394 lines
16 KiB
Python

from collections import defaultdict
from statistics import median
from typing import Dict, Any, List
import logging
import os
import math
from dotenv import load_dotenv
load_dotenv()
LOG_LEVEL = os.environ.get('LOG_LEVEL', 'INFO').upper()
import logging
logging.basicConfig(
format='%(asctime)s %(levelname)s: %(message)s',
datefmt='%Y-%m-%d %H:%M:%S',
level=LOG_LEVEL
)
from helpers.SimpleBeatLeaderAPI import SimpleBeatLeaderAPI
from helpers.BeatLeaderAPI import BeatLeaderAPI
from saberlist.utils import prompt_for_player_id, load_history, save_history, normalize_difficulty_name
from helpers.ScoreSaberAPI import ScoreSaberAPI
from clients.scoresaber.models.get_api_player_player_id_scores_sort import GetApiPlayerPlayerIdScoresSort
"""Testing
api = ScoreSaberAPI()
song_count = 40
bin_size = 0.25
bin_sort = False
"""
def playlist_strategy_scoresaber_accuracy_gaps(
api: ScoreSaberAPI,
song_count: int = 40,
bin_size: float = 0.25,
bin_sort: bool = False
) -> List[Dict[str, Any]]:
"""
Build a playlist of songs where the player's accuracy is furthest below the median accuracy
for their star rating range. Songs are grouped into bins by star rating to ensure fair comparison.
:param api: ScoreSaberAPI instance for making API calls
:param song_count: Number of songs to include in the playlist
:param bin_size: Size of star rating bins for grouping similar difficulty songs
:param bin_sort: Whether to sort the bins by star rating
:return: A tuple containing (list of song dictionaries, playlist title string)
"""
player_id = prompt_for_player_id()
history = load_history()
history.setdefault('scoresaber_accuracy_gaps', {})
history.setdefault('playlist_counts', {})
# Get the current count and increment it
count_key = 'scoresaber_accuracy_gaps'
current_count = history['playlist_counts'].get(count_key, 0)
new_count = current_count + 1
history['playlist_counts'][count_key] = new_count
# Fetch player scores
scores_data = api.get_player_scores(
player_id=player_id,
use_cache=True,
limit=100, # per page
sort=GetApiPlayerPlayerIdScoresSort.RECENT
)
ranked_scores = [score for score in scores_data.get('playerScores', [])
if score.get('leaderboard', {}).get('stars', 0) != 0]
if not ranked_scores:
logging.warning(f"No ranked scores found for player ID {player_id} on ScoreSaber.")
return [], ""
logging.debug(f"Found {len(ranked_scores)} ranked scores for player ID {player_id} on ScoreSaber.")
# Get min and max star ratings
min_stars = min(score['leaderboard']['stars'] for score in ranked_scores)
max_stars = max(score['leaderboard']['stars'] for score in ranked_scores)
star_range = max_stars - min_stars
# Determine number of bins
num_bins = math.ceil(star_range / bin_size)
logging.info(f"Using bin size: {bin_size}, resulting in {num_bins} bins.")
# Group accuracies by bins
bin_to_accuracies = defaultdict(list)
for score in ranked_scores:
# Calculate accuracy
try:
modified_score = score['score']['modifiedScore']
max_score = score['leaderboard']['maxScore']
accuracy = modified_score / max_score if max_score else 0
score['accuracy'] = accuracy
except Exception as e:
logging.error(f"Error calculating accuracy for score {score}: {e}")
continue
stars = score['leaderboard'].get('stars')
if stars is not None and accuracy is not None:
bin_index = int((stars - min_stars) / bin_size)
bin_to_accuracies[bin_index].append(accuracy)
# Calculate median accuracy for each bin
bin_to_median = {}
for bin_index, accuracies in bin_to_accuracies.items():
bin_to_median[bin_index] = median(accuracies)
bin_start = min_stars + bin_index * bin_size
bin_end = bin_start + bin_size
logging.debug(f"Median accuracy for bin {bin_index} (stars {bin_start:.2f} to {bin_end:.2f}): {bin_to_median[bin_index]:.4f}")
# Compute difference from median for each score
for score in ranked_scores:
stars = score['leaderboard'].get('stars')
accuracy = score.get('accuracy')
if stars is not None and accuracy is not None:
bin_index = int((stars - min_stars) / bin_size)
median_acc = bin_to_median.get(bin_index)
if median_acc is not None:
score['diff_from_median'] = accuracy - median_acc
else:
score['diff_from_median'] = float('inf') # Place entries with missing data at the end
else:
score['diff_from_median'] = float('inf') # Place entries with missing data at the end
# Sort scores by difference from median (ascending: most below median first)
ranked_scores.sort(key=lambda x: x.get('diff_from_median', float('inf')))
playlist_data = []
for score in ranked_scores:
if len(playlist_data) >= song_count:
break
accuracy = score['score'].get('accuracy', 0)
stars = score['leaderboard'].get('stars')
song_hash = score['leaderboard'].get('songHash')
if not song_hash or stars is None:
logging.debug(f"Skipping score due to missing hash or stars: {score}")
continue
difficulty_raw = score['leaderboard']['difficulty'].get('difficultyRaw', '')
game_mode = score['leaderboard']['difficulty'].get('gameMode', 'Standard')
game_mode = game_mode.replace('Solo', '') # Remove prefix 'Solo' from the game mode
difficulty = normalize_difficulty_name(difficulty_raw)
# Avoid reusing song+difficulty
if song_hash in history['scoresaber_accuracy_gaps'] and difficulty in history['scoresaber_accuracy_gaps'][song_hash]:
logging.debug(f"Skipping song {song_hash} with difficulty {difficulty} as it's in history.")
