dodox/smaug
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: address sanity-check issues + rebrand to Smaug

Browse Source 
Co-authored-by: dodox <dodox@users.noreply.local>
This commit is contained in:
claude
2026-05-04 16:32:00 +00:00
co-authored by dodox
parent 8c0085e503
commit 2e2be3e9c7
7 changed files with 167 additions and 103 deletions
Download Patch File Download Diff File Expand all files Collapse all files
backtest/backtest.py
+49 -28
View File
@@ -1,25 +1,49 @@
import logging
import math
from datetime import datetime, timedelta
from typing import Optional
import sqlite3
import config
logger = logging.getLogger(__name__)
def _load_signals_from_db(db_path: str) -> list[dict]:
def _load_signals_from_db(db_path: str, min_score: float, min_cluster_size: int) -> list[dict]:
import sqlite3
conn = sqlite3.connect(db_path)
conn.row_factory = sqlite3.Row
rows = conn.execute(
"SELECT s.*, f.role FROM signals s "
"LEFT JOIN filings f ON f.ticker = s.ticker AND f.transaction_date = s.trigger_date "
"WHERE s.cluster_size >= 1"
"""
SELECT s.*, f.role FROM signals s
LEFT JOIN filings f ON f.ticker = s.ticker AND f.transaction_date = s.trigger_date
WHERE s.score >= ? AND s.cluster_size >= ?
""",
(min_score, min_cluster_size),
).fetchall()
conn.close()
return [dict(r) for r in rows]
def _first_close_on_or_after(price_data, target_date: datetime) -> float:
"""Return the closing price on the first trading day on or after target_date."""
for ts, row in price_data["Close"].items():
ts_date = ts.to_pydatetime().replace(tzinfo=None)
if ts_date.date() >= target_date.date():
return float(row)
raise ValueError(f"No price data on or after {target_date.date()}")
def _first_close_before(price_data, target_date: datetime) -> float:
"""Return the closing price on the last trading day before or on target_date."""
result = None
for ts, row in price_data["Close"].items():
ts_date = ts.to_pydatetime().replace(tzinfo=None)
if ts_date.date() <= target_date.date():
result = float(row)
if result is None:
raise ValueError(f"No price data on or before {target_date.date()}")
return result
def run_backtest(
db_path: str = None,
holding_days: int = None,
@@ -34,15 +58,14 @@ def run_backtest(
db_path = db_path or config.DB_PATH
holding_days = holding_days or config.HOLDING_PERIOD_DAYS
signals = _load_signals_from_db(db_path)
signals = [s for s in signals if s["score"] >= min_score and s["cluster_size"] >= min_cluster_size]
signals = _load_signals_from_db(db_path, min_score, min_cluster_size)
if not signals:
logger.warning("No signals found matching criteria")
return {}
results = []
spy_returns = {}
spy_cache: dict[tuple, float] = {}
for signal in signals:
ticker = signal["ticker"]
@@ -58,16 +81,17 @@ def run_backtest(
try:
stock_data = yf.download(
ticker,
start=(entry_date - timedelta(days=5)).strftime("%Y-%m-%d"),
start=entry_date.strftime("%Y-%m-%d"),
end=(exit_date + timedelta(days=5)).strftime("%Y-%m-%d"),
progress=False,
auto_adjust=True,
)
if stock_data.empty:
logger.debug(f"No price data for {ticker}")
continue
entry_price = float(stock_data["Close"].iloc[0])
exit_price = float(stock_data["Close"].iloc[-1])
entry_price = _first_close_on_or_after(stock_data, entry_date)
exit_price = _first_close_before(stock_data, exit_date)
stock_return = (exit_price - entry_price) / entry_price
except Exception as e:
@@ -75,25 +99,25 @@ def run_backtest(
continue
period_key = (entry_date_str, holding_days)
if period_key not in spy_returns:
if period_key not in spy_cache:
try:
spy_data = yf.download(
"SPY",
start=(entry_date - timedelta(days=5)).strftime("%Y-%m-%d"),
start=entry_date.strftime("%Y-%m-%d"),
end=(exit_date + timedelta(days=5)).strftime("%Y-%m-%d"),
progress=False,
auto_adjust=True,
)
if not spy_data.empty:
spy_entry = float(spy_data["Close"].iloc[0])
spy_exit = float(spy_data["Close"].iloc[-1])
spy_returns[period_key] = (spy_exit - spy_entry) / spy_entry
spy_entry = _first_close_on_or_after(spy_data, entry_date)
spy_exit = _first_close_before(spy_data, exit_date)
spy_cache[period_key] = (spy_exit - spy_entry) / spy_entry
else:
spy_returns[period_key] = 0.0
spy_cache[period_key] = 0.0
except Exception:
spy_returns[period_key] = 0.0
spy_cache[period_key] = 0.0
spy_return = spy_returns.get(period_key, 0.0)
spy_return = spy_cache[period_key]
alpha = stock_return - spy_return
results.append({
@@ -114,12 +138,10 @@ def run_backtest(
win_rate = sum(1 for r in returns if r > 0) / len(returns)
avg_return = sum(returns) / len(returns)
avg_alpha = sum(alphas) / len(alphas)
import math
std_dev = math.sqrt(sum((r - avg_return) ** 2 for r in returns) / len(returns))
sharpe = (avg_return / std_dev * math.sqrt(252 / holding_days)) if std_dev > 0 else 0.0
summary = {
return {
"total_signals": len(results),
"win_rate": round(win_rate, 4),
"avg_return": round(avg_return, 4),
@@ -129,19 +151,18 @@ def run_backtest(
"results": results,
}
return summary
def print_summary(summary: dict):
if "error" in summary:
print(f"Error: {summary['error']}")
return
print(f"\n{'='*40}")
width = 40
print(f"\n{'=' * width}")
print(f"Backtest Results ({summary['holding_days']}-day hold)")
print(f"{'='*40}")
print(f"{'=' * width}")
print(f"Total signals: {summary['total_signals']}")
print(f"Win rate: {summary['win_rate']:.1%}")
print(f"Avg return: {summary['avg_return']:.2%}")
print(f"Avg alpha vs SPY: {summary['avg_alpha_vs_spy']:.2%}")
print(f"Sharpe ratio: {summary['sharpe_ratio']:.2f}")
print(f"{'='*40}\n")
print(f"{'=' * width}\n")