GPP Stacking with Park Factors: A Conclusion
In 2015, Logan won an FSTA Award for “Best Recurring Series” by way of his work here with Park Factors. This is the conclusion of that series from the 2015 MLB season. Check it out.
How many times can the fat lady sing in one week? This week I was the victim of her song not once, but twice. After coming to a conclusion in my Victiv PGA experiment, it’s now time to put a bow on the “DraftKings GPP Stacking with Park Factors experiment.”
For the entirety of this season, I’ve been toiling over at DraftKings and experimenting with some different techniques for tournament stacking. The strategies I’ve used have fallen primarily on the shoulders of park factors. For those that don’t know, a park factor is a relative rating of a ball park’s aptitude for offense. Unlike other sports, MLB game play is greatly influenced by the venue, and as DFS players, we’re inherently aware of that fact.
Using the buzz that surrounds games at Coors Field or Yankee Stadium, and the same lack thereof for offenses in Petco Park or Safeco Field, I set out to experiment with stacking lineups from the best and worst parks on a given slate. To have a standard measure of park factor, I slightly manipulated the FanGraphs park factors by handedness and got a relative ranking of each ball park. The rest was easy. On a given night, I would stack a lineup on DraftKings that featured players from the best park on the slate, as well as a stack from the worst park according to my rankings.
I recorded my scores and winnings and lived to tell you guys about it each week, touching on some of the things I thought were most important from week to week. If you’d like to take a look back at any individual week, you can choose any of the previous articles by selecting “Park Factors” from the MLB Free dropdown menu.
Since this is a turning of the page event, instead of focusing on the small success or failures from this week, I want to look at the big picture and discuss some of the bigger themes that I’ve focused on since the beginning of the year.
First however, I want to show you how the last month and a half ended up. The table below shows the ROI, average scores, as well as the standard deviation for both of the respective strategies.
|Strategy||ROI||Avg. Score||Standard Deviation|
While the goal of the experiment wasn’t to break the bank, I’m not disappointed that I actually walked away with some extra money. Thanks to a big hit from a “worst park” stack early in the year, I was able to operate in the black for the experiment’s duration.
The actual goal of this experiment was to make it as similar to a real scientific experiment as possible. That means, I created some hypotheses, tested them each week, and can draw conclusions that support or reject them.
In this regards, I was very successful.
Here are the two main hypotheses I made in the first few weeks and some conclusive analysis after gathering all the data from the experiment.
1. Stacks including the teams from the best park will yield higher average scores, will be more consistent and have a chance to cash more frequently than those stacks from the worst parks.
2. Stacks including the teams from the worst park carry high upsides thanks to the inherited low ownership percentages associated with parks that feature lower scoring game.
The experimental data speaks for itself and allows me to conclude that both of these hypotheses are true and accurate.
How do I know that?
First, when looking at the best park data, I can make a couple of important notes.
- The best park averaged nearly seven fantasy points more night.
- The best park had a smaller standard deviation than the stacks from the worst park, resulting in a more consistent scores.
- The best park compiled a positive ROI.
All three of these points are important to help confirm the hypothesis I derived for the best park. The best park carried a higher average score, a more consistent set of scores and still compiled a positive ROI.
In fact, to compare just how well the best park average would have done on a nightly basis, I took the average cashing score from the tournaments I entered for the last month and cross-referenced it with all of the scores I had ever collected from best park stacks.
Using this process, which yielded an average cashing score of 113.43 best park stacks would have cashed at a 35% clip. On the other hand, the worst park would have only cashed at a rate of 23%.
While the best park stacks bested the worst parks in cash rate, the worst park stacks accumulated a much higher ROI during the experiment, and mainly thanks to one big score.
This is what I anticipated with my second hypothesis about stacking the worst parks. Low ownership levels create a lot of upside and the potential to separate yourself from the field if a game at a bad park happens to yield a lot of runs. While using these stacks might work out less often, they present the chance to reap big rewards.
In general, the experiment as a whole speaks to the relative success of stacking as a strategy. Because of the correlated at-bats for hitters on the same team during a game, and the importance tied to statistics such as runs and RBIs that can only be acquired with the help of a teammate (unless one player is just hitting a bunch of home runs) stacking is a viable and potentially lucrative strategy for large field tournaments.
The stacking in this experiment though, was perhaps a bit extreme. While I was fortunate enough to acquire positive ROIs, the best course of action while stacking a team might be to not only look for great run scoring environments, but to also mix and match them with the best possible players with the best opportunities. Since I was strict in my process and only selected players from a particular park, sometimes I had to take a player with a less than ideal matchup in lieu of selecting one from a better park and in a better matchup.
However, regardless of the extreme nature, the experiment was an overwhelming success. Even though I might not be stacking entire parks on a nightly basis, I now realize the importance and the potential that stacks of this nature carry.