Tango Game AI Bot

This repository contains an AI bot developed in Python to automatically solve the Tango game available on LinkedIn.

How It Works

• Board Reading: Uses pyautogui to capture the game board.
• Edge Prediction: Applies a trained CNN to detect "equals" and "crosses" on the board edges.
• Sun & Moon Detection: Identifies initial sun and moon positions by checking the color of certain pixels in the miniboxes.
• Optimization: Formulates the board state as a MILP problem and solves it using Gurobi.
• Automation: Automatically clicks the solution on the board.

Example

Below an example of how the Bot works in a real Tango game.

example.mp4

Repository Structure

• mouse_info.py – Utilities to select the board pixels for pyautogui.
• data_loader.py – Dataset and preprocessing utilities.
• model.py – CNN model architecture.
• train.py – Model training script.
• cnn.pth – Pytorch trained model.
• screenreader.py – Board reading and prediction logic.
• main.py – MILP model and clicking automation.
• tango_solver.py – Deprecated attempt using html information.

Requirements

• Python 3.8+
• PyTorch
• torchvision
• pyautogui
• numpy
• pandas
• Gurobi (with license)

Usage

1. Train the Model:
   Run train.py to train the CNN on your labeled data.

2. Run the Bot:
   Execute the main bot script to read the board, predict edges, solve the MILP, and play the game.