stats.py

import polars as pl
from data import data_df

# from enum import Enum
from types import SimpleNamespace

from convert import verify_and_return_presult, verify_and_return_basic_ball_kind_code

class Player:
  PITCHER = 'pitcher'
  BATTER = 'batter'
  BOTH = 'both'

STATS = {}

def assert_value(value, options):
  assert value in options, f'Expected one of {options}, got {value}'

def register_stat(name, expr, percent, percentile, batted_ball=False):
  assert name not in STATS, f'"{name}" already registered, returns {STATS[name]}'
  assert_value(percentile, ('pitcher', 'batter', 'both', None))
  STATS[name] = dict(expr=expr.alias(name), percent=percent, percentile=percentile, batted_ball=batted_ball)

def get_stat(stat):
  return STATS[stat]

def get_stats(stats):
  return [get_stat(stat) for stat in stats]

def get_stat_val(stat, key, default=None):
  return STATS[stat][key] if stat in STATS else default

def get_stats_val(stats, key, default=None):
  return [get_stat_val(stat, key, default) for stat in stats]
  
valid_pitch = pl.col('x').is_not_null() & pl.col('y').is_not_null() & (pl.col('ballSpeed') > 0)
is_ball = pl.col('presult').is_in(verify_and_return_presult(['Ball', 'Walk']))
is_non_ball = pl.col('pitch') & ~is_ball # pitches that are not balls i.e. no catcher interference, etc.
is_two_str = pl.col('before_s') == 2 # named this way in case I use two_str for 2-Str%
first_count = (pl.col('before_s') == 0) & (pl.col('before_b') == 0)
is_bip_out = pl.col('presult').is_in(verify_and_return_presult([
  'Groundout', 'Flyout', 'Lineout', 'Groundout (Double play)',
  'Foul fly', 'Foul line (?)',
  'Sacrifice bunt', 'Sacrifice fly',
  "Fielder's choice", "Sacrifice fielder's choice"
]))
pa = pl.col('pa_code').unique().len()

