pitcher_update.py

import pandas as pd
import numpy as np
import joblib
import math
import pickle

loaded_model = joblib.load('joblib_model/barrel_model.joblib')
in_zone_model = joblib.load('joblib_model/in_zone_model_knn_20240410.joblib')
attack_zone_model = joblib.load('joblib_model/model_attack_zone.joblib')
xwoba_model = joblib.load('joblib_model/xwoba_model.joblib')
px_model = joblib.load('joblib_model/linear_reg_model_x.joblib')
pz_model = joblib.load('joblib_model/linear_reg_model_z.joblib')


def percentile(n):
    def percentile_(x):
        return np.nanpercentile(x, n)
    percentile_.__name__ = 'percentile_%s' % n
    return percentile_


def df_update(df=pd.DataFrame()):
    df.loc[df['sz_top']==0,'sz_top'] = np.nan
    df.loc[df['sz_bot']==0,'sz_bot'] = np.nan


    df['in_zone'] = [x < 10 if x > 0 else np.nan for x in df['zone']]
    if len(df.loc[(~df['x'].isnull())&(df['px'].isnull()),'px']) > 0:
        df.loc[(~df['x'].isnull())&(df['px'].isnull()),'px'] = px_model.predict(df.loc[(~df['x'].isnull())&(df['px'].isnull())][['x']])
        df.loc[(~df['y'].isnull())&(df['pz'].isnull()),'pz'] = px_model.predict(df.loc[(~df['y'].isnull())&(df['pz'].isnull())][['y']]) + 3.2

    
        # df['in_zone'] =  [x < 10 if x > 0 else np.nan for x in df['zone']]
    if len(df.loc[(~df['px'].isna())&
                  (df['in_zone'].isna())&
                  (~df['sz_top'].isna())]) > 0:
            print('We found missing data')
            df.loc[(~df['px'].isna())&
                            (df['in_zone'].isna())&
                            (~df['sz_top'].isna())&
                            (~df['pz'].isna())&
                            (~df['sz_bot'].isna())
                            ,'in_zone'] = in_zone_model.predict(df.loc[(~df['px'].isna())&
                                                                (df['in_zone'].isna())&
                                                                (~df['sz_top'].isna())&
                                                                (~df['pz'].isna())&
                                                                (~df['sz_bot'].isna())][['px','pz','sz_top','sz_bot']].values)
    hit_codes = ['single',
        'double','home_run', 'triple']

    ab_codes = ['single', 'strikeout', 'field_out',
        'grounded_into_double_play', 'fielders_choice', 'force_out',
        'double', 'field_error', 'home_run', 'triple',
        'double_play',
        'fielders_choice_out', 'strikeout_double_play',
        'other_out','triple_play']


    obp_true_codes = ['single', 'walk',
        'double','home_run', 'triple',
        'hit_by_pitch', 'intent_walk']

    obp_codes = ['single', 'strikeout', 'walk', 'field_out',
        'grounded_into_double_play', 'fielders_choice', 'force_out',
        'double', 'sac_fly', 'field_error', 'home_run', 'triple',
        'hit_by_pitch', 'double_play', 'intent_walk',
        'fielders_choice_out', 'strikeout_double_play',
        'sac_fly_double_play',
        'other_out','triple_play']


    contact_codes = ['In play, no out',
            'Foul', 'In play, out(s)',
        'In play, run(s)',
        'Foul Bunt']



    conditions_hit = [df.event_type.isin(hit_codes)]
    choices_hit = [True]
    df['hits'] = np.select(conditions_hit, choices_hit, default=False)

    conditions_ab = [df.event_type.isin(ab_codes)]
    choices_ab = [True]
    df['ab'] = np.select(conditions_ab, choices_ab, default=False)

    conditions_obp_true = [df.event_type.isin(obp_true_codes)]
    choices_obp_true = [True]
    df['on_base'] = np.select(conditions_obp_true, choices_obp_true, default=False)

    conditions_obp = [df.event_type.isin(obp_codes)]
    choices_obp = [True]
    df['obp'] = np.select(conditions_obp, choices_obp, default=False)

