added poisson math to stat sims

app.py

@@ -1,8 +1,10 @@
 import streamlit as st
 import numpy as np
+from numpy import where as np_where
 import pandas as pd
 import gspread
 import plotly.express as px
+import scipy.stats as stats
 st.set_page_config(layout="wide")
 
 @st.cache_resource
@@ -47,6 +49,13 @@ gcservice_account, gcservice_account2, NFL_Data = init_conn()
 game_format = {'Win%': '{:.2%}', 'Vegas': '{:.2%}', 'Win% Diff': '{:.2%}'}
 american_format = {'First Inning Lead Percentage': '{:.2%}', 'Fifth Inning Lead Percentage': '{:.2%}'}
 
+def calculate_poisson(row):
+    mean_val = row['Mean_Outcome']
+    threshold = row['Prop']
+    cdf_value = stats.poisson.cdf(threshold, mean_val)
+    probability = 1 - cdf_value
+    return probability
+
 @st.cache_resource(ttl=600)
 def init_baselines():
     sh = gcservice_account.open_by_url(NFL_Data)
@@ -83,7 +92,9 @@ def init_baselines():
 
 game_model, overall_stats, timestamp, prop_frame, prop_trends, pick_frame = init_baselines()
 qb_stats = overall_stats[overall_stats['Position'] == 'QB']
+qb_stats = qb_stats.drop_duplicates(subset=['Player', 'Position'])
 non_qb_stats = overall_stats[overall_stats['Position'] != 'QB']
+non_qb_stats = non_qb_stats.drop_duplicates(subset=['Player', 'Position'])
 team_dict = dict(zip(prop_frame['Player'], prop_frame['Team']))
 t_stamp = f"Last Update: " + str(timestamp) + f" CST"
 
@@ -107,7 +118,9 @@ with tab1:
               st.cache_data.clear()
               game_model, overall_stats, timestamp, prop_frame, prop_trends, pick_frame = init_baselines()
               qb_stats = overall_stats[overall_stats['Position'] == 'QB']
+              qb_stats = qb_stats.drop_duplicates(subset=['Player', 'Position'])
               non_qb_stats = overall_stats[overall_stats['Position'] != 'QB']
+              non_qb_stats = non_qb_stats.drop_duplicates(subset=['Player', 'Position'])
               team_dict = dict(zip(prop_frame['Player'], prop_frame['Team']))
               t_stamp = f"Last Update: " + str(timestamp) + f" CST"
     line_var1 = st.radio('How would you like to display odds?', options = ['Percentage', 'American'], key='line_var1')
@@ -166,7 +179,9 @@ with tab3:
               st.cache_data.clear()
               game_model, overall_stats, timestamp, prop_frame, prop_trends, pick_frame = init_baselines()
               qb_stats = overall_stats[overall_stats['Position'] == 'QB']
+              qb_stats = qb_stats.drop_duplicates(subset=['Player', 'Position'])
               non_qb_stats = overall_stats[overall_stats['Position'] != 'QB']
+              non_qb_stats = non_qb_stats.drop_duplicates(subset=['Player', 'Position'])
               team_dict = dict(zip(prop_frame['Player'], prop_frame['Team']))
               t_stamp = f"Last Update: " + str(timestamp) + f" CST"
     split_var2 = st.radio("Would you like to view all teams or specific ones?", ('All', 'Specific Teams'), key='split_var2')
@@ -192,7 +207,9 @@ with tab4:
               st.cache_data.clear()
               game_model, overall_stats, timestamp, prop_frame, prop_trends, pick_frame = init_baselines()
               qb_stats = overall_stats[overall_stats['Position'] == 'QB']
+              qb_stats = qb_stats.drop_duplicates(subset=['Player', 'Position'])
               non_qb_stats = overall_stats[overall_stats['Position'] != 'QB']
+              non_qb_stats = non_qb_stats.drop_duplicates(subset=['Player', 'Position'])
               team_dict = dict(zip(prop_frame['Player'], prop_frame['Team']))
               t_stamp = f"Last Update: " + str(timestamp) + f" CST"
     split_var5 = st.radio("Would you like to view all teams or specific ones?", ('All', 'Specific Teams'), key='split_var5')
@@ -221,7 +238,9 @@ with tab5:
               st.cache_data.clear()
               game_model, overall_stats, timestamp, prop_frame, prop_trends, pick_frame = init_baselines()
               qb_stats = overall_stats[overall_stats['Position'] == 'QB']
+              qb_stats = qb_stats.drop_duplicates(subset=['Player', 'Position'])
               non_qb_stats = overall_stats[overall_stats['Position'] != 'QB']
+              non_qb_stats = non_qb_stats.drop_duplicates(subset=['Player', 'Position'])
               team_dict = dict(zip(prop_frame['Player'], prop_frame['Team']))
               t_stamp = f"Last Update: " + str(timestamp) + f" CST"
     col1, col2 = st.columns([1, 5])
@@ -317,12 +336,14 @@ with tab5:
                         player_outcomes = pd.merge(players_only, overall_file, left_index=True, right_index=True)
 