# to-do: unify PA calculation
# pl.col('pa_code').unique().len() or pl.col('PA').first()
register_stat('FB Velo', pl.col('FB Velo').max(), False, Player.PITCHER)
register_stat('K%', pl.when(pl.col('presult').str.contains('strikeout')).then(1).otherwise(0).sum() / pl.col('pa_code').unique().len(), True, Player.PITCHER)
register_stat('BB%', pl.when(pl.col('presult') == 'Walk').then(1).otherwise(0).sum() / pl.col('pa_code').unique().len(), True, Player.BATTER)
register_stat('Swing%', pl.col('swing').sum() / pl.col('pitch').sum(), True, Player.BOTH)
register_stat('Z-Swing%', (pl.col('swing') & pl.col('zone')).sum() / pl.col('zone').sum(), True, Player.BATTER)
register_stat('Chase%', (pl.col('swing') & ~pl.col('zone')).sum() / (~pl.col('zone')).sum(), True, Player.PITCHER)
register_stat('Contact%', (pl.col('swing') & ~pl.col('whiff')).sum()/pl.col('swing').sum(), True, Player.BATTER)
register_stat('Z-Con%', (pl.col('zone') & pl.col('swing') & ~pl.col('whiff')).sum()/(pl.col('zone') & pl.col('swing')).sum(), True, Player.BATTER)
register_stat('O-Con%', (~pl.col('zone') & pl.col('swing') & ~pl.col('whiff')).sum()/(~pl.col('zone') & pl.col('swing')).sum(), True, Player.BATTER)
register_stat('Whiff%', pl.col('whiff').sum() / pl.col('swing').sum(), True, Player.PITCHER)
register_stat('SwStr%', pl.col('whiff').sum() / pl.col('pitch').sum(), True, Player.PITCHER)
register_stat('CSW%', pl.col('csw').sum() / pl.col('pitch').sum(), True, Player.PITCHER)
register_stat('Ball%', is_ball.sum() / pl.col('pitch').sum(), True, Player.BATTER)
register_stat('Strike%', is_non_ball.sum() / pl.col('pitch').sum(), True, Player.PITCHER)
register_stat('F-Str%', (is_non_ball  & first_count).sum() / first_count.sum(), True, Player.PITCHER)
register_stat('PAR%', ((is_two_str & pl.col('presult').str.contains('strikeout')).sum()) / is_two_str.sum(), True, Player.PITCHER)
register_stat('PLUS%', (pl.col('csw') | (pl.col('presult') == 'Foul') | is_bip_out).sum() / pl.col('pitch').sum(), True, Player.PITCHER)
register_stat('Behind%', ((pl.col('before_b') > pl.col('before_s')) & (pl.col('before_s') < 2) & (pl.col('before_b') > 1)).sum() / pl.len(), True, Player.BATTER)
register_stat('Zone%', pl.col('zone').sum() / pl.col('pitch').sum(), True, Player.PITCHER)
register_stat('Glove%', (pl.when(pl.col('pitLR') == 'r').then(pl.col('x') < 0).otherwise(pl.col('x') > 0)).mean(), True, None)
register_stat('Arm%', (pl.when(pl.col('pitLR') == 'r').then(pl.col('x') >= 0).otherwise(pl.col('x') <= 0)).mean(), True, None)
register_stat('High%', (pl.col('y') > 125).mean(), True, None)
register_stat('Low%', (pl.col('y') <= 125).mean(), True, None)
register_stat('MM%', (pl.col('x').is_between(-20, 20) & pl.col('y').is_between(100, 100+50)).mean(), True, None)
register_stat('Sec%', (pl.col('basic_ballKind_code').is_in(verify_and_return_basic_ball_kind_code(['BR', 'OS']))).sum() / pl.col('pitch').sum(), True, None) 
register_stat('GB%', pl.col('G') + pl.col('B'), True, Player.PITCHER, True)
register_stat('FB%', pl.col('F') + pl.col('P'), True, Player.BATTER, True)
register_stat('LD%', pl.col('L'), True, Player.BATTER, True)
register_stat('IFFB%', pl.col('P'), True, Player.PITCHER, True)
register_stat('OFFB%', pl.col('F'), True, Player.BATTER, True)
register_stat('AIR%', pl.col('F') + pl.col('P') + pl.col('L'), True, Player.BATTER, True)
register_stat('HR%', (pl.col('presult') == 'Home run').sum() / pa, True, Player.BATTER)
register_stat('HR/FB', (pl.col('presult') == 'Home run').sum() / (pl.col('aux_bresult').struct.field('batType').is_in(['F', 'P'])).sum(), True, Player.BATTER)
# note: to match bouno-san's data, HR/FB bust be /FB for pitchers and /OFFB for batters; 
# /OFFB matches both player types for DeltaGraphs, so we use /OFFB in implementaiton and /FB in na,e

# register_stat('Usage', pl.col('count')/pl.sum('count').over('pitId'), True, None)
register_stat('Usage', pl.len()/pl.first('Pitches'), True, None)
register_stat('Avg Velo', pl.when(valid_pitch).then('mph').mean(), False, None)
register_stat('Max Velo', pl.col('mph').max(), False, None)


def filter_data_by_date_and_game_kind(data, start_date=None, end_date=None, game_kind=None):
  if start_date is not None:
    data = data.filter(pl.col('date') >= start_date)
  if end_date is not None:
    data = data.filter(pl.col('date') <= end_date)
  if game_kind is not None:
    data = data.filter(pl.col('coarse_game_kind') == game_kind)
  return data

def compute_team_games(data):
  data = (
      data
      .with_columns(
          pl.col('gameId').unique().len().over('HomeTeamNameES').alias('home_games'),
          pl.col('gameId').unique().len().over('VisitorTeamNameES').alias('visitor_games')
      )
  )
  game_data = (
      data
      .group_by('HomeTeamNameES')
      .first()
      [['HomeTeamNameES', 'home_games']]
      .rename({'HomeTeamNameES': 'team'})
      .join(
          (
              data
              .group_by('VisitorTeamNameES')
              .first()
              [['VisitorTeamNameES', 'visitor_games']]
              .rename({'VisitorTeamNameES': 'team'})
          ),
          on='team',
          how='full'
      )
      .fill_null(0)
      .with_columns(
        (pl.col('home_games')+pl.col('visitor_games')).alias('games'),
        pl.when(pl.col('team').is_null())
        .then(pl.col('team_right'))
        .otherwise(pl.col('team')).alias('team')
      )
  )