    bip_codes = ['In play, no out', 'In play, run(s)','In play, out(s)']
    
    conditions_bip = [df.play_description.isin(bip_codes)]
    choices_bip = [True]
    df['bip'] = np.select(conditions_bip, choices_bip, default=False)

    conditions = [
        (df['launch_speed'].isna()),
        (df['launch_speed']*1.5 - df['launch_angle'] >= 117 ) & (df['launch_speed'] + df['launch_angle'] >= 124) & (df['launch_speed'] > 98) & (df['launch_angle'] >= 8) & (df['launch_angle'] <= 50)
    ]
    df['bip_div'] = ~df.launch_speed.isna()
    choices = [False,True]
    df['barrel'] = np.select(conditions, choices, default=np.nan)
    df['barrel'] = loaded_model.predict(df[['launch_speed','launch_angle']].fillna(0).values)
    conditions_ss = [
        (df['launch_angle'].isna()),
        (df['launch_angle'] >= 8 ) * (df['launch_angle'] <= 32 )
    ]

    choices_ss = [False,True]
    df['sweet_spot'] = np.select(conditions_ss, choices_ss, default=np.nan)

    conditions_hh = [
        (df['launch_speed'].isna()),
        (df['launch_speed'] >= 94.5 )
    ]

    choices_hh = [False,True]
    df['hard_hit'] = np.select(conditions_hh, choices_hh, default=np.nan)


    conditions_tb = [
        (df['event_type']=='single'),
        (df['event_type']=='double'),
        (df['event_type']=='triple'),
        (df['event_type']=='home_run'),
    ]

    choices_tb = [1,2,3,4]

    df['tb'] = np.select(conditions_tb, choices_tb, default=np.nan)

    conditions_woba = [
        (df['event_type'].isin(['strikeout', 'field_out', 'sac_fly', 'force_out',
                                'grounded_into_double_play', 'fielders_choice', 'field_error',
                                'sac_bunt', 'double_play', 'fielders_choice_out', 'strikeout_double_play',
                                 'sac_fly_double_play', 'other_out'])),
        (df['event_type']=='walk'),
        (df['event_type']=='hit_by_pitch'),
        (df['event_type']=='single'),
        (df['event_type']=='double'),
        (df['event_type']=='triple'),
        (df['event_type']=='home_run'),
    ]

    choices_woba =    [0,
                    0.696,
                    0.726,
                    0.883,
                    1.244,
                    1.569,
                    2.004]

    df['woba'] = np.select(conditions_woba, choices_woba, default=np.nan)


    woba_codes = ['strikeout', 'field_out', 'single', 'walk', 'hit_by_pitch',
        'double', 'sac_fly', 'force_out', 'home_run',
        'grounded_into_double_play', 'fielders_choice', 'field_error',
        'triple', 'sac_bunt', 'double_play',
        'fielders_choice_out', 'strikeout_double_play',
        'sac_fly_double_play', 'other_out']


    
    
    

    conditions_woba_code = [
        (df['event_type'].isin(woba_codes))
    ]

    choices_woba_code = [1]

    df['woba_codes'] = np.select(conditions_woba_code, choices_woba_code, default=np.nan)


    df['woba_contact'] = [df['woba'].values[x] if df['bip'].values[x] == 1 else np.nan for x in range(len(df['woba_codes']))]

    #df['in_zone'] =  [x < 10 if type(x) == int else np.nan for x in df['zone']]

    # df['in_zone_2'] = in_zone_model.predict(df[['x','y','sz_bot','sz_top']].fillna(0).values)
    # df['in_zone_3'] = df['in_zone_2'] < 10
    # df.loc[df['in_zone'].isna(),'in_zone'] = df.loc[df['in_zone'].isna(),'in_zone_3'].fillna(0)


    df['whiffs'] = [1 if ((x == 'S')|(x == 'W')|(x =='T')) else 0 for x in df.play_code]
    df['csw'] = [1 if ((x == 'S')|(x == 'W')|(x =='T')|(x == 'C')) else 0 for x in df.play_code]
    df['swings'] = [1 if x == True else 0 for x in df.is_swing]


