Maximizing Plate Appearances Conclusion
This is the conclusion of Logan’s experiment from the 2015 MLB season pertaining to plate appearances. Still relevant, so check it out!
About a month ago, I challenged all readers to a game of “Would You Rather?” I know it’s been a month, and not all of you are as sharp or as young as I am, so your memory has probably served you poorly and you’ve forgotten. Never fear, take a look back, and refresh yourself on the game that I posed to you – READ ME. Too lazy to read that? The Cliff Notes version is for a month straight I experimented with different lineups and self-imposed constraints to test the importance of maximizing plate appearances in MLB DFS cash games.
Now that you’ve refreshed your memory, let me tell you that I’ve completed the month of experimenting, and it’s time to dive into the data and draw some conclusions. Before I jump into the actual data and results from the experiment, I first wanted to take some time to point out some of the day to day nuances of doing an experiment like this, and how it differed from lineup creation on a normal day.
Based on the parameters of this experiment, my daily lineup creation required a large shift in paradigm. For the control team, I merely had to create a standard cash game lineup, like one that I would create on any other night. I was careful to pay attention to run totals, platoon splits and salary, particularly dollar to fantasy point values.
However, when constructing the max plate appearance lineup, or the independent variable of this experiment (back to 8th grade science!) I had a restricted player pool. Given the parameters I set up in the preview to the experiment for the independent group, I had to select players on the road hitting within the first four spots of the order. In case you can’t conceptualize what that does to the player pool, let me break it down for you.
- On a full slate of games, 15 teams are playing on the road, while the opposing 15 are in their home ballpark. Cut the pool in half.
- Of the half of the players available that are on the road, I may only select players that are hitting in the top four spots in the order. That means that I can only select 44% of the remaining players (a number that was already cut in half).
- Consider that of the 15 teams that are on the road, each can only start one position player. That left me with only 15 players per lineup spot, and many were filtered out by not hitting in the top four spots in the order.
While I want you to clearly understand the constraints of the experiment, and the drastic cut in player pool, it actually wasn’t impossible, as I was able to collect data over the past month 25 different times. As noted in the preview, some slates I was forced to pass over because of the constraints on the player pool.
I noted above that selecting players for the control group was no different from the process I would use every single night. Using my usual process and research methods, I went about selecting the players that I best saw fit based on salary and matchup. When creating the independent group, I was no longer able to be so picky about matchup and salary due to the constraints of the experiment. For as many times as it worked out perfectly that a player I was targeting for my normal team was on the road and in the first four spots in the order, there was an equal amount of times where I was forced to roster a player that I wouldn’t have used on a particular night just because he was the only one that fit the mold. Because of this, I sometimes was forced to limit both the floor and ceiling that surrounded the max plate appearance team.
One other important note about player selection during this experiment was the disregard I was essentially forced to use in terms of salary. When a player fit the parameters of the experiment and was a possible selection, I could only compare him to others at his position before deciding whether or not to put him in my lineup. On more than one occasion, there were only one or two players per position and I had to put salary and value on the backburner, swallow my pride and make the lineup work. Sometimes this meant that I was forced to sacrifice a little bit in terms of pitching (the safest and most predictable points) and other times this meant I needed to leave a ton of salary on the table, because there was seemingly no other choice.
After the teams were finally selected, I got to sit back and watch. There were some really awesome nights, and some really bad ones as you’ll be able to see from the table below. I wanted to make sure to show the great variance that came with doing an experiment like this, along with the standard variance that we assume to come from the game of baseball.
Below is a summary of the data.
|Team||FanDuel Average||Standard Deviation||High Score||Low Score|
Taking a look at this table, there are a few things that I want to point out. First off, the control group, or the standard lineup finished with a higher average score, along with a smaller standard deviation.
This is fully what I expected when starting this experiment. With the ability to truly mix and match players regardless of ballpark and lineup position, I was able to best take advantage of salary and find the best values to make an overall productive lineup. This is what we as DFS players strive to do every night. By selecting players who we deem to be in optimal situations, we expect a safe floor and a solid baseline night in and night out. This is what I got and anticipated from the control group, posting a higher average and a smaller standard deviation.
The independent group, or the max plate appearance lineup averaged just slightly less in terms of points from night to night, however it came with a much higher standard deviation and much more variance. Again, this is what I hypothesized prior to conducting the experiment. Based on the parameters and being unable to always select players in preferable matchups, I expected the scores from night in and night out to vary more, and come back with a lower average. To show just how crazy the variance was from night to night, I included the high and low scores from both the control and independent group, showing a staggering change that could happen on any night.
To help you visualize it, I’ve included below a screenshot of a night’s worth of games, with the max plate appearance team showing the immense upside of gathering a lot of plate appearances and having more opportunity, versus the typical cash game lineup.
This screenshot exemplifies exactly the goal of this experiment – to try and truly value “opportunity.” Yes, there are plenty of nights where simply having more chances does not equate to success (take for instance the night that I scored 5.5), however, when things do fall correctly, selecting players near the top of the order that are on the road, produces more opportunity and greater chance for more fantasy points.
In the preview of this experiment, I tasked myself to find the answer to one main question. After finding out just how valuable the volume of opportunity is as compared to the overall quality of opportunity, I wanted to be able to say whether or not a specific, schematic lineup strategy would be more or less successful than simply selecting a team that you deemed “the most optimal.” After doing a month’s worth of this experiment and collecting data for both the control and independent groups, I am able to conclude that simplifying lineup construction to a schematic strategy like the one used above, is not as successful over time as compared to building a lineup without strict construction parameters (outside of the salary cap). However, if the parameters of this experiment were tweaked to allow more free-form lineup construction while still adhering to a strategy of maximizing plate appearances, I believe you would see a positive increase in average score, with a decrease in standard deviation, pointing to a more consistently successful strategy. Editor’s Note: Unbeknownst to Logan, I’ve already volunteered him to tweak this experiment over the first month of next season, loosening the constraints a bit to make the experiment more indicative of the nightly dilemma in cash games (better lineup spot or better matchup) and placing an emphasis on tracking the actual number of plate appearances between the two lineups.