How to Draft the Perfect Team: Part I (The Hypothesis)

Malcolm McMillan (@McLateralFF)
The Lateral Editor

How to Draft the Perfect Team

Part I: The Hypothesis

Drafting is obviously important when building a championship roster. Yes, you need to still use FAAB or waiver claims and make trades to fix mistakes or capitalize on new trends, but you cannot rely only on add/drops and trades. That is like deciding to run a marathon with a 20 lb weight around your neck. If you are really good, maybe you will win, but it is much more likely that someone else is going to win that race.

Would it not be great then, if there was a way to map out the perfect draft? Or at least get really close? What if instead of being the person who took Le'Veon Bell at ADP RB18 you were the person who Melvin Gordon two spots later? Could you reasonably have reached that conclusion in the first place? Quite possibly! With only a two-spot difference in ADP, there is certainly evidence that people rated the two RBs similarly.

What about the less logical choices though? In hindsight, you would have loved to have locked up future RB4 James Robinson in round 13 instead of his teammate RB Ryquell Armstead. But would you have logically ever done that? Would you have reached five rounds to snag an undrafted free agent that was going at ADP 222 overall? If we are being honest, the answer is probably no. You were much more likely to snag Robinson as a waiver add in the preseason or after Week 1.

This gets at the core of the issue: can you logically draft perfectly, or because of unpredictable variables, is it simply impossible to reasonably draft a championship roster? That is the question we will be trying to answer over the next several articles.

The Scientific Method

For those of you who do not remember high school biology anymore, the scientific method looks something like this:
  1. Develop a research question
  2. Formulate a hypothesis
  3. Design an experiment to prove the hypothesis
  4. Perform the experiment to prove the hypothesis
  5. Conclude whether or not the hypothesis was proven
  6. Report the results of the experiment
Below is how I will apply the scientific method to attempt to create the perfect draft:

Develop a Research Question

The research question is: Is it possible to draft a championship roster using logic?

Formulate a Hypothesis

My hypothesis is that it is possible to create a championship roster using available data and logic rather than solely on hindsight. 

Design an Experiement

Below are the parameters for this experiment to create a championship roster:
  • This is a draft for a redraft team going into the 2020 season
  • The format for this team will be PPR
  • The league is a 12-team league
  • The roster construction will be 1 QB, 2 RB, 3 WR, 1 TE, 1 FLEX (W/R/T), 1 K, 1 DST, and 6 bench spots
    • Therefore, this hypothetical draft will be a 16-round draft
  • This roster will be constructed drafting from the sixth spot in a snake draft
  • In the first ten rounds, players can only be selected that have an ADP equal to the ADP of the pick or an ADP of up to one round later from the respective pick 
    • E.g. If picking 1.06, I can select anyone with an ADP of ADP6 to ADP19
  • In the final six rounds, players can only be selected that have an ADP equal to the ADP of the pick or an ADP of up to two rounds later from the respective pick
    • E.g. If picking 11.06, I can select anyone with an ADP of ADP131 to ADP158
    • As there are ten starting spots and six bench spots, this is to simulate fantasy managers getting riskier with their later round/depth picks
  • Initial ADP will be sourced from FantasyPros ADP
    • This itself is sourced from ESPN, RTSports, and Fantrax
  • ADP will also be sourced from fantasydata and Fantasy Football Calculator when possible to account for any skewed data
  • Playoff weeks will be considered Weeks 14-16
    • Week 17 data will be ignored entirely for this experiment
  • Playoff rounds will be one week per round
  • Players will be selected round by round
    • I.e. I cannot retroactively change a pick because a bye week/bad performance issue popped up. If that is an issue, I need to be mindful of it when making the pick
Perform the Experiment

To perform the experiment:
  1. I will draft and fill a roster based on the parameters I established in my experiment design. 
  2. The players will be selected to score the most points on a week-to-week basis. 
  3. After all the players have been drafted, I will construct weekly scoring totals for my hypothetical teams.
  4. I will then compare those point totals against the best possible rosters for each week and whatever other historical data I can find
  5. Based on step 4, I will construct a likely record for my drafted team over the course of a season
Conclude if the Hypothesis Was Proven

The hypothesis will be considered proven if the drafted team is determined to have won enough games to make the playoffs, advance to the championship game, and then subsequently win the championship.

Report the Results

Results will be reported over the coming parts of this multi-part series. This will likely consist of multiple parts for both the drafting process and the in-season performance analysis in the interest of producing easily consumable content while still being thorough.

Important Caveats

Here are some caveats that are important to keep in mind:
  • In real drafts, people act irrationally
    • For example, I cannot draft Alvin Kamara in this draft as he had an ADP of 1.05, and I will be drafting 1.06. However, in a real draft, there would be plenty of instances where a fantasy manager drafted a player with a higher ADP (higher ADP = larger number) such as RB Dalvin Cook or WR Davante Adams
  • These ADP sources are not perfect
    • These data sets are (understandably) unable to account for all fantasy drafts conducted in 2020. However, I feel confident enough that they represent a properly sized and reputable sample of the population
  • Limiting myself to certain ranges of ADP per pick is an arbitrary choice
    • There is no scientific data I am aware of that suggests that logically people will likely pick within a round of a player's ADP (or two rounds in the final six rounds). However, I do stand by the inherent logic of the parameter, as expanding my choices would allow me to reach for people I likely know scored more in 2020 due to hindsight, rather than logical conclusions. This would defeat the purpose of the experiment
  • Determining weekly "wins" will have to be at least partially subjective
    • Again, it is tough (impossible) to account for every possible permutation of winning rosters in a given week. However, it is possible to construct the best possible roster in a given week, and then use that as a launching point for further analysis.
  • While I have selected most sources for my data, as of this writing I have not selected all my sources
    • There is a chance that some historical data I am hoping to find will simply not exist. However, I do believe there is enough data out there based on what I do already have available, to credibly prove or disprove my hypothesis
  • Finally, this is ultimately still done in hindsight. My goal is to create as realistic a roster as possible, but hindsight cannot be eliminated completely. You will just have to accept that to appreciate the thought exercise

Follow The Lateral on Twitter (@TheLateralFF)