    df['out_zone'] = df.in_zone == False
    df['zone_swing'] = (df.in_zone == True)&(df.swings == 1)
    df['zone_contact'] = (df.in_zone == True)&(df.swings == 1)&(df.whiffs == 0)
    df['ozone_swing'] = (df.in_zone==False)&(df.swings == 1)
    df['ozone_contact'] = (df.in_zone==False)&(df.swings == 1)&(df.whiffs == 0)

    df['k'] = df.event_type.isin(list(filter(None, [x if 'strikeout' in x else '' for x in df.event_type.dropna().unique()])))
    df['bb'] = df.event_type.isin(['walk','intent_walk'])

    df['k_minus_bb'] = df['k'].astype(np.float32)-df['bb'].astype(np.float32)
    df['bb_minus_k'] = df['bb'].astype(np.float32)-df['k'].astype(np.float32)

    df['pa']  = [1 if isinstance(x, str) else 0 for x in df.event_type]
    df['pitches']  = [1 if x else 0 for x in df.is_pitch]


    df.loc[df['launch_speed'].isna(),'barrel'] = np.nan


    pitch_cat = {'FA':'Fastball',
    'FF':'Fastball',
    'FT':'Fastball',
    'FC':'Fastball',
    'FS':'Off-Speed',
    'FO':'Off-Speed',
    'SI':'Fastball',
    'ST':'Breaking',
    'SL':'Breaking',
    'CU':'Breaking',
    'KC':'Breaking',
    'SC':'Off-Speed',
    'GY':'Off-Speed',
    'SV':'Breaking',
    'CS':'Breaking',
    'CH':'Off-Speed',
    'KN':'Off-Speed',
    'EP':'Breaking',
    'UN':np.nan,
    'IN':np.nan,
    'PO':np.nan,
    'AB':np.nan,
    'AS':np.nan,
    'NP':np.nan}
    #df['pitch_type'] = df['pitch_type'].map(pitch_cat).fillna('Unknown')
    df['average'] = 'average'

    df.loc[df['trajectory'] == 'bunt_popup','trajectory'] = 'popup'
    df.loc[df['trajectory'] == 'bunt_grounder','trajectory'] = 'ground_ball'
    df.loc[df['trajectory'] == '','trajectory'] = np.nan
    df.loc[df['trajectory'] == 'bunt_line_drive','trajectory'] = 'line_drive'
    df[['trajectory_fly_ball','trajectory_ground_ball','trajectory_line_drive','trajectory_popup']] = pd.get_dummies(df['trajectory'], prefix='trajectory')

    df['attack_zone'] = np.nan



    df.loc[df[['px','pz','sz_top','sz_bot']].isnull().sum(axis=1)==0,'attack_zone'] = attack_zone_model.predict(df.loc[df[['px','pz','sz_top','sz_bot']].isnull().sum(axis=1)==0][['px','pz','sz_top','sz_bot']])



    df['heart'] = df['attack_zone'] == 0
    df['shadow'] = df['attack_zone'] == 1
    df['chase'] = df['attack_zone'] == 2
    df['waste'] = df['attack_zone'] == 3

    df['heart_swing'] = (df['attack_zone'] == 0)&(df['swings']==1)
    df['shadow_swing'] = (df['attack_zone'] == 1)&(df['swings']==1)
    df['chase_swing'] = (df['attack_zone'] == 2)&(df['swings']==1)
    df['waste_swing'] = (df['attack_zone'] == 3)&(df['swings']==1)

    df['heart_whiff'] = (df['attack_zone'] == 0)&(df['whiffs']==1)
    df['shadow_whiff'] = (df['attack_zone'] == 1)&(df['whiffs']==1)
    df['chase_whiff'] = (df['attack_zone'] == 2)&(df['whiffs']==1)
    df['waste_whiff'] = (df['attack_zone'] == 3)&(df['whiffs']==1)  

    df['woba_pred'] = np.nan
    df['woba_pred_contact'] = np.nan

    if len(df.loc[df[['launch_angle','launch_speed']].isnull().sum(axis=1)==0,'woba_pred']) > 0:
        

        df.loc[df[['launch_angle','launch_speed']].isnull().sum(axis=1)==0,'woba_pred']  = [sum(x) for x in xwoba_model.predict_proba(df.loc[df[['launch_angle','launch_speed']].isnull().sum(axis=1)==0][['launch_angle','launch_speed']]) * ([0, 0.883,1.244,1.569,2.004])]

