Pushing the deadline to get bots submitted to October 1st.
Run LeagueRunner.py to simulate fantasy football leagues. Requires Python 3.4.3 to run.
robots/Robot1.py is the example class that you can add your own logic to (and rename)
The idea for the first season is that the robots only handle drafting, so you only need to implement the one method: draft_players. The method takes in a list of available players and your currently drafted team. Players are sorted by average draft position (ADP), you can think of their ADP as an identifier instead of their name. Before any leagues are ran the AI is passed a list of player stats and projections, also listed by adp. Use these stats, projections, and player's average draft position to determine your optimal draft selection.
To evaluate run LeagueRunner.py and the output indicates the percent of playoffs made and percent of leagues won.
By default it will run against 2014 data (your AI is passed in 2014 projections and can see historic stats from 2013 and back).
The league runner fills in the robots to reach the needed 12 players with AdpDrafter.py which simply drafts by average draft position.
Feel free to make any improvements to the code (while avoiding breaking existing APIs) or documentation and submit them.
To enter your AI just implement the drafting AI and submit your .py file
For the first season it would be great to open source the AIs, but we should be able to work with .pyc if needed.
September 10th: First NFL game, you must submit your AI by this date
- League size is set to 12 (this doesn't mean only 12 people can submit AIs, there is no limit on people submitting AIs and they can be shuffled in and out)
- Scoring is the standard found at sites like ESPN, but with 0.5 points per reception
- There are 12 regular season weeks
- The top four teams by wins and then points scored go to playoffs
- No divisions
- Playoff match-ups are 2 weeks long
- Decimal scoring is used
- Projections are given (contains some but not all individual stats)
- Yearly data is given for all years that a players has played
- Weekly data is given for the past 3 seasons
Make sure to check for None being returned anywhere if there is no data for that player stat
Uncomment the block in robot1.py to see the data printed out for one player of each position.
Player stats passed as a list of PlayerHistory (player_history_init)
PlayerHistory
.position => string in {"QB", "RB", "WR", "TE", "K", "DEF"}
.adp
.current_year
.current_team
.yearly_data => PlayerYear
PlayerYear
.season_totals => PlayerSeasonStats
.weekly_data => PlayerWeeklyStats
PlayerSeasonStats extends PlayerStats
.games_played
.points_per_game
.team
PlayerWeeklyStats extends PlayerStats
.home_game => true/false
.opponent => string
.team_won => true/false
.team_score
.opponent_score
PlayerStats
.points
.completions
.attempts
.completion_percent
.passing_yards
.passing_td
.interceptions_thrown
.rushing_attempts
.rushing_yards
.rushing_avg
.rushing_td
.receiving_targets
.receiving_receptions
.receiving_yards
.receiving_avg
.receiving_td
.fg_make
.fg_attempt
.fg_percent
.ep_made
.ep_attempt
.sacks
.fumble_recovery
.interception_caught
.defensive_td
.points_allowed
.passing_yards_allowed_per_game
.rushing_yards_allowed_per_game
.safeties
.kick_return_td
TeamData (static to get information about a team)
.head_coach
.bye_week
[ Usage: team_data = TeamData.get_team_data(player_team, current_year) ]
And any of the above APIs will return None if there wasn't data for it.
(See PlayerClasses.py to dig in further)
The winning AI is the one which wins the highest percentage of leagues. We plan to run at least 100,000 league simulations to take luck out of the equation as much as possible.
- No player names, the idea of creating these AIs is to evaluate players more anonymously based on data which should allow crossover from year to year
- No random so that we can get consistent results
- One submission per person
- No in season management means no injury subs, no benches, no waivers, etc
Reports should be able to be generated to give an idea of how well each AI is doing and possibly other interesting stats on how their drafting has helped / hurt them so far. (To be implemented)