                         players_only['Mean_Outcome'] = overall_file.mean(axis=1)
+                        players_only['Prop'] = prop_var
+                        players_only['poisson_var'] = players_only.apply(calculate_poisson, axis=1)
                         players_only['10%'] = overall_file.quantile(0.1, axis=1)
                         players_only['90%'] = overall_file.quantile(0.9, axis=1)
                         if ou_var == 'Over':
-                            players_only['beat_prop'] = overall_file[overall_file > prop_var].count(axis=1)/float(total_sims)
+                            players_only['beat_prop'] = np.where(players_only['Prop'] <= 3, players_only['poisson_var'], overall_file[overall_file > prop_var].count(axis=1)/float(total_sims))
                         elif ou_var == 'Under':
-                            players_only['beat_prop'] = (overall_file[overall_file < prop_var].count(axis=1)/float(total_sims))
+                            players_only['beat_prop'] = np.where(players_only['Prop'] <= 3, 1 - players_only['poisson_var'], (overall_file[overall_file < prop_var].count(axis=1)/float(total_sims)))
 
                         players_only['implied_odds'] = np.where(line_var <= 0, (-(line_var)/((-(line_var))+100)), 100/(line_var+100))
 
@@ -368,7 +389,9 @@ with tab6:
               st.cache_data.clear()
               game_model, overall_stats, timestamp, prop_frame, prop_trends, pick_frame = init_baselines()
               qb_stats = overall_stats[overall_stats['Position'] == 'QB']
+              qb_stats = qb_stats.drop_duplicates(subset=['Player', 'Position'])
               non_qb_stats = overall_stats[overall_stats['Position'] != 'QB']
+              non_qb_stats = non_qb_stats.drop_duplicates(subset=['Player', 'Position'])
               team_dict = dict(zip(prop_frame['Player'], prop_frame['Team']))
               t_stamp = f"Last Update: " + str(timestamp) + f" CST"
     col1, col2 = st.columns([1, 5])
@@ -499,16 +522,17 @@ with tab6:
                                     prop_check = (overall_file - prop_file)
             
                                     players_only['Mean_Outcome'] = overall_file.mean(axis=1)
+                                    players_only['Book'] = players_only['Player'].map(book_dict)
+                                    players_only['Prop'] = players_only['Player'].map(prop_dict)
+                                    players_only['poisson_var'] = players_only.apply(calculate_poisson, axis=1)
                                     players_only['10%'] = overall_file.quantile(0.1, axis=1)
                                     players_only['90%'] = overall_file.quantile(0.9, axis=1)
-                                    players_only['Over'] = prop_check[prop_check > 0].count(axis=1)/float(total_sims)
+                                    players_only['Over'] = np_where(players_only['Prop'] <= 3, players_only['poisson_var'], prop_check[prop_check > 0].count(axis=1)/float(total_sims))
                                     players_only['Imp Over'] = players_only['Player'].map(over_dict)
                                     players_only['Over%'] = players_only[["Over", "Imp Over"]].mean(axis=1)
-                                    players_only['Under'] = prop_check[prop_check < 0].count(axis=1)/float(total_sims)
+                                    players_only['Under'] = np_where(players_only['Prop'] <= 3, 1 - players_only['poisson_var'], prop_check[prop_check < 0].count(axis=1)/float(total_sims))
                                     players_only['Imp Under'] = players_only['Player'].map(under_dict)
                                     players_only['Under%'] = players_only[["Under", "Imp Under"]].mean(axis=1)
-                                    players_only['Prop'] = players_only['Player'].map(prop_dict)
-                                    players_only['Book'] = players_only['Player'].map(book_dict)
                                     players_only['Prop_avg'] = players_only['Prop'].mean() / 100
                                     players_only['prop_threshold'] = .10
                                     players_only = players_only[players_only['Mean_Outcome'] > 0]
@@ -652,16 +676,17 @@ with tab6:
                                 prop_check = (overall_file - prop_file)
         
                                 players_only['Mean_Outcome'] = overall_file.mean(axis=1)
+                                players_only['Book'] = players_only['Player'].map(book_dict)
+                                players_only['Prop'] = players_only['Player'].map(prop_dict)
+                                players_only['poisson_var'] = players_only.apply(calculate_poisson, axis=1)
                                 players_only['10%'] = overall_file.quantile(0.1, axis=1)
                                 players_only['90%'] = overall_file.quantile(0.9, axis=1)
-                                players_only['Over'] = prop_check[prop_check > 0].count(axis=1)/float(total_sims)
+                                players_only['Over'] = np_where(players_only['Prop'] <= 3, players_only['poisson_var'], prop_check[prop_check > 0].count(axis=1)/float(total_sims))
                                 players_only['Imp Over'] = players_only['Player'].map(over_dict)
                                 players_only['Over%'] = players_only[["Over", "Imp Over"]].mean(axis=1)
-                                players_only['Under'] = prop_check[prop_check < 0].count(axis=1)/float(total_sims)
+                                players_only['Under'] = np_where(players_only['Prop'] <= 3, 1 - players_only['poisson_var'], prop_check[prop_check < 0].count(axis=1)/float(total_sims))
                                 players_only['Imp Under'] = players_only['Player'].map(under_dict)
                                 players_only['Under%'] = players_only[["Under", "Imp Under"]].mean(axis=1)
-                                players_only['Book'] = players_only['Player'].map(book_dict)
-                                players_only['Prop'] = players_only['Player'].map(prop_dict)
                                 players_only['Prop_avg'] = players_only['Prop'].mean() / 100
                                 players_only['prop_threshold'] = .10
                                 players_only = players_only[players_only['Mean_Outcome'] > 0]