        ## Assign a value of 0.696 to every walk in the dataset
        df.loc[df['event_type'].isin(['walk']),'woba_pred'] = 0.696

        ## Assign a value of 0.726 to every hit by pitch in the dataset
        df.loc[df['event_type'].isin(['hit_by_pitch']),'woba_pred'] = 0.726

        ## Assign a value of 0 to every Strikeout in the dataset 
        df.loc[df['event_type'].isin(['strikeout','strikeout_double_play']),'woba_pred'] = 0

        
        df.loc[df[['launch_angle','launch_speed']].isnull().sum(axis=1)==0,'woba_pred_contact']  = [sum(x) for x in xwoba_model.predict_proba(df.loc[df[['launch_angle','launch_speed']].isnull().sum(axis=1)==0][['launch_angle','launch_speed']]) * ([0, 0.883,1.244,1.569,2.004])]
    
    df['xwoba_codes'] = np.nan
    df.loc[df[['launch_angle','launch_speed']].isnull().sum(axis=1)==0,'xwoba_codes']  = 1
    ## Assign a value of 0.696 to every walk in the dataset
    df.loc[df['event_type'].isin(['walk']),'xwoba_codes'] = 1

    ## Assign a value of 0.726 to every hit by pitch in the dataset
    df.loc[df['event_type'].isin(['hit_by_pitch']),'xwoba_codes'] = 1

    ## Assign a value of 0 to every Strikeout in the dataset 
    df.loc[df['event_type'].isin(['strikeout','strikeout_double_play']),'xwoba_codes'] = 1
    return df

def df_update_summ(df=pd.DataFrame()):
    df_summ = df.groupby(['pitcher_id','pitcher_name']).agg(
        pa = ('pa','sum'),
        ab = ('ab','sum'),
        obp_pa = ('obp','sum'),
        hits = ('hits','sum'),
        on_base = ('on_base','sum'),
        k = ('k','sum'),
        bb = ('bb','sum'),
        bb_minus_k = ('bb_minus_k','sum'),
        csw = ('csw','sum'),
        bip = ('bip','sum'),
        bip_div = ('bip_div','sum'),
        tb = ('tb','sum'),
        woba = ('woba','sum'),
        woba_contact = ('woba_contact','sum'),
        xwoba = ('woba_pred','sum'),
        xwoba_contact = ('woba_pred_contact','sum'),       
        woba_codes = ('woba_codes','sum'),
        xwoba_codes = ('xwoba_codes','sum'),
        hard_hit = ('hard_hit','sum'),
        barrel = ('barrel','sum'),
        sweet_spot = ('sweet_spot','sum'),
        max_launch_speed = ('launch_speed','max'),
        launch_speed_90 =  ('launch_speed',percentile(90)),
        launch_speed = ('launch_speed','mean'),
        launch_angle = ('launch_angle','mean'),
        pitches = ('is_pitch','sum'),
        swings =  ('swings','sum'),
        in_zone = ('in_zone','sum'),
        out_zone = ('out_zone','sum'),
        whiffs = ('whiffs','sum'),
        zone_swing = ('zone_swing','sum'),
        zone_contact = ('zone_contact','sum'),
        ozone_swing = ('ozone_swing','sum'),
        ozone_contact = ('ozone_contact','sum'),
        ground_ball = ('trajectory_ground_ball','sum'),
        line_drive = ('trajectory_line_drive','sum'),
        fly_ball =('trajectory_fly_ball','sum'),
        pop_up = ('trajectory_popup','sum'),
        attack_zone = ('attack_zone','count'),
        heart = ('heart','sum'),
        shadow = ('shadow','sum'),
        chase = ('chase','sum'),
        waste = ('waste','sum'),
        heart_swing = ('heart_swing','sum'),
        shadow_swing = ('shadow_swing','sum'),
        chase_swing = ('chase_swing','sum'),
        waste_swing = ('waste_swing','sum'),
        ).reset_index()
    return df_summ

def df_update_summ_avg(df=pd.DataFrame()):
    df_summ_avg = df.groupby(['average']).agg(
       
