James McCool
Refactor showdown baseline initialization in Streamlit app to use specific slate names for improved clarity and consistency, and remove unnecessary slate filtering in showdown functions.
40e7b2a
| import streamlit as st | |
| import pandas as pd | |
| import numpy as np | |
| from pymongo import MongoClient | |
| from database import db | |
| def init_DK_SD_seed_frames(slate, split, translation_dict): | |
| # Now dynamic | |
| collection = db[translation_dict[slate]] | |
| cursor = collection.find().limit(split) | |
| raw_display = pd.DataFrame(list(cursor)) | |
| raw_display = raw_display[['CPT', 'FLEX1', 'FLEX2', 'FLEX3', 'FLEX4', 'FLEX5', 'salary', 'proj', 'Team', 'Team_count', 'Secondary', 'Secondary_count', 'Own']] | |
| DK_seed = raw_display.to_numpy() | |
| return DK_seed | |
| def init_FD_SD_seed_frames(slate, split, translation_dict): | |
| # Now dynamic | |
| collection = db[translation_dict[slate]] | |
| cursor = collection.find().limit(split) | |
| raw_display = pd.DataFrame(list(cursor)) | |
| raw_display = raw_display[['CPT', 'FLEX1', 'FLEX2', 'FLEX3', 'FLEX4', 'FLEX5', 'salary', 'proj', 'Team', 'Team_count', 'Secondary', 'Secondary_count', 'Own']] | |
| FD_seed = raw_display.to_numpy() | |
| return FD_seed | |
| def init_SD_baselines(slate_var): | |
| collection = db['DK_SD_NFL_ROO'] | |
| cursor = collection.find() | |
| raw_display = pd.DataFrame(list(cursor)) | |
| raw_display = raw_display[raw_display['version'] == 'overall'] | |
| raw_display = raw_display[['Player', 'Position', 'Team', 'Opp', 'Salary', 'Floor', 'Median', 'Ceiling', 'Top_finish', 'Top_5_finish', 'Top_10_finish', '20+%', '2x%', '3x%', '4x%', | |
| 'Own', 'Small_Field_Own', 'Large_Field_Own', 'Cash_Field_Own', 'CPT_Own', 'LevX', 'version', 'slate', 'timestamp', 'player_id', 'site']] | |
| raw_display = raw_display.rename(columns={'player_id': 'player_ID'}) | |
| raw_display['Small_Field_Own'] = raw_display['Large_Field_Own'] | |
| raw_display['small_CPT_Own_raw'] = (raw_display['Small_Field_Own'] / 2) * ((100 - (100-raw_display['Small_Field_Own']))/100) | |
| small_cpt_own_var = 300 / raw_display['small_CPT_Own_raw'].sum() | |
| raw_display['small_CPT_Own'] = raw_display['small_CPT_Own_raw'] * small_cpt_own_var | |
| raw_display['cpt_Median'] = raw_display['Median'] * 1.25 | |
| raw_display['STDev'] = raw_display['Median'] / 4 | |
| raw_display['CPT_STDev'] = raw_display['cpt_Median'] / 4 | |
| dk_raw = raw_display.dropna(subset=['Median']) | |
| collection = db['FD_SD_NFL_ROO'] | |
| cursor = collection.find() | |
| raw_display = pd.DataFrame(list(cursor)) | |
| raw_display = raw_display[raw_display['version'] == 'overall'] | |
| raw_display = raw_display[['Player', 'Position', 'Team', 'Opp', 'Salary', 'Floor', 'Median', 'Ceiling', 'Top_finish', 'Top_5_finish', 'Top_10_finish', '20+%', '2x%', '3x%', '4x%', | |
| 'Own', 'Small_Field_Own', 'Large_Field_Own', 'Cash_Field_Own', 'CPT_Own', 'LevX', 'version', 'slate', 'timestamp', 'player_id', 'site']] | |
| raw_display = raw_display.rename(columns={'player_id': 'player_ID'}) | |
| raw_display['Small_Field_Own'] = raw_display['Large_Field_Own'] | |
| raw_display['small_CPT_Own_raw'] = (raw_display['Small_Field_Own'] / 2) * ((100 - (100-raw_display['Small_Field_Own']))/100) | |
| small_cpt_own_var = 300 / raw_display['small_CPT_Own_raw'].sum() | |
| raw_display['small_CPT_Own'] = raw_display['small_CPT_Own_raw'] * small_cpt_own_var | |
| raw_display['cpt_Median'] = raw_display['Median'] * 1.25 | |
| raw_display['STDev'] = raw_display['Median'] / 4 | |
| raw_display['CPT_STDev'] = raw_display['cpt_Median'] / 4 | |
| fd_raw = raw_display.dropna(subset=['Median']) | |
| return dk_raw, fd_raw | |
| def convert_df(array): | |
| array = pd.DataFrame(array, columns=column_names) | |
| return array.to_csv().encode('utf-8') | |
| def calculate_SD_value_frequencies(np_array): | |
| unique, counts = np.unique(np_array[:, :6], return_counts=True) | |
| frequencies = counts / len(np_array) # Normalize by the number of rows | |
| combined_array = np.column_stack((unique, frequencies)) | |
| return combined_array | |
| def sim_SD_contest(Sim_size, seed_frame, maps_dict, Contest_Size): | |
| SimVar = 1 | |
| Sim_Winners = [] | |
| # Pre-vectorize functions | |
| vec_projection_map = np.vectorize(maps_dict['Projection_map'].__getitem__) | |
| vec_cpt_projection_map = np.vectorize(maps_dict['cpt_projection_map'].__getitem__) | |
| vec_stdev_map = np.vectorize(maps_dict['STDev_map'].__getitem__) | |
| vec_cpt_stdev_map = np.vectorize(maps_dict['cpt_STDev_map'].__getitem__) | |
| st.write('Simulating contest on frames') | |
| while SimVar <= Sim_size: | |
| fp_random = seed_frame[np.random.choice(seed_frame.shape[0], Contest_Size)] | |
| sample_arrays1 = np.c_[ | |
| fp_random, | |
| np.sum(np.random.normal( | |
| loc=np.concatenate([ | |
| vec_cpt_projection_map(fp_random[:, 0:1]), # Apply cpt_projection_map to first column | |
| vec_projection_map(fp_random[:, 1:-7]) # Apply original projection to remaining columns | |
| ], axis=1), | |
| scale=np.concatenate([ | |
| vec_cpt_stdev_map(fp_random[:, 0:1]), # Apply cpt_projection_map to first column | |
| vec_stdev_map(fp_random[:, 1:-7]) # Apply original projection to remaining columns | |
| ], axis=1)), | |
| axis=1) | |
| ] | |
| sample_arrays = sample_arrays1 | |
| final_array = sample_arrays[sample_arrays[:, 7].argsort()[::-1]] | |
| best_lineup = final_array[final_array[:, -1].argsort(kind='stable')[::-1][:1]] | |
| Sim_Winners.append(best_lineup) | |
| SimVar += 1 | |
| return Sim_Winners |