continue
song_dict = {
'hash': song_hash,
'difficulties': [
{
'name': difficulty,
'characteristic': game_mode
}
]
}
playlist_data.append(song_dict)
song_name = score['leaderboard']['songName']
song_artist = score['leaderboard']['songAuthorName']
logging.debug(f"Selected song for playlist: Name={song_name}, Artist={song_artist}, "
f"Accuracy={accuracy*100:.2f}%, Diff from Median={score['diff_from_median']*100:.2f}%")
# Update history
history['scoresaber_accuracy_gaps'].setdefault(song_hash, []).append(difficulty)
if not playlist_data:
logging.info("No new songs found to add to the playlist based on history for ScoreSaber accuracy gaps.")
else:
for song in playlist_data:
song_hash = song['hash']
difficulty = song['difficulties'][0]['name']
logging.info(f"Song added: Hash={song_hash}, Difficulty={difficulty}")
logging.info(f"Total songs added to playlist from ScoreSaber accuracy gaps: {len(playlist_data)}")
save_history(history)
playlist_title = f"scoresaber_accgraph-{new_count:02d}"
return playlist_data, playlist_title
def playlist_strategy_beatleader_accuracy_gaps(
api: SimpleBeatLeaderAPI,
song_count: int = 40,
bin_size: float = 0.25,
bin_sort: bool = False
) -> List[Dict[str, Any]]:
"""
Build a playlist of songs where the player's accuracy is furthest below the median accuracy
for their star rating range. Songs are grouped into bins by star rating to ensure fair comparison.
:param api: SimpleBeatLeaderAPI instance for making API calls
:param song_count: Number of songs to include in the playlist
:param bin_size: Size of star rating bins for grouping similar difficulty songs
:param bin_sort: Whether to sort the bins by star rating
:return: A tuple containing (list of song dictionaries, playlist title string)
"""
player_id = prompt_for_player_id()
history = load_history()
history.setdefault('beatleader_accuracy_gaps', {})
history.setdefault('playlist_counts', {})
# Get the current count and increment it
count_key = 'beatleader_accuracy_gaps'
current_count = history['playlist_counts'].get(count_key, 0)
new_count = current_count + 1
history['playlist_counts'][count_key] = new_count
# Fetch accuracy graph data
all_scores = api.get_player_accgraph(player_id)
if not all_scores:
logging.warning(f"No accgraph data found for player ID {player_id} on BeatLeader.")
return [], ""
logging.debug(f"Found {len(all_scores)} accgraph entries for player ID {player_id} on BeatLeader.")
# Collect all star ratings
star_ratings = [entry['stars'] for entry in all_scores if entry.get('stars') is not None]
if not star_ratings:
logging.warning("No star ratings found in accgraph data.")
return [], ""
min_stars = min(star_ratings)
max_stars = max(star_ratings)
star_range = max_stars - min_stars
# Remove the bin size calculation logic
num_bins = math.ceil(star_range / bin_size)
logging.info(f"Using bin size: {bin_size}, resulting in {num_bins} bins.")
# Group accuracies by bins
bin_to_accuracies = defaultdict(list)
for entry in all_scores:
stars = entry.get('stars')
acc = entry.get('acc')
if stars is not None and acc is not None:
bin_index = int((stars - min_stars) / bin_size)
bin_to_accuracies[bin_index].append(acc)
# Calculate median accuracy for each bin
bin_to_median = {}
for bin_index, accs in bin_to_accuracies.items():
bin_to_median[bin_index] = median(accs)
bin_start = min_stars + bin_index * bin_size
bin_end = bin_start + bin_size
logging.debug(f"Median accuracy for bin {bin_index} (stars {bin_start:.2f} to {bin_end:.2f}): {bin_to_median[bin_index]:.4f}")
# Compute difference from median for each score
for entry in all_scores:
stars = entry.get('stars')
acc = entry.get('acc')
if stars is not None and acc is not None:
bin_index = int((stars - min_stars) / bin_size)
median_acc = bin_to_median.get(bin_index)
if median_acc is not None:
entry['diff_from_median'] = acc - median_acc
else:
entry['diff_from_median'] = float('inf') # Place entries with missing data at the end
else:
entry['diff_from_median'] = float('inf') # Place entries with missing data at the end
# Sort scores by difference from median (ascending: most below median first)
all_scores.sort(key=lambda x: x.get('diff_from_median', float('inf')))
playlist_data = []
for score_entry in all_scores:
if len(playlist_data) >= song_count:
break
acc = score_entry.get('acc', 0)
stars = score_entry.get('stars')
song_hash = score_entry.get('hash')
if not song_hash or stars is None:
logging.debug(f"Skipping entry due to missing hash or stars: {score_entry}")
continue
# Use stars as a proxy for difficulty; adjust if you have actual difficulty levels
difficulty = score_entry.get('diff', '')
difficulty_characteristic = score_entry.get('mode', 'Standard')
if song_hash in history['beatleader_accuracy_gaps'] and difficulty in history['beatleader_accuracy_gaps'][song_hash]:
logging.debug(f"Skipping song {song_hash} with difficulty {difficulty} as it's in history.")