        ).reset_index()
    return df_summ_avg

def df_summ_changes(df_summ=pd.DataFrame()):
    df_summ['avg'] = [df_summ.hits[x]/df_summ.ab[x] if df_summ.ab[x] != 0 else np.nan for x in range(len(df_summ))]
    df_summ['obp'] = [df_summ.on_base[x]/df_summ.obp_pa[x] if df_summ.obp_pa[x] != 0 else np.nan for x in range(len(df_summ))]
    df_summ['slg'] =  [df_summ.tb[x]/df_summ.ab[x] if df_summ.ab[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['ops'] = df_summ['obp']+df_summ['slg']

    df_summ['k_percent'] = [df_summ.k[x]/df_summ.pa[x] if df_summ.pa[x] != 0 else np.nan for x in range(len(df_summ))]
    df_summ['bb_percent'] =[df_summ.bb[x]/df_summ.pa[x] if df_summ.pa[x] != 0 else np.nan for x in range(len(df_summ))]
    df_summ['bb_minus_k_percent'] =[(df_summ.bb_minus_k[x])/df_summ.pa[x] if df_summ.pa[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['bb_over_k_percent'] =[df_summ.bb[x]/df_summ.k[x] if df_summ.k[x] != 0 else np.nan for x in range(len(df_summ))]




    df_summ['csw_percent'] =[df_summ.csw[x]/df_summ.pitches[x] if df_summ.pitches[x] != 0 else np.nan for x in range(len(df_summ))]


    df_summ['sweet_spot_percent'] = [df_summ.sweet_spot[x]/df_summ.bip_div[x] if df_summ.bip_div[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['woba_percent'] = [df_summ.woba[x]/df_summ.woba_codes[x] if df_summ.woba_codes[x] != 0 else np.nan for x in range(len(df_summ))]
    df_summ['woba_percent_contact'] = [df_summ.woba_contact[x]/df_summ.bip[x] if df_summ.bip[x] != 0 else np.nan for x in range(len(df_summ))]
    #df_summ['hard_hit_percent'] = [df_summ.sweet_spot[x]/df_summ.bip[x] if df_summ.bip[x] != 0 else np.nan for x in range(len(df_summ))]
    df_summ['hard_hit_percent'] = [df_summ.hard_hit[x]/df_summ.bip_div[x] if df_summ.bip_div[x] != 0 else np.nan for x in range(len(df_summ))]


    df_summ['barrel_percent'] = [df_summ.barrel[x]/df_summ.bip_div[x] if df_summ.bip_div[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['zone_contact_percent'] = [df_summ.zone_contact[x]/df_summ.zone_swing[x] if df_summ.zone_swing[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['zone_swing_percent'] = [df_summ.zone_swing[x]/df_summ.in_zone[x] if df_summ.in_zone[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['zone_percent'] = [df_summ.in_zone[x]/df_summ.pitches[x] if df_summ.pitches[x] > 0 else np.nan for x in range(len(df_summ))]

    df_summ['chase_percent'] = [df_summ.ozone_swing[x]/(df_summ.pitches[x] - df_summ.in_zone[x]) if (df_summ.pitches[x]- df_summ.in_zone[x]) != 0 else np.nan for x in range(len(df_summ))]

    df_summ['chase_contact'] = [df_summ.ozone_contact[x]/df_summ.ozone_swing[x] if df_summ.ozone_swing[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['swing_percent'] = [df_summ.swings[x]/df_summ.pitches[x] if df_summ.pitches[x] > 0 else np.nan for x in range(len(df_summ))]

    df_summ['whiff_rate'] = [df_summ.whiffs[x]/df_summ.swings[x] if df_summ.swings[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['swstr_rate'] = [df_summ.whiffs[x]/df_summ.pitches[x] if df_summ.pitches[x] > 0 else np.nan for x in range(len(df_summ))]