  return (
      data
      .drop('home_games', 'visitor_games')
      .join(
          game_data[['team', 'games']].rename({'games': 'home_games'}),
          left_on='HomeTeamNameES',
          right_on='team'
      )
      .join(
          game_data[['team', 'games']].rename({'games': 'visitor_games'}),
          left_on='VisitorTeamNameES',
          right_on='team'
      )
  )


def compute_pitch_stats(data, player_type, pitch_class_type, min_pitches=1, pitcher_lr='both', batter_lr='both', group_by_team=False):
  assert pitcher_lr in ('both', 'l', 'r')
  assert batter_lr in ('both', 'l', 'r')
  assert player_type in ('pitcher', 'batter', 'team pitching', 'team batting') 
  assert pitch_class_type in ('general', 'specific')

  # pitching or batting, player or team
  pitching = player_type in ('pitcher', 'team pitching')
  team = player_type in ('team pitching', 'team batting')

  # handedness filters
  if pitcher_lr != 'both':
    data = data.filter(pl.col('pitLR') == pitcher_lr)
  if batter_lr != 'both':
    data = data.filter(pl.col('batLR') == batter_lr)

  if pitching:
    over_col = 'pitId' if not team else 'pitcher_team_name_short'
  else:
    over_col = 'batId' if not team else 'batter_team_name_short'
    
  # id_cols = ['pitId' if player_type == 'pitcher' else 'batId']
  # team_col = 'pitcher_team_name_short' if pitching else 'batter_team_name_short'
  # if group_by_team:
    # id_cols.append(team_col)

  # col names
  match player_type:
    case 'pitcher':
      id_cols = ['pitId']
      name_col = 'pitcher_name'
    case 'batter':
      id_cols = ['batId']
      name_col = 'batter_name'
    case _:
      id_cols = []
      name_col = None
      
  team_col = 'pitcher_team_name_short' if pitching else 'batter_team_name_short'
  if group_by_team or team:
    id_cols.append(team_col)
  
  handedness_col = 'pitLR' if pitching else 'batLR'
  new_handedness_col = 'Throws' if pitching else 'Bats'
  
  # name_col = 'pitcher_name' if player_type == 'pitcher' else 'batter_name'

  pitch_col = 'ballKind_code' if pitch_class_type == 'specific' else 'general_ballKind_code'
  pitch_name_col = 'ballKind' if pitch_class_type == 'specific' else 'general_ballKind'
  pitch_stats = (
      data
      .with_columns(
        (pl.col('ballSpeed') / 1.609).round(1).alias('mph'),
        pl.when(pl.col('general_ballKind_code').is_in(['4S', 'FC', 'SI'])).then(pl.when(valid_pitch).then('ballSpeed').mean().over(over_col, 'general_ballKind_code')).mul(1/1.609).round(1).alias('FB Velo'),
        pl.len().over(over_col).alias('Pitches')
      )
      .group_by(*id_cols, pitch_col)
      .agg(
        *([pl.col(name_col).first()] if not team else []),
        *([] if group_by_team or team else [pl.col(team_col).last()]),
        *(
          [pl.col(handedness_col).first().str.to_uppercase().alias(new_handedness_col) ]
          if not (team and ((pitcher_lr == 'both') if pitching else (batter_lr == 'both')))
          else []
        ),
        # pl.first(name_col),
        # pl.col('pitLR').first().str.to_uppercase().alias('Throws'),
        *([pl.first('general_ballKind')] if pitch_class_type == 'specific' else []),
        pl.first(pitch_name_col),
        pl.len().alias('count'),
        # pl.when(pl.col('x').is_not_null() & pl.col('y').is_not_null() & (pl.col('ballSpeed') > 0)).then('ballSpeed').mean().alias('Avg KPH'),
        # pl.col('ballSpeed').max().alias('Max KPH'),
        # pl.when(pl.col('x').is_not_null() & pl.col('y').is_not_null() & (pl.col('ballSpeed') > 0)).then('mph').mean().round(1).alias('Avg MPH'),
        # pl.col('mph').max().alias('Max MPH'),
        pl.col('aux_bresult').struct.field('batType').drop_nulls().value_counts(normalize=True),
        # swing,
        # z_swing,
        # chase,
        # contact,
        # z_con,
        # o_con,
        # whiff,
        # swstr,
        # csw,
        # strike,
        # ball,
        # f_strike,
        # par,
        # zone,
        # glove,
        # arm,
        # high,
        # low,
        # mm,
        # behind
        *[stat['expr'] for stat in STATS.values() if not stat['batted_ball']]
      )
      .with_columns(
          # (pl.col('count')/pl.sum('count').over('pitId')).alias('usage'),
          # get_stat_val('Usage', 'expr'),
          (pl.col('count') >= min_pitches).alias('qualified'),
      )
      .explode('batType')
      .unnest('batType')
      .pivot(on='batType', values='proportion')
      .fill_null(0)
      .with_columns(
          *[stat['expr'] for stat in STATS.values() if stat['batted_ball']]
          # (pl.col('G') + pl.col('B')).alias('GB%'),
          # (pl.col('F') + pl.col('P')).alias('FB%'),
          # pl.col('L').alias('LD%'),
          # pl.col('P').alias('IFFB%'),
          # pl.col('F').alias('OFFB%'),
          # (pl.col('F') + pl.col('P') + pl.col('L')).alias('AIR%')
          
