paper-dynasty-card-creation/test_retrosheet_arms.py

"""
Validation script for Retrosheet-based OF arm ratings

This script:
1. Loads 2005 Retrosheet events data
2. Calculates arm ratings using the new Retrosheet method
3. Compares against known strong/weak arms from 2005
4. Generates a detailed report

Usage:
    python test_retrosheet_arms.py
"""

import pandas as pd
from defenders.retrosheet_arm_calculator import (
    calculate_of_arms_from_retrosheet,
    calculate_position_baselines,
    calculate_player_arm_rating
)


def main():
    print("="*80)
    print("RETROSHEET ARM RATING VALIDATION - 2005 SEASON")
    print("="*80)
    print()

    # Load Retrosheet events
    print("Loading Retrosheet events data...")
    events_file = 'data-input/retrosheet/retrosheets_events_2005.csv'
    df_events = pd.read_csv(events_file, low_memory=False)
    print(f"Loaded {len(df_events):,} events")
    print()

    # Calculate arm ratings
    print("Calculating arm ratings for all outfielders...")
    arm_ratings = calculate_of_arms_from_retrosheet(df_events, season_pct=1.0)
    print(f"Calculated ratings for {len(arm_ratings)} outfielders")
    print()

    # Distribution summary
    print("ARM RATING DISTRIBUTION")
    print("-"*80)
    ratings_series = pd.Series(list(arm_ratings.values()))
    for rating in sorted(ratings_series.unique()):
        count = (ratings_series == rating).sum()
        pct = count / len(ratings_series) * 100
        stars = '*' * int(pct / 2)
        print(f"Rating {rating:+2d}: {count:3d} players ({pct:5.1f}%) {stars}")
    print()

    # Show top and bottom performers
    print("TOP 20 ARM RATINGS (Elite Arms)")
    print("-"*80)
    sorted_players = sorted(arm_ratings.items(), key=lambda x: x[1])
    top_20 = sorted_players[:20]

    for i, (player_id, rating) in enumerate(top_20, 1):
        # Get their stats
        player_stats = get_player_arm_stats(df_events, player_id)
        print(f"{i:2d}. {player_id:12s} Rating: {rating:+2d}  "
              f"Assists: {player_stats['assists']:2d}  "
              f"Home Throws: {player_stats['home_throws']:2d}  "
              f"Assist Rate: {player_stats['assist_rate']:.2f}%  "
              f"Balls: {player_stats['balls_fielded']:3d}")
    print()

    print("BOTTOM 15 ARM RATINGS (Weak Arms)")
    print("-"*80)
    bottom_15 = sorted_players[-15:]
    for i, (player_id, rating) in enumerate(bottom_15, 1):
        player_stats = get_player_arm_stats(df_events, player_id)
        print(f"{i:2d}. {player_id:12s} Rating: {rating:+2d}  "
              f"Assists: {player_stats['assists']:2d}  "
              f"Assist Rate: {player_stats['assist_rate']:.2f}%  "
              f"Balls: {player_stats['balls_fielded']:3d}")
    print()

    # Known strong arms to validate
    print("VALIDATION: Known Strong Arms (2005)")
    print("-"*80)
    known_strong_arms = [
        ('suzui001', 'Ichiro Suzuki', 'RF', 'Gold Glove, legendary arm'),
        ('crawc002', 'Carl Crawford', 'LF', 'Multiple stolen base deterrent'),
        ('edmoj001', 'Jim Edmonds', 'CF', 'Gold Glove, strong arm'),
        ('guerv001', 'Vladimir Guerrero', 'RF', 'Feared arm'),
    ]

    for player_id, name, pos, notes in known_strong_arms:
        rating = arm_ratings.get(player_id, 'NOT FOUND')
        if rating != 'NOT FOUND':
            stats = get_player_arm_stats(df_events, player_id)
            print(f"{name:20s} ({pos}): Rating {rating:+2d}  "
                  f"Assists: {stats['assists']:2d}  "
                  f"Note: {notes}")
        else:
            print(f"{name:20s} ({pos}): NOT FOUND IN DATA - Check player_id")
    print()

    # Position breakdown
    print("ARM RATINGS BY POSITION")
    print("-"*80)
    position_stats = get_position_breakdown(df_events, arm_ratings)
    for pos, stats in position_stats.items():
        print(f"\n{pos}:")
        print(f"  Total Players: {stats['count']}")
        print(f"  Average Rating: {stats['avg_rating']:.2f}")
        print(f"  Std Dev: {stats['std_rating']:.2f}")
        print(f"  Elite Arms (≤-3): {stats['elite_count']}")
        print(f"  Weak Arms (≥+2): {stats['weak_count']}")
    print()