    df_summ['ground_ball_percent'] = [df_summ.ground_ball[x]/df_summ.bip[x] if df_summ.bip[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['line_drive_percent'] = [df_summ.line_drive[x]/df_summ.bip[x] if df_summ.bip[x] != 0 else np.nan for x in range(len(df_summ))]    

    df_summ['fly_ball_percent'] = [df_summ.fly_ball[x]/df_summ.bip[x] if df_summ.bip[x] != 0 else np.nan for x in range(len(df_summ))]    

    df_summ['pop_up_percent'] = [df_summ.pop_up[x]/df_summ.bip[x] if df_summ.bip[x] != 0 else np.nan for x in range(len(df_summ))]  


    
    df_summ['heart_zone_percent'] = [df_summ.heart[x]/df_summ.attack_zone[x] if df_summ.attack_zone[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['shadow_zone_percent'] = [df_summ.shadow[x]/df_summ.attack_zone[x] if df_summ.attack_zone[x] != 0 else np.nan for x in range(len(df_summ))]    

    df_summ['chase_zone_percent'] = [df_summ.chase[x]/df_summ.attack_zone[x] if df_summ.attack_zone[x] != 0 else np.nan for x in range(len(df_summ))]    

    df_summ['waste_zone_percent'] = [df_summ.waste[x]/df_summ.attack_zone[x] if df_summ.attack_zone[x] != 0 else np.nan for x in range(len(df_summ))]    
    
    
    df_summ['heart_zone_swing_percent'] = [df_summ.heart_swing[x]/df_summ.heart[x] if df_summ.heart[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ['shadow_zone_swing_percent'] = [df_summ.shadow_swing[x]/df_summ.shadow[x] if df_summ.shadow[x] != 0 else np.nan for x in range(len(df_summ))]    

    df_summ['chase_zone_swing_percent'] = [df_summ.chase_swing[x]/df_summ.chase[x] if df_summ.chase[x] != 0 else np.nan for x in range(len(df_summ))]    

    df_summ['waste_zone_swing_percent'] = [df_summ.waste_swing[x]/df_summ.waste[x] if df_summ.waste[x] != 0 else np.nan for x in range(len(df_summ))]    
    



    df_summ['xwoba_percent'] = [df_summ.xwoba[x]/df_summ.xwoba_codes[x] if df_summ.xwoba_codes[x] != 0 else np.nan for x in range(len(df_summ))]
    df_summ['xwoba_percent_contact'] = [df_summ.xwoba_contact[x]/df_summ.bip[x] if df_summ.bip[x] != 0 else np.nan for x in range(len(df_summ))]

    df_summ = df_summ.dropna(subset=['bip'])
    return df_summ

def df_summ_filter_out(df_summ=pd.DataFrame(),batter_select = 0):
    df_summ_filter = df_summ[df_summ['pa'] >= min(math.floor(df_summ.xs(batter_select,level=0)['pa']/10)*10,500)]
    df_summ_filter_pct = df_summ_filter.rank(pct=True,ascending=True)
    df_summ_player = df_summ.xs(batter_select,level=0)
    df_summ_player_pct = df_summ_filter_pct.xs(batter_select,level=0)
    return df_summ_filter,df_summ_filter_pct,df_summ_player,df_summ_player_pct

