import pandas as pd import numpy as np import statsmodels.formula.api as smf import statsmodels.api as sm from skmisc import * from plotnine import * from plotnine.data import * import matplotlib.pyplot as plt import seaborn as sns from mpl_toolkits.mplot3d import Axes3D

SELECT game_date, event_type,event_detail, event_team,event_player_1,coords_x, coords_y,home_goalie,away_goalie, home_team,away_team,home_skaters, away_skaters,pbp_distance,event_angle, season, CASE WHEN event_type = 'GOAL' THEN 1 ELSE 0 END AS goal, CASE WHEN home_goalie is NULL THEN home_team ELSE 'NONE' END AS pulled_home_goal, CASE WHEN away_goalie is NULL THEN away_team ELSE 'NONE' END AS pulled_away_goal FROM data WHERE event_type IN ('SHOT','MISS','GOAL') AND event_detail IN ('Wrist','Slap','Snap','Backhand','Wrap-around') AND pbp_distance<=60

df=df.copy() df['opp_goalie_pulled']=np.where( (df.event_team==df.home_team) & (df.pulled_away_goal==df.away_team),1,0) df['opp_goalie_pulled']=np.where( (df.event_team==df.away_team) & (df.pulled_home_goal==df.home_team),1,df.opp_goalie_pulled) df['is_home']=np.where(df['event_team']==df['home_team'],1,0) df['is_home']=df['is_home'].astype('category') df['man_advantage']=np.where(df['is_home']==1,df['home_skaters']-df['away_skaters'],df['away_skaters']-df['home_skaters']) df['event_detail']=df['event_detail'].astype('category') df['Quadrant']='NA' df['Quadrant']=np.where( (df['coords_y']>=0) & (df['coords_x']>=0),"I",df['Quadrant']) df['Quadrant']=np.where( (df['coords_y']>=0) & (df['coords_x']<=0),"II",df['Quadrant']) df['Quadrant']=np.where( (df['coords_y']<0) & (df['coords_x']<0),"III",df['Quadrant']) df['Quadrant']=np.where( (df['coords_y']<0) & (df['coords_x']>0),"IV",df['Quadrant']) df['AdjAngle']=np.where( (df['Quadrant']=="II")| (df['Quadrant']=="IV"),-1*df['event_angle'],df['event_angle']) df['OppGoalie']=np.where(df['event_team']==df['home_team'],df['away_goalie'],df['home_goalie']) df['OppGoalie']=np.where(df['event_team']==df['away_team'],df['home_goalie'],df['away_goalie']) df['IsGoalieOfInterest']=np.where(df['OppGoalie']=='ANDREI.VASILEVSKIY','Andrei Vasilevskiy','Other') df['IsShooterOfInterest']=np.where( df['event_player_1']=='AUSTON.MATTHEWS','Auston Matthews','Other') df=df[ (df['man_advantage']>=-2) & (df['man_advantage']<=2)]

MODEL='goal~pbp_distance+AdjAngle+I(AdjAngle**2)+AdjAngle*C(event_detail)+pbp_distance*C(event_detail)+C(IsGoalieOfInterest)+C(IsShooterOfInterest)+C(opp_goalie_pulled)+man_advantage' mod=smf.logit(formula=MODEL,data=df).fit() mod.summary()

df['y_pred_prob']=mod.predict(df) df['DistanceBin']=pd.qcut(df['pbp_distance'],9, precision=2) df['AngleBin']=pd.qcut(df['AdjAngle'],9, precision=2) heatmap=df.groupby(['DistanceBin','AngleBin'])['y_pred_prob'].mean().reset_index() heatmap=heatmap.pivot('DistanceBin','AngleBin','y_pred_prob') heatmap=heatmap.sort_index(level=0,ascending=False) heatmap=heatmap.fillna(0) f,ax = plt.subplots(figsize=(12, 9)) sns.heatmap(heatmap, annot=True, linewidths=.1, ax=ax)

d3_plots=df.sample(3000).dropna() sns.set(style = "darkgrid") fig = plt.figure(figsize=(12,12)) ax = fig.add_subplot(111, projection = '3d') x = d3_plots['pbp_distance'] y = d3_plots['AdjAngle'] z = d3_plots['y_pred_prob'] c=d3_plots['y_pred_prob'] ax.set_xlabel("Angle") ax.set_ylabel("Distance") ax.set_zlabel("PredProb") ax.scatter(x, y, z,c=c) plt.show()

fig = plt.figure(figsize=(12,12)) ax = fig.gca(projection='3d') ax.plot_trisurf( d3_plots['pbp_distance'],d3_plots['AdjAngle'],d3_plots['y_pred_prob'], cmap=plt.cm.viridis, linewidth=0.2) plt.show()

theme1 = theme( plot_title = element_text(size = 12, color = "white"), axis_text_x = element_text(size = 12, color = "white"), axis_text_y = element_text(size = 12, color = "white"), axis_title_y = element_text(size = 12, color = "white"), axis_title_x = element_text(size = 12, color = "white"), strip_text_x=element_text(size = 12, color = "white"), plot_background = element_rect(fill = "#676a6e"), panel_background = element_rect(fill = "#676a6e"), legend_title = element_text(size = 16, color = "white"), legend_background = element_rect(fill = "#676a6e",color="#676a6e",size=1), legend_text = element_text(size = 14, color = "white"), strip_background = element_rect(fill = "#676a6e"), legend_key=element_rect(fill = "#676a6e"), panel_spacing=(0.5), aspect_ratio=4/3 )

df_plots=df.sample(40000).dropna()

df_plots=df_plots[ (df_plots['man_advantage']<=2) & (df_plots['man_advantage']>=-2)] (ggplot(df_plots,aes(x='AdjAngle',y='y_pred_prob',color='factor(IsGoalieOfInterest)'))+geom_smooth(span=.7,method='loess')+ labs(x='Angle',y='Predicted Probability',color='Distance')+coord_cartesian(ylim = (0, .1)) +theme1)

(ggplot(df_plots,aes(x='AdjAngle',y='y_pred_prob',color='factor(IsShooterOfInterest)'))+geom_smooth(span=.5,method='loess')+ labs(x='Angle',y='Predicted Probability',color='Distance')+coord_cartesian(ylim = (0, .20))+theme1)

(ggplot(df_plots.dropna(),aes(x='AdjAngle',y='y_pred_prob',color='factor(event_detail)')) +geom_smooth(span=.75,method='loess')+labs(x='Angle',y='Predicted Probability',color='Distance')+theme1)