    # Detailed player report
    print("DETAILED PLAYER REPORT (Elite Arms Only)")
    print("-"*80)
    elite_players = [(p, r) for p, r in arm_ratings.items() if r <= -3]
    elite_players.sort(key=lambda x: x[1])

    for player_id, rating in elite_players:
        print(f"\nPlayer: {player_id} (Rating: {rating:+2d})")
        detailed_stats = get_detailed_player_stats(df_events, player_id)
        for key, value in detailed_stats.items():
            print(f"  {key:20s}: {value}")

    print("\n" + "="*80)
    print("VALIDATION COMPLETE")
    print("="*80)


def get_player_arm_stats(df_events, player_id):
    """Get basic arm stats for a player across all OF positions."""
    stats = {
        'assists': 0,
        'home_throws': 0,
        'balls_fielded': 0,
        'assist_rate': 0.0
    }

    for a_col, po_col, f_col in [('a7', 'po7', 'f7'), ('a8', 'po8', 'f8'), ('a9', 'po9', 'f9')]:
        player_plays = df_events[df_events[f_col] == player_id]

        if len(player_plays) > 0:
            fielder_num = int(a_col[-1])
            assists = player_plays[player_plays[a_col] > 0].shape[0]
            balls = player_plays[(player_plays[po_col] > 0) | (player_plays[a_col] > 0)].shape[0]

            home = player_plays[
                (player_plays[a_col] > 0) &
                ((player_plays['brout1'] == fielder_num) |
                 (player_plays['brout2'] == fielder_num) |
                 (player_plays['brout3'] == fielder_num))
            ].shape[0]

            stats['assists'] += assists
            stats['home_throws'] += home
            stats['balls_fielded'] += balls

    if stats['balls_fielded'] > 0:
        stats['assist_rate'] = stats['assists'] / stats['balls_fielded'] * 100

    return stats


def get_detailed_player_stats(df_events, player_id):
    """Get detailed breakdown of arm-related plays."""
    detailed = {}

    for pos_name, a_col, po_col, f_col in [
        ('LF', 'a7', 'po7', 'f7'),
        ('CF', 'a8', 'po8', 'f8'),
        ('RF', 'a9', 'po9', 'f9')
    ]:
        player_plays = df_events[df_events[f_col] == player_id]

        if len(player_plays) == 0:
            continue

        fielder_num = int(a_col[-1])

        balls_fielded = player_plays[
            (player_plays[po_col] > 0) | (player_plays[a_col] > 0)
        ].shape[0]

        if balls_fielded < 50:
            continue

        assists = player_plays[player_plays[a_col] > 0].shape[0]

        throwouts = player_plays[
            (player_plays[a_col] > 0) &
            ((player_plays['brout1'] == fielder_num) |
             (player_plays['brout2'] == fielder_num) |
             (player_plays['brout3'] == fielder_num) |
             (player_plays['brout_b'] == fielder_num))
        ].shape[0]

        home_throws = player_plays[
            (player_plays[a_col] > 0) &
            ((player_plays['brout1'] == fielder_num) |
             (player_plays['brout2'] == fielder_num) |
             (player_plays['brout3'] == fielder_num))
        ].shape[0]

        batter_extra = player_plays[
            (player_plays[a_col] > 0) &
            (player_plays['brout_b'] == fielder_num)
        ].shape[0]

        detailed[f'{pos_name} Balls Fielded'] = balls_fielded
        detailed[f'{pos_name} Total Assists'] = assists
        detailed[f'{pos_name} Throwouts'] = throwouts
        detailed[f'{pos_name} Home Throws'] = home_throws
        detailed[f'{pos_name} Batter Extra Outs'] = batter_extra
        detailed[f'{pos_name} Assist Rate'] = f"{assists/balls_fielded*100:.2f}%" if balls_fielded > 0 else "N/A"

    return detailed


def get_position_breakdown(df_events, arm_ratings):
    """Get statistical breakdown by position."""
    positions = {}

    for pos_name, f_col in [('LF', 'f7'), ('CF', 'f8'), ('RF', 'f9')]:
        fielders = df_events[f_col].dropna().unique()
        pos_ratings = []

        for fielder in fielders:
            if fielder in arm_ratings:
                # Check if they qualified at this position
                player_plays = df_events[df_events[f_col] == fielder]
                a_col = f'a{f_col[-1]}'
                po_col = f'po{f_col[-1]}'
                balls = player_plays[
                    (player_plays[po_col] > 0) | (player_plays[a_col] > 0)
                ].shape[0]

                if balls >= 50:
                    pos_ratings.append(arm_ratings[fielder])

        if pos_ratings:
            positions[pos_name] = {
                'count': len(pos_ratings),
                'avg_rating': pd.Series(pos_ratings).mean(),
                'std_rating': pd.Series(pos_ratings).std(),
                'elite_count': sum(1 for r in pos_ratings if r <= -3),
                'weak_count': sum(1 for r in pos_ratings if r >= 2)
            }

    return positions


if __name__ == '__main__':
    main()