def df_summ_batter_pitch_up(df=pd.DataFrame()):
    df_summ_batter_pitch = df.dropna(subset=['pitch_type']).groupby(['pitcher_id','pitcher_name','pitch_type']).agg(
        pa = ('pa','sum'),
        ab = ('ab','sum'),
        obp_pa = ('obp','sum'),
        hits = ('hits','sum'),
        on_base = ('on_base','sum'),
        k = ('k','sum'),
        bb = ('bb','sum'),
        bb_minus_k = ('bb_minus_k','sum'),
        csw = ('csw','sum'),
        bip = ('bip','sum'),
        bip_div = ('bip_div','sum'),
        tb = ('tb','sum'),
        woba = ('woba','sum'),
        woba_contact = ('woba_pred_contact','sum'),
        xwoba = ('woba_pred','sum'),
        xwoba_contact = ('woba_pred','sum'),       
        woba_codes = ('woba_codes','sum'),
        xwoba_codes = ('xwoba_codes','sum'),
        hard_hit = ('hard_hit','sum'),
        barrel = ('barrel','sum'),
        sweet_spot = ('sweet_spot','sum'),
        max_launch_speed = ('launch_speed','max'),
        launch_speed_90 =  ('launch_speed',percentile(90)),
        launch_speed = ('launch_speed','mean'),
        launch_angle = ('launch_angle','mean'),
        pitches = ('is_pitch','sum'),
        swings =  ('swings','sum'),
        in_zone = ('in_zone','sum'),
        out_zone = ('out_zone','sum'),
        whiffs = ('whiffs','sum'),
        zone_swing = ('zone_swing','sum'),
        zone_contact = ('zone_contact','sum'),
        ozone_swing = ('ozone_swing','sum'),
        ozone_contact = ('ozone_contact','sum'),
        ground_ball = ('trajectory_ground_ball','sum'),
        line_drive = ('trajectory_line_drive','sum'),
        fly_ball =('trajectory_fly_ball','sum'),
        pop_up = ('trajectory_popup','sum'),
        attack_zone = ('attack_zone','count'),
        heart = ('heart','sum'),
        shadow = ('shadow','sum'),
        chase = ('chase','sum'),
        waste = ('waste','sum'),
        heart_swing = ('heart_swing','sum'),
        shadow_swing = ('shadow_swing','sum'),
        chase_swing = ('chase_swing','sum'),
        waste_swing = ('waste_swing','sum'),
        ).reset_index()

    #return df_summ_batter_pitch
    df_summ_batter_pitch['avg'] = [df_summ_batter_pitch.hits[x]/df_summ_batter_pitch.ab[x] if df_summ_batter_pitch.ab[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]
    df_summ_batter_pitch['obp'] = [df_summ_batter_pitch.on_base[x]/df_summ_batter_pitch.obp_pa[x] if df_summ_batter_pitch.obp_pa[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]
    df_summ_batter_pitch['slg'] =  [df_summ_batter_pitch.tb[x]/df_summ_batter_pitch.ab[x] if df_summ_batter_pitch.ab[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['ops'] = df_summ_batter_pitch['obp']+df_summ_batter_pitch['slg']

    df_summ_batter_pitch['k_percent'] = [df_summ_batter_pitch.k[x]/df_summ_batter_pitch.pa[x] if df_summ_batter_pitch.pa[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]
    df_summ_batter_pitch['bb_percent'] =[df_summ_batter_pitch.bb[x]/df_summ_batter_pitch.pa[x] if df_summ_batter_pitch.pa[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]
    df_summ_batter_pitch['bb_minus_k_percent'] =[(df_summ_batter_pitch.bb_minus_k[x])/df_summ_batter_pitch.pa[x] if df_summ_batter_pitch.pa[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['bb_over_k_percent'] =[df_summ_batter_pitch.bb[x]/df_summ_batter_pitch.k[x] if df_summ_batter_pitch.k[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]




    df_summ_batter_pitch['csw_percent'] =[df_summ_batter_pitch.csw[x]/df_summ_batter_pitch.pitches[x] if df_summ_batter_pitch.pitches[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]


    df_summ_batter_pitch['sweet_spot_percent'] = [df_summ_batter_pitch.sweet_spot[x]/df_summ_batter_pitch.bip_div[x] if df_summ_batter_pitch.bip_div[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['woba_percent'] = [df_summ_batter_pitch.woba[x]/df_summ_batter_pitch.woba_codes[x] if df_summ_batter_pitch.woba_codes[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]
    df_summ_batter_pitch['woba_percent_contact'] = [df_summ_batter_pitch.woba_contact[x]/df_summ_batter_pitch.bip[x] if df_summ_batter_pitch.bip[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]
    #df_summ_batter_pitch['hard_hit_percent'] = [df_summ_batter_pitch.sweet_spot[x]/df_summ_batter_pitch.bip[x] if df_summ_batter_pitch.bip[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]
    df_summ_batter_pitch['hard_hit_percent'] = [df_summ_batter_pitch.hard_hit[x]/df_summ_batter_pitch.bip_div[x] if df_summ_batter_pitch.bip_div[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]


