Refactor large_field_preset function to improve team-based filtering by introducing median and finish percentile rankings for enhanced sorting accuracy. Clean up temporary columns after processing to maintain DataFrame integrity.

global_func/large_field_preset.py

@@ -7,7 +7,11 @@ def large_field_preset(portfolio: pd.DataFrame, lineup_target: int):
 
         for team in portfolio['Stack'].unique():
             rows_to_drop = []
-            working_portfolio = portfolio[portfolio['Stack'] == team].sort_values(by='Finish_percentile', ascending = True)
+            working_portfolio = portfolio.copy()
+            working_portfolio['median_rank'] = working_portfolio['median_rank'].rank(method='first')
+            working_portfolio['finish_percentile_rank'] = working_portfolio['Finish_percentile'].rank(method='first')
+            working_portfolio['rank_agg'] = (working_portfolio['median_rank'] + working_portfolio['finish_percentile_rank']) / 2
+            working_portfolio = working_portfolio[working_portfolio['Stack'] == team].sort_values(by='rank_agg', ascending = True)
             working_portfolio = working_portfolio[working_portfolio['Lineup Edge'] > 0]
             working_portfolio = working_portfolio.reset_index(drop=True)
             curr_own_type_max = working_portfolio.loc[0, 'Own'] + (slack_var / 20 * working_portfolio.loc[0, 'Own'])
@@ -19,6 +23,7 @@ def large_field_preset(portfolio: pd.DataFrame, lineup_target: int):
                     curr_own_type_max = working_portfolio.loc[i, 'Own'] + (slack_var / 20 * working_portfolio.loc[i, 'Own'])
 
             working_portfolio = working_portfolio.drop(rows_to_drop).reset_index(drop=True)
+            working_portfolio = working_portfolio.drop(columns=['median_rank', 'finish_percentile_rank', 'rank_agg'])
             concat_portfolio = pd.concat([concat_portfolio, working_portfolio])
         if len(concat_portfolio) >= lineup_target:
             return concat_portfolio.sort_values(by='Finish_percentile', ascending=True).head(lineup_target)