continue
song_dict = {
'hash': song_hash,
'difficulties': [
{
'name': difficulty,
'characteristic': difficulty_characteristic
}
]
}
playlist_data.append(song_dict)
logging.debug(f"Selected song for playlist: Hash={song_hash}, Difficulty={difficulty}, "
f"Accuracy={acc*100:.2f}%, Diff from Median={score_entry['diff_from_median']*100:.2f}%")
# Update history
history['beatleader_accuracy_gaps'].setdefault(song_hash, []).append(difficulty)
if not playlist_data:
logging.info("No new songs found to add to the playlist based on history for BeatLeader accuracy gaps.")
else:
for song in playlist_data:
song_hash = song['hash']
difficulty = song['difficulties'][0]['name']
logging.info(f"Song added: Hash={song_hash}, Difficulty={difficulty}")
logging.info(f"Total songs added to playlist from BeatLeader accuracy gaps: {len(playlist_data)}")
save_history(history)
playlist_title = f"accgraph-{new_count:02d}"
return playlist_data, playlist_title
def playlist_strategy_beatleader_lowest_acc(
api: BeatLeaderAPI,
song_count: int = 20
) -> List[Dict[str, Any]]:
player_id = prompt_for_player_id()
history = load_history()
history.setdefault('beatleader_lowest_acc', {})
history.setdefault('playlist_counts', {})
"""Selects songs with the lowest accuracy, avoiding reusing the same song+difficulty."""
# Get the current count and increment it
count_key = 'beatleader_lowest_acc'
current_count = history['playlist_counts'].get(count_key, 0)
new_count = current_count + 1
history['playlist_counts'][count_key] = new_count
scores_data = api.get_player_scores(player_id)
all_scores = scores_data.get('playerScores', [])
if not all_scores:
logging.warning(f"No scores found for player ID {player_id} on BeatLeader.")
return [], ""
logging.debug(f"Found {len(all_scores)} scores for player ID {player_id} on BeatLeader.")
# Sort by accuracy in ascending order (lowest first)
all_scores.sort(key=lambda x: x.get('score', {}).get('accuracy', float('inf')))
playlist_data = []
for score_entry in all_scores:
if len(playlist_data) >= song_count:
break
score = score_entry.get('score', {})
leaderboard = score_entry.get('leaderboard', {})
song_hash = leaderboard.get('songHash')
difficulty_raw = int(leaderboard.get('difficulty', ''))
game_mode = leaderboard.get('modeName', 'Standard')
accuracy = score.get('accuracy', 0)
if not song_hash or not difficulty_raw:
logging.debug(f"Skipping score due to missing song_hash or difficulty_raw: {score_entry}")
continue
difficulty = normalize_difficulty_name(difficulty_raw)
# avoid reusing song+difficulty
if song_hash in history['beatleader_lowest_acc'] and difficulty in history['beatleader_lowest_acc'][song_hash]:
logging.debug(f"Skipping song {song_hash} with difficulty {difficulty} as it's in history.")
continue
song_dict = {
'hash': song_hash,
'difficulties': [
{
'name': difficulty,
'characteristic': game_mode
}
]
}
playlist_data.append(song_dict)
logging.debug(f"Selected song for playlist: Hash={song_hash}, Difficulty={difficulty}, Accuracy={accuracy*100:.2f}%")
# Update history
history['beatleader_lowest_acc'].setdefault(song_hash, []).append(difficulty)
if not playlist_data:
logging.info("No new songs found to add to the playlist based on history for BeatLeader lowest accuracy.")
else:
for song in playlist_data:
song_hash = song['hash']
difficulty = song['difficulties'][0]['name']
logging.info(f"Song added: Hash={song_hash}, Difficulty={difficulty}")
logging.info(f"Total songs added to playlist from BeatLeader lowest accuracy: {len(playlist_data)}")
save_history(history)
return playlist_data, f"beatleader_lowest_acc-{new_count:02d}"