# In training notebook cell-14:
USE_EMPIRICAL_THRESHOLDS = True   # Enable empirical mode
TARGET_CANDIDATES = 1500          # Target candidate count

# Thresholds are automatically derived in cell-16

from alpaca_trading.selection import SymbolDatabase
from alpaca_trading.selection.empirical_config import build_config_from_database

db = SymbolDatabase(db_path='data/symbol_database.db')
result = build_config_from_database(
    db=db,
    target_candidates=1500,       # How many candidates you want
    volume_percentile=50,         # P50 = median (top 50% by volume)
    price_percentile_low=5,       # Exclude bottom 5% (penny stocks)
    price_percentile_high=99,     # Exclude top 1% (too expensive)
)

# Use the derived config
config = result.config

# See what was derived
print(result.describe())
# Output:
# ======================================================================
# EMPIRICAL CONFIGURATION (derived from market data)
# ======================================================================
#
# DERIVED THRESHOLDS:
#   min_volume_equity             : $180,432  [P50 of equity volume]
#   min_price                     : 1.25  [P5 of equity price]
#   max_price                     : 892.50  [P99 of equity price]
#   sector_top_pct                : 41.67%  [calculated for 1500 target candidates]
#   min_per_sector                : 45  [median_sector_size / 10]
#   max_per_sector                : 450  [median_sector_size]

from alpaca_trading.selection.empirical_config import build_full_empirical_config

result = build_full_empirical_config(
    db=db,
    data_fetcher=fetcher,         # Required for correlation
    target_candidates=1500,
    estimate_correlations=True,   # Compute actual correlations
)

# max_correlation is now derived from P75 of pairwise correlations
print(f"max_correlation: {result.config.max_correlation:.2f}")

@dataclass
class EmpiricalConfigResult:
    config: SelectionConfig       # Ready-to-use config
    thresholds: Dict[str, Any]    # All derived values
    derivation_method: Dict[str, str]  # How each was derived
    data_summary: Dict[str, Any]  # Market stats used