      )
      .drop('G', 'F', 'B', 'P', 'L', 'null')
      .with_columns(
          # (pl.when(pl.col('qualified')).then(pl.col(stat)).rank(descending=((stat in ['FB%', 'LD%', 'OFFB%', 'AIR%', 'Ball%', 'Behind%'] or 'Contact%' in stat)))/pl.when(pl.col('qualified')).then(pl.col(stat)).count()).alias(f'{stat}_pctl')
          # for stat in ['Avg KPH', 'Max KPH', 'Avg MPH', 'Max MPH', 'Swing%', 'Z-Swing%', 'Chase%', 'Contact%', 'Z-Contact%', 'O-Contact%', 'SwStr%', 'Whiff%', 'CSW%', 'Strike%', 'Ball%', 'F-Str%', 'PAR%', 'GB%', 'FB%', 'LD%', 'OFFB%', 'IFFB%', 'AIR%', 'Zone%', 'Behind%']
          (pl.when(pl.col('qualified')).then(pl.col(stat)).rank(descending=(get_stat_val(stat, 'percentile')) not in  (Player.PITCHER if pitching else Player.BATTER, Player.BOTH))
          /
          pl.when(pl.col('qualified')).then(pl.col(stat)).count()).alias(f'{stat}_pctl')
          for stat in STATS.keys()
      )
      .rename({pitch_col: 'ballKind_code', pitch_name_col: 'ballKind'} if pitch_class_type == 'general' else {})
      .sort(id_cols[0], 'count', descending=[False, True])
  )
  return pitch_stats

def compute_player_stats(data, player_type, qual='qualified', pitcher_lr='both', batter_lr='both', group_by_team=False):
  # TO-DO: figure out if I still need group_by_team
  assert pitcher_lr in ('both', 'l', 'r')
  assert batter_lr in ('both', 'l', 'r')
  assert player_type in ('pitcher', 'batter', 'team pitching', 'team batting') 

  # pitching or batting, player or team
  pitching = player_type in ('pitcher', 'team pitching')
  team = player_type in ('team pitching', 'team batting')