    df_summ_batter_pitch['barrel_percent'] = [df_summ_batter_pitch.barrel[x]/df_summ_batter_pitch.bip_div[x] if df_summ_batter_pitch.bip_div[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['zone_contact_percent'] = [df_summ_batter_pitch.zone_contact[x]/df_summ_batter_pitch.zone_swing[x] if df_summ_batter_pitch.zone_swing[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['zone_swing_percent'] = [df_summ_batter_pitch.zone_swing[x]/df_summ_batter_pitch.in_zone[x] if df_summ_batter_pitch.in_zone[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['zone_percent'] = [df_summ_batter_pitch.in_zone[x]/df_summ_batter_pitch.pitches[x] if df_summ_batter_pitch.pitches[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['chase_percent'] = [df_summ_batter_pitch.ozone_swing[x]/(df_summ_batter_pitch.pitches[x] - df_summ_batter_pitch.in_zone[x]) if (df_summ_batter_pitch.pitches[x]- df_summ_batter_pitch.in_zone[x]) != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['chase_contact'] = [df_summ_batter_pitch.ozone_contact[x]/df_summ_batter_pitch.ozone_swing[x] if df_summ_batter_pitch.ozone_swing[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['swing_percent'] = [df_summ_batter_pitch.swings[x]/df_summ_batter_pitch.pitches[x] if df_summ_batter_pitch.pitches[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['whiff_rate'] = [df_summ_batter_pitch.whiffs[x]/df_summ_batter_pitch.swings[x] if df_summ_batter_pitch.swings[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['swstr_rate'] = [df_summ_batter_pitch.whiffs[x]/df_summ_batter_pitch.pitches[x] if df_summ_batter_pitch.pitches[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['heart_zone_percent'] = [df_summ_batter_pitch.heart[x]/df_summ_batter_pitch.attack_zone[x] if df_summ_batter_pitch.attack_zone[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['shadow_zone_percent'] = [df_summ_batter_pitch.shadow[x]/df_summ_batter_pitch.attack_zone[x] if df_summ_batter_pitch.attack_zone[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]    

    df_summ_batter_pitch['chase_zone_percent'] = [df_summ_batter_pitch.chase[x]/df_summ_batter_pitch.attack_zone[x] if df_summ_batter_pitch.attack_zone[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]    

    df_summ_batter_pitch['waste_zone_percent'] = [df_summ_batter_pitch.waste[x]/df_summ_batter_pitch.attack_zone[x] if df_summ_batter_pitch.attack_zone[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]    
    
    
    df_summ_batter_pitch['heart_zone_swing_percent'] = [df_summ_batter_pitch.heart_swing[x]/df_summ_batter_pitch.heart[x] if df_summ_batter_pitch.heart[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]

    df_summ_batter_pitch['shadow_zone_swing_percent'] = [df_summ_batter_pitch.shadow_swing[x]/df_summ_batter_pitch.shadow[x] if df_summ_batter_pitch.shadow[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]    

    df_summ_batter_pitch['chase_zone_swing_percent'] = [df_summ_batter_pitch.chase_swing[x]/df_summ_batter_pitch.chase[x] if df_summ_batter_pitch.chase[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]    

    df_summ_batter_pitch['waste_zone_swing_percent'] = [df_summ_batter_pitch.waste_swing[x]/df_summ_batter_pitch.waste[x] if df_summ_batter_pitch.waste[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]    
    



    df_summ_batter_pitch['xwoba_percent'] = [df_summ_batter_pitch.xwoba[x]/df_summ_batter_pitch.xwoba_codes[x] if df_summ_batter_pitch.xwoba_codes[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]
    df_summ_batter_pitch['xwoba_percent_contact'] = [df_summ_batter_pitch.xwoba_contact[x]/df_summ_batter_pitch.bip[x] if df_summ_batter_pitch.bip[x] != 0 else np.nan for x in range(len(df_summ_batter_pitch))]




    df_summ_batter_pitch['bip'] = df_summ_batter_pitch['bip'].fillna(0)

    return df_summ_batter_pitch