---
title: Dota 2 Edge Ensemble
emoji: 🎮
colorFrom: red
colorTo: green
sdk: gradio
sdk_version: "5.12.0"
app_file: app.py
pinned: false
license: mit
---

# Dota 2 Edge — Multi-Model Ensemble Predictor

Predict professional Dota 2 match outcomes using a 5-model ensemble
(XGBoost, LightGBM, CatBoost, LogReg, MLP) with Stacking meta-learner.

## Prediction Moments

| Moment | Description | Stacking AUC |
|--------|-------------|-------------|
| m0 | Draft (pre-game) | 0.587 |
| m1 | @10 minutes | 0.735 |
| m15 | @15 minutes | 0.807 |
| m2 | @20 minutes | 0.838 |
| m3 | @30 minutes | 0.963 |

## Dataset

Trained on 1,645 professional Dota 2 matches from OpenDota API.
Dataset available at [Finish-him/dota2-pro-matches](https://huggingface.co/datasets/Finish-him/dota2-pro-matches).