  # handedness filters
  if pitcher_lr != 'both':
    data = data.filter(pl.col('pitLR') == pitcher_lr)
  if batter_lr != 'both':
    data = data.filter(pl.col('batLR') == batter_lr)

  if pitching:
    over_col = 'pitId' if not team else 'pitcher_team_name_short'
  else:
    over_col = 'batId' if not team else 'batter_team_name_short'
  data = (
      compute_team_games(data)
      .with_columns(
          pl.when(pl.col('half_inning').str.ends_with('1')).then('home_games').otherwise('visitor_games').first().over('pitId').alias('games'),
          # pl.col('inning_code').unique().len().over(over_col).alias('IP'),
          (pl.col('bso').struct.field('o').cast(pl.Int32) - pl.col('beforeBso').struct.field('o').cast(pl.Int32)).sum().mul(1/3).over(over_col).alias('IP'),
          pl.col('pa_code').unique().len().over(over_col).alias('PA'),
          pl.col('presult').is_in(verify_and_return_presult([
            'Single', 'Double', 'Triple', 'Home run', 'Inside-the-park home run',
            'Groundout', 'Flyout', 'Lineout', 'Groundout (Double play)',
            'Foul fly', 'Foul line (?)',
            'Error', 'Sacrifice hit error', 'Sacrifice fly error',
            "Fielder's choice",
            'Bunt strikeout', 'Swinging strikeout', 'Looking strikeout'
          ])).sum().over(over_col).alias('AB'),
          pl.len().over(over_col).alias('Pitches')
          # pl.col('presult').is_in(verify_and_return_presult([
            # 'Groundout', 'Flyout', 'Lineout', 'Groundout (Double play)',
            # 'Foul fly', 'Foul line (?)',
            # 'Sacrifice bunt', 'Sacrifice fly',
            # "Fielder's choice", "Sacrifice fielder's choice",
            # 'Bunt strikeout', 'Swinging strikeout', 'Looking strikeout'
          # ])).sum().over('pitId').mul(1/3).alias('IP')
      )
  )

  # qualifiers
  qualified_factor = 1 if pitching else 3.1
  qual_col = 'IP' if pitching else 'PA'
  if qual == 'qualified':
    data = data.with_columns((pl.col(qual_col) >= qualified_factor * pl.col('games')).alias('qualified'))
  else:
    data = data.with_columns((pl.col(qual_col) >= qual).alias('qualified'))

  # percentile ascending/descending
  # if pitching:
    # stat_descending_pctl = lambda stat: stat in ['BB%', 'Ball%', 'FB%', 'LD%', 'OFFB%', 'AIR%', 'Z-Swing%', 'Behind%', 'OBP'] or 'Contact%' in stat
  # else:
    # stat_descending_pctl = lambda stat: not (stat in ['BB%', 'Ball%', 'FB%', 'LD%', 'OFFB%', 'AIR%', 'Swing%', 'Z-Swing%', 'Behind%', 'OBP'] or 'Contact%' in stat)

  # col names
  match player_type:
    case 'pitcher':
      id_cols = ['pitId']
      name_col = 'pitcher_name'
    case 'batter':
      id_cols = ['batId']
      name_col = 'batter_name'
    case _:
      id_cols = []
      name_col = None
      
  team_col = 'pitcher_team_name_short' if pitching else 'batter_team_name_short'
  if group_by_team or team:
    id_cols.append(team_col)
  
