confpickem - NFL Confidence Pick'em Analyzer A Python package for analyzing and optimizing picks for NFL Confidence Pick'em pools. This package provides tools for: - Scraping Yahoo Pick'em league data - Analyzing pick distributions and trends - Simulating outcomes and optimizing picks - Evaluating different picking strategies

Installation

You can install the package using pip:

pip install confpickem

Quick Start

Command Line Interface (Recommended)

The easiest way to use confpickem is through the command-line tools:

# Optimize your picks (analytical optimizer, runs in seconds)
confpickem --week 10 --mode beginning

# Re-optimize once games start -- locks what's finished or kicked off
confpickem --week 10 --mode midweek

# Check win probabilities for all players
confpickem-win-probability --week 10 --live-odds

# Update player skills from historical data
confpickem-player-skills update --years 2024,2025

New here? docs/USAGE.md walks through a full week end to end (getting cookies.txt, the weekly run, reading the output).

Installation: Install the package to get these commands:

pip install -e .   # From project root

See the CLI Documentation for full details.

Python API

import json
from confpickem import YahooPickEm, ConfidencePickEmSimulator, Player
from confpickem.yahoo_pickem_integration import convert_yahoo_to_simulator_format

yahoo = YahooPickEm(week=10, league_id=YOUR_LEAGUE_ID, cookies_file='cookies.txt')

sim = ConfidencePickEmSimulator(num_sims=2000)
sim.add_games_from_dataframe(convert_yahoo_to_simulator_format(yahoo, ignore_results=True))

skills = json.load(open('current_player_skills.json'))   # or {} for average skills
sim.players = [
    Player(nm,
           skills.get(nm, {}).get('skill_level', 0.6),
           skills.get(nm, {}).get('crowd_following', 0.5),
           skills.get(nm, {}).get('confidence_following', 0.5))
    for nm in yahoo.players['player_name']
]

picks = sim.optimize_picks_analytic("Your Yahoo Name", verbose=True)
print(picks)   # {TEAM: confidence}, a full 1..N assignment

See examples/yahoo_pickem_demo.ipynb for the full flow (sanity checks, game importance, mid-week re-optimization).

Features

🎯 Unified CLI Tools

• optimize.py - Comprehensive pick optimization with live odds support
• win_probability.py - Monte Carlo win probability calculator
• player_skills.py - Historical performance analysis and skill modeling

📊 Yahoo Data Scraping

• Scrape pick distributions and crowd confidence levels
• Track actual picks and results from your league
• Cache responses to avoid excessive requests

🎲 Simulation and Analysis

• Monte Carlo simulation of game outcomes
• Player skill modeling and analysis
• Pick optimization algorithms
• Risk/reward and game importance analysis

🔴 Live Vegas Odds Integration

• Real-time betting line integration via The Odds API
• More accurate win probabilities than Yahoo spreads
• Automatic fallback to Yahoo data when API unavailable

🧠 Strategy Optimization

• Analytical Poisson-binomial P(win) optimizer (default; noise-free, runs in seconds)
• Optimize confidence point assignments
• Mid-week / mid-Sunday re-optimization — locks games already finished or kicked off
• --greedy / --hill-climb for the older simulation-based optimizers

Dependencies

• Python ≥ 3.8
• requests
• pandas
• numpy
• beautifulsoup4
• scipy

Methodology

The pick optimizer computes your probability of winning the week analytically (the weekly score is a Poisson-binomial distribution) rather than by Monte Carlo simulation. This matters: a naive hill climb on simulated win probability optimizes noise, not strategy, and never beats "pick the favorites." The analytical objective is noise-free, ~250× faster to evaluate, and in a 29-week backtest produced 2 outright weekly wins vs. the greedy optimizer's 1, at a mean finish of 34th vs. 43rd.

This is the default optimizer (confpickem --week N --mode beginning). It handles mid-week runs too — games already finished or kicked off are locked to your submitted picks and the rest optimized around them. Pass --greedy for the old sequential optimizer or --hill-climb for the slower simulation search.

See docs/optimization-methodology.md for the full write-up.

Documentation

• Usage Walkthrough - A full week end to end, from the command line
• Python API Demo - The same flow from Python
• Optimization Methodology - How the optimizer works and why
• CLI Tools Guide - Full flag reference for the command-line tools
• CLI Tools (in package) - Detailed CLI documentation
• GitHub Repository - Source code and issues

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.