Framework to use Artificial Intelligence (AI) to generate optimal fantasy football teams.

Getting started | Quick run

The goal of Fanta-AI is to suggest an optimal auction strategy to maximize your chances of winning the next season of your fantasy league. The project is divided into two main parts:

• Evaluation of all the players
• Auction strategy

All the current data and results are based on Lega Serie A, but you are invited to modify the framework in order to accommodate your favorite international league. Let's discuss step by step Fanta-AI using as example the Serie A stagione 22/23, and good luck with your next season!

If you are passionate for fantacalcio like me, you probably know that the winner of the fantasy league is mainly determined by luck, but in order to be a candidate for this lottery you have to build a solid team. To build an optimal team it is necessary to pick good players. The first step is then to estimate the potential of each player.

In this project there are 3 available alternatives for the player evaluation that can be selected using the --ev_type [manuale/fantagazzetta/piopy] argument, which correspond to:

• Manual evaluation: for hardcore players that know better than anyone else, the optimal strategy is to manually rank the players. To perform this operation the user has to follow the excel table Players_evaluation/Le_mie_valutazioni.xlsx, which has the following structure:

You can edit this file and change the rank (Valutazione) as you prefer. For future seasons and different leagues, you have to update the player list and their corresponding role.

• Fantagazzetta: you can download the list of players from fantacalcio.it. This has to be placed in Players_evaluation/Quotazioni_Fantagazzetta.xlsx. The evaluation of the players will be made according to the professional journalists of fantagazzetta.
• Fantaciclopedia (piopy): based on piopy tool, this evaluation uses the real statistics of the last two seasons, available on fantaciclopedia. To perform this type of evaluation you have to run

python Evaluate_players.py 

Notice that the output Players_evaluation/giocatori_excel.xls is highly informative and detailed, and can serve as a valuable tool in the arsenal of the good fantallenatore. I invite you to explore its details here.

After you have evaluated all the players according to your favorite metric, you have to build your team during the auction. This project is based on the following 2 assumptions to model the auction:

• All the auctioneers (your friends in the fantalega) are decent players and their evaluation will not be much different compared to yours.
• The auction is a classic English auction, with open ascending bids (la classica asta a chiamata)

Accepting this assumptions, we can model the fantacalcio auction as a knapsack problem which is a typical problem in combinatorial optimization.

To solve this problem, FantaAI implements a genetic algorithm that evolves to generate the best teams that you can realistically expect to build during your real auction. You can run it using

python Pick_my_team.py # --[arguments]

There are several arguments that can be set for the specific use case, and to develop your favorite strategy:

• --crediti : initial budget for each team
• --n[P/D/C/A] : number of players per role P/D/C/A for each team
• --max_b_[P/D/C/A] : maximum percentage of the budget to invest in the specific role P/D/C/A
• --t[P/D/C/A] : target of good players per role. An optimal team invest most of its budget in a few solid picks rater than spreading too much
• --bonus_multiplier : how much to weight the top players for each role
• --pop_size : size of the population for the genetic algorithm
• --num_gen : number of generations of evolution
• --mutation_rate : rate of mutations for the genetic algorithm
• --swap_mutation_rate : rate of recombinations for the genetic algorithm

1. Clone this repository

git clone https://github.com/SCiarella/FantaLega_Serie_AI.git

2. Setup virtual environment

cd FantaLega_Serie_AI/
python -m venv fantaAI-venv
source ./fantaAI-venv/bin/activate

3. Install requirements

pip install -r requirements.txt

1. (Optional) if you want to use piopy evaluation method

python Evaluate_players.py 

2. Pick an optimal team

python Pick_my_team.py

Fanta-AI implements as a possible model for player evaluation the project fantacalcio-py developed by piopy and based on the data from Fantaciclopedia