  handedness_col = 'pitLR' if pitching else 'batLR'
  new_handedness_col = 'Throws' if pitching else 'Bats'
  player_stats = (
    data
    .with_columns(
      (pl.col('ballSpeed') / 1.609).round(1).alias('mph'),
      pl.when(pl.col('general_ballKind_code').is_in(['4S', 'FC', 'SI'])).then(pl.when(valid_pitch).then('ballSpeed').mean().over(over_col, 'general_ballKind_code')).mul(1/1.609).round(1).alias('FB Velo')
      )
    .group_by(id_cols)
    .agg(
        *([pl.col(name_col).first()] if not team else []),
        *([] if group_by_team or team else [pl.col(team_col).last()]),
        *(
          [pl.col(handedness_col).first().str.to_uppercase().alias(new_handedness_col) ]
          if not (team and ((pitcher_lr == 'both') if pitching else (batter_lr == 'both')))
          else []
        ),
        pl.col('IP').first(),
        pl.col('PA').first(),
        # pl.col('FB Velo').max(),
        # (pl.when(pl.col('presult').str.contains('strikeout')).then(1).otherwise(0).sum() / pl.col('pa_code').unique().len()).alias('K%'),
        # (pl.when(pl.col('presult') == 'Walk').then(1).otherwise(0).sum() / pl.col('pa_code').unique().len()).alias('BB%'),
        pl.col('aux_bresult').struct.field('batType').drop_nulls().value_counts(normalize=True),
        # swing,
        # z_swing,
        # chase,
        # contact,
        # z_con,
        # o_con,
        # whiff,
        # swstr,
        # csw,
        # strike,
        # ball,
        # f_strike,
        # par,
        # zone,
        # glove,
        # arm,
        # high,
        # low,
        # mm,
        # behind,
        # pl.col('AB').first(),
        # h,
        # bb,
        # hbp,
        # sf,
        # obp,
        # pl.first('qualified')
        pl.first('qualified'),
        *[stat['expr'] for stat in STATS.values() if not stat['batted_ball']]
    )
    .explode('batType')
    .unnest('batType')
    .pivot(on='batType', values='proportion')
    .fill_null(0)
    .with_columns(
        *[stat['expr'] for stat in STATS.values() if stat['batted_ball']]
        # (pl.col('G') + pl.col('B')).alias('GB%'),
        # (pl.col('F') + pl.col('P')).alias('FB%'),
        # pl.col('L').alias('LD%'),
        # pl.col('P').alias('IFFB%'),
        # pl.col('F').alias('OFFB%'),
        # (pl.col('F') + pl.col('P') + pl.col('L')).alias('AIR%')
    )
    .drop('G', 'F', 'B', 'P', 'L')
    .with_columns(
        # (pl.when(pl.col('qualified')).then(pl.col(stat)).rank(descending=stat_descending_pctl(stat))/pl.when(pl.col('qualified')).then(pl.col(stat)).count()).alias(f'{stat}_pctl')
        # for stat in ['FB Velo', 'K%', 'BB%', 'Swing%', 'Z-Swing%', 'Chase%', 'Contact%', 'Z-Contact%', 'O-Contact%', 'SwStr%', 'Whiff%', 'CSW%', 'Strike%', 'Ball%', 'F-Str%', 'PAR%', 'GB%', 'FB%', 'LD%', 'OFFB%', 'IFFB%', 'AIR%', 'Zone%', 'Behind%', 'OBP']
        (pl.when(pl.col('qualified')).then(pl.col(stat)).rank(descending=(get_stat_val(stat, 'percentile')) not in (Player.PITCHER if pitching else Player.BATTER, Player.BOTH))
        /
        pl.when(pl.col('qualified')).then(pl.col(stat)).count()).alias(f'{stat}_pctl')
        for stat in STATS.keys()
    )
    .sort(qual_col, descending=True)
  )
  return player_stats


def get_pitcher_stats(id, lr='both', game_kind=None, start_date=None, end_date=None, min_ip=1, min_pitches=1, pitch_class_type='specific'):
  
  source_data = data_df
  source_data = filter_data_by_date_and_game_kind(source_data, start_date=start_date, end_date=end_date, game_kind=game_kind)

  # if lr is not None:
    # source_data =

  pitch_stats = compute_pitch_stats(source_data, player_type='pitcher', pitch_class_type=pitch_class_type, min_pitches=min_pitches, batter_lr=lr, group_by_team=False).filter(pl.col('pitId') == id)

  pitch_shapes = (
      (source_data.filter(pl.col('batLR') == lr) if lr != 'both' else source_data)
      .filter(
          (pl.col('pitId') == id) &
          pl.col('x').is_not_null() &
          pl.col('y').is_not_null() &
          (pl.col('ballSpeed') > 0)
      )
      [['pitId', 'general_ballKind_code', 'ballKind_code', 'ballSpeed', 'x', 'y']]
      .with_columns((pl.col('ballSpeed')/1.609).alias('ballSpeed_mph'))
  )

  pitcher_stats = compute_player_stats(source_data, player_type='pitcher', qual=min_ip, batter_lr=lr, group_by_team=False).filter(pl.col('pitId') == id)
  return SimpleNamespace(pitcher_stats=pitcher_stats, pitch_stats=pitch_stats, pitch_shapes=pitch_shapes)