind’s batting average follows:

Assumptions: average league split for LH vs. RHP: .025 (25 batting points)

  1. Avg vs RHP – Avg vs. LHP = .313 – .193 = .120 (120 batting average points)
  2. .120(1-1000/(1000+249))+.025(1000/(1000+249)) = .044 (44 batting average points)
  3. .291 = (249/1411)(x-.044) + (1162/1411)(x)

.291 = (.176)(x-.044) + (.824)(x)

.291 = .176x – .008 +.824x

  1. .299 = x

Your first impression, as I admit mine was, might be to write this concept off and assume that the players that you have identified are good bets to maintain their current splits. While this might be the case – and I hope it is with Moreland – I have added a third column for career splits. (3 year average/regressed/career splits). A few hitters are too young to have had enough at bats to balance out a three year sample, a few others are shells of their former selves which inflates the career rate, a few hint at decline, and a few show consistency.

Player AVG OPS
3-yr avg/regressed/career 3-yr avg/regressed/career
Adam Lind .313/.299/.293 .912/.871/.863
Matt Adams .298/.286/.296 .829/.802/.822
Gerardo Parra .291/.286/.296 .778/.761/.767
Andre Ethier .286/.279/.304 .828/.806/.889
Denard Span .311/.303/.293 .800/.785/.760
Pedro Alvarez .245/.241/.246 .807/.787/.794
Ender Inciarte .309/.301/.309 .768/.752/.768
Carl Crawford .294/.292/.303 .760/.764/.802
Scooter Gennett .307/.291/.307 .796/.759/.796
Lucas Duda .250/.249/.254 .863/.848/.842

In my opinion, Lind, Adams, Parra, Ethier, Span, and Duda appear to have established and maintained a significant split advantage and would be worth platooning when they are starting against a right handed pitcher.

The other hitters have either not established a long enough track record or accrued a large part of their split during earlier parts of their career and no longer offer the same upside.

The strategy outlined above is a good way to get a few extra rotisserie points out of your fantasy lineup. If you are mathematically inclined, it may be worth your effort to regress towards the league average. Either way, a number of the players listed above would make excellent part time fantasy plays.

See (here) http://www.insidethebook.com/ee/index.php/site/comments/fangraphs_now_has_some_splits_data/#comments  and (here) http://www.fangraphs.com/blogs/estimating-hitter-platoon-skill/ for articles that inspired the mathematics behind this post, as well as (here) http://www.baseballprospectus.com/article.php?articleid=19675 for a related article by Baseball Prospectus’ Bret Sayre.

The Author

Jesse MacPherson

Jesse MacPherson

4 Comments

  1. Alex
    March 23, 2016 at 12:08 pm — Reply

    The splits you’re using are different than the situation you want to use the players in, right? i.e. you’re using their splits for when they face RHP, not for games where the starter is a RHP. You’re not going to be able to avoid him facing LHP relievers, so the stats you end up with will be much worse than the ones you’re calculating here.

    • March 23, 2016 at 1:22 pm — Reply

      Yes, I used stats against right handed pitchers not stats in starts against right handed pitchers. There isn’t usually a ton of difference though. The guy I based this article on:

      Mitch Moreland .294/.290 and .876/.877 was basically the same.

      Adam Lind had a more pronounced difference, but he played nearly every day. This meant the brewers thought he was their best option regardless of pitching hand he was going against. Someone like Moreland is fortuitously subbed out for you when a lefty comes in the game.

      Excellent point, however.

  2. CTBB
    March 25, 2016 at 6:10 pm — Reply

    Got any good recommendations for 3B eligible players? After the elite guys there aren’t many I like and I was thinking about waiting and doing something like this

  3. July 2, 2016 at 2:02 pm — Reply

    […] The Case for the Frankenplayer: I believe that the most important finding – by Tom Tango, not me – is that regression to the mean is a powerful phenomena that should be dismissed at your own risk. The three players that I selected as the best fits for this system – using the strong side of a platoon split to make a single player – were Mitch Moreland, Matt Adams, and Adam Lind. I actually implemented this strategy for a period of time, platooning Moreland and Adams. Eventually, Mitch Moreland made this impossible and I dropped him after not one owner would give me a single FAAB dollar for him. Moreland is a warning flag to anyone who chooses to trust bias resulting from personal interest over advanced statistics. The fact that Moreland never really hit against either armed pitchers prior to 2015 should have been a red flag that 2015 might be an aberration. And it was. Moreland is 30 years old and he’s repeatedly shown that he’s a below average major league hitter. Lind’s extreme regression was a bit more surprising as he had a larger sample size to look back upon. But alas, he’s also struggling to hit against righties much less lefties. There are other underlying factors that could be confusing the matter such as changing stadiums and leagues, but my second target was also a failure. The ray of light was Matt Adams. Unfortunately, although not for his owners, his success has not validated this frankenplayer approach. He’s simply been so good that despite acquiring him to platoon, I’ve used him nearly every day. I do believe this process has value, but I am discouraged at how unpredictable splits seem to be. Verdict: Productive Ground Out. […]

Leave a Reply

Previous post

Travis Shaw is on the Rise

Next post

TDGX Draft Recap - Round Two