README.md

metadata

license: mit
title: Football Elite Player Predictor
sdk: gradio
emoji: 👀
colorFrom: green
colorTo: blue
short_description: Predicting Elite Receivers

title: Football Elite Player Predictor

emoji: 🏈

colorFrom: green

colorTo: blue

sdk: gradio

sdk_version: 4.0.0

app_file: app.py

pinned: false

license: mit

---

Football: Will this player be Elite? 🏈

Predict whether a football player will be classified as "Elite" based on their performance statistics using an AutoGluon TabularPredictor model.

Overview

This application uses machine learning to classify football players as either Elite or Not Elite based on their receiving statistics. The model analyzes 8 key performance metrics to make predictions with confidence probabilities.

Features

• Real-time Predictions: Enter player stats and get instant classification
• Probability Scores: See confidence levels for each class
• Interactive Interface: Adjust sliders and inputs to explore different scenarios
• Example Players: Pre-loaded examples including star players and benchmarks

Input Features

The model uses the following 8 statistics:

1. Targets (TGT): Number of passes thrown to the player
2. Receptions (REC): Number of catches made
3. Yards (YDS): Total receiving yards
4. Yards Before Catch per Reception (YBC_R): Average yards before catch
5. Yards After Catch per Reception (YAC_R): Average yards after catch
6. Average Depth of Target (ADOT): Average distance from line of scrimmage
7. Drop Percentage (DROP_PCT): Percentage of dropped passes
8. Rating (RAT): Overall passer rating when targeted

Model

• Framework: AutoGluon TabularPredictor
• Task: Binary Classification (Elite vs Not Elite)
• Output: Class prediction with probability distribution

How to Use

1. Enter a player name (optional, for tracking)
2. Adjust the statistical inputs using sliders and number fields
3. View the real-time prediction and probability scores
4. Try the example players to see different scenarios

Examples Included

• Justin Jefferson: Elite receiver profile
• Cooper Kupp: High-volume elite target
• Rookie WR: Developing player profile
• Tyreek Hill: Elite deep threat profile
• Bench Player: Minimal playing time

Technical Details

The model is loaded from Hugging Face Hub and makes predictions using ensemble methods via AutoGluon's TabularPredictor.

Limitations

• Model performance depends on training data quality and representativeness
• Predictions are probabilistic and should not be used as sole decision-making criteria
• Statistics should be from comparable game situations and sample sizes

Acknowledgments

Built with AutoGluon and Gradio.

---