A sports betting environment for Farma Foundation Gymnasium based environments.
pip install oddsgym
The starting bank is X (X > 0), representing X available bets. Actions are all available bets for a game (depends on sport), placing 1 bet for each option. Also, the agent cannot bet over his current bank (i.e., can't place 3 bets when the current bank is 2).
For example, in 3-way betting for soccer, the available actions are:
1. Bet on home team
2. Bet on away team
3. Bet on draw
4. Bet on home team and away team
5. Bet on home team and draw
6. Bet on away team and draw
7. Bet on home team and away team and draw
8. Don't place a bet for this game
A step is placing a bet on a single game. In each step, the agent knows the betting odds for this game. The reward for each step is the amount of money won (negative reward when losing money).
An episode is betting for a whole year or when "striking out" (losing all the money).
Here is an example of how to use the ThreeWaySoccerOddsEnv environment:
import gymnasium as gym
from oddsgym.envs import ThreeWaySoccerOddsEnv
# Create the environment
env = ThreeWaySoccerOddsEnv(soccer_bets_dataframe)
# Reset the environment
obs, info = env.reset()
# Take a random action
action = env.action_space.sample()
obs, reward, done, truncated, info = env.step(action)You can also use the pre-registered environments directly with gym.make:
import gymnasium as gym
# Create the environment
env = gym.make("FootballDataDaily-v0")
# Reset the environment
obs, info = env.reset()
# Take a random action
action = env.action_space.sample()
obs, reward, done, truncated, info = env.step(action)To create a custom environment, you can extend the base classes provided by oddsgym. Here is an example of how to create a custom environment for a new sport:
import pandas as pd
from oddsgym.envs.base import BaseOddsEnv
class CustomSportOddsEnv(BaseOddsEnv):
sport = "custom_sport"
odds_column_names = ["outcome1", "outcome2", "outcome3"]
def __init__(self, custom_sport_bets_dataframe, *args, **kwargs):
odds = custom_sport_bets_dataframe[self.odds_column_names]
results = custom_sport_bets_dataframe["result"]
super().__init__(odds, self.odds_column_names, results, *args, **kwargs)
def create_info(self, action):
info = super().create_info(action)
# Add custom info here
return infoYou can register your custom environment with gym to use it like any other pre-registered environment:
import gymnasium as gym
from oddsgym.envs import CustomSportOddsEnv
gym.register(
id="CustomSport-v0",
entry_point="path.to.your.module:CustomSportOddsEnv",
max_episode_steps=365,
)To use the environments with rllib, you need to register them with rllib. Here is an example of how to register the FootballDataDailyEnv environment:
import gymnasium as gym
from ray import tune
from oddsgym.envs import FootballDataDailyEnv
# Register the environments with rllib
tune.register_env(
"FootballDataDaily-ray-v0",
lambda env_config: gym.wrappers.FlattenObservation(FootballDataDailyEnv(env_config))
)Once registered, you can use these environments with rllib models.
The full documentation is hosted at: https://oryjonay.github.io/Odds-Gym