Dreamteam

Dreamteam is a pulp-based integer programming formulation for the generalised dream team problem. It is based on the premise that for a given 'expected score' per player and price per player, acting within the selection rules of the game and some salary cap requirement, an optimal team exists and can be selected

Installation

1. Clone or download the repository

2. Install with pip

   pip install path/to/cloned/repository
   

Usage

To use Dreamteam to create a team:

• Sign up for a dreamteam/fantasy sports account of your choice.
• Scrape or otherwise extract the player data for all players available for selection. This can often be done simply by viewing the requests sent on the dreamteam website you are using and finding the one which has json data for all players.
• Create a new Python 3 virtualenv and install the dreamteam package (see Installation above)
• Create a list of Position objects for each position in the comp that you need to put a specific number of players into (e.g. forwards, backs, midfielders)

from dreamteam import Position
positions = [Position('Forward'), Position('Back')]

• Create a list of Squad objects for each squad in the comp that you need to create players for (e.g. Main Team, Reserves, Bench, etc). For each squad, provide a multiplier to apply to the expected score for players in that squad. For example, you may want to have a multiplier of 1.0 for the Main Team and 0.5 for bench players to encourage the solver to put more emphasis on the selection of players in the Main Team squad. For each squad, also supply a dictionary of Position: int representing the number of players which are required to be selected for each position in that squad (for example, the Main Team may need 4 Forwards and 6 Backs).

from dreamteam import Squad
squads = [
    Squad('Main Team', value_multiplier=1.0, position_caps = {prositions[0]: 4, positions[1]:6)
]

• Create a list of players, which should include every player available for selection (as found in Step 2). For each player, you will need to provide a name, the team that the player plays for, the elgible squads the player can play in, the eligible positions the player can be selected for, and an expected score. The expected score is something that you will need to determine via your own methods. The dream team website may provide an estimate in the json data that you scrape, or you may wish to use your own advanced machine learning methods to determine this, or anything in between.

from dreamteam import Player
players = [
    Player('Jeff', 'Harvey', 'Broncos', 70, positions[1:], squads),
    Player('Brett', 'Jones', 'Cowboys', 85, positions[:1], squads),
    ...
]

• Pass your list of positions, squads and players to the engine and wait for it to return. If your dream team comp has the concept of a 'captain' who receives a bonus score, you can also pass the captain_value_bonus kwarg to the engine (0.0 is no bonus, 1.0 would be a captain that gets 'double score')

from dreamteam import solve
problem = solve(
    players, positions, squads, salary_cap=120000, captain_value_bonus=0.0,
)

• Once the engine returns the solved problem, you can inspect the problem object itself (which is returned straight from Pulp) or simply inspect your player objects which will now have been mutated and have the properties 'selected_in_solution', solution_position, solution_squad and solution_captain set on them (which will be False, None, None, and False respectively for players not selected in the solution)

solution_players = [p for p in players if p.selected_in_solution]
for player in solution_players:
    print(player)

License

Dreamteam is supplied as open source under the MIT license. See LICENSE for details.

API Documentation

API Documentation is a work in progress. Please feel free to inspect the code in the meantime (there are only two fairly small modules)

Mathematical Formulation

The integer programming (IP) model that is used to solve the Dreamteam problem will be included as part of the API documentation when it is complete.