Methodology
Universe
S&P 500 as of each month-end, reconstructed point-in-time from Wikipedia's constituents and changes tables. Tickers that were added or removed during the lookback window are tracked correctly — a stock that joined the index in 2024 will only appear in screens from 2024 onward.Wikipedia is well-maintained but not authoritative. We plan to swap to a SynapseDiscovery index-constituents endpoint when one is available.
Rebalance
Monthly, at the last trading day of each month. Holdings change only on rebalance dates — within a month, the portfolio holds the same names.
Construction
Long-only, equal-weight, top quintile (Q5). For each factor we compute a signal cross-sectionally on the as-of universe, winsorize at the 1st and 99th percentiles, z-score within the universe, and rank into 5 quintiles. The Q5 portfolio is the top 20% of names, equal-weighted.
Returns
Daily total returns (splits + reinvested dividends), pulled from SynapseDiscovery as
adjustment=total_return. Portfolio return on day d is the equal-weight mean of held names' daily total returns. No transaction costs or slippage are modeled.Benchmark
SPY total return, on the same daily basis.
Trailing windows
1W / 1M / 3M / 6M / YTD / 1Y are computed from the cumulative return series: latest cumulative value divided by the cumulative value at the window anchor.
Factors
- Momentum
- 12-month total return, skipping the most recent 21 trading days (the 12-1 convention) to sidestep short-term reversal. Higher = better.
- Value
- Trailing earnings yield — sum of the last four reported quarterly
eps_diluteddivided by the latest price. Higher E/P = cheaper = top quintile. - Quality
- Most recently reported Return on Equity, taken directly from Synapse's
return_on_equityfield. Higher = better. - Low Volatility
- Negative of the trailing 252-day daily-return standard deviation. Lower realized vol = top quintile. The defensive factor: captures the empirical low-vol anomaly.
- Size
- Negative log market cap (shares × price). Smaller companies rank higher — the classic small-cap premium.
Fundamentals (Value, Quality, Size) are read point-in-time: a quarterly row is treated as "known" only after period_end + 60 days, approximating the 10-Q filing lag. No look-ahead bias from later restatements.
Known limitations
- Universe sourced from Wikipedia rather than an authoritative index provider — small day-to-day membership drift possible.
- No transaction costs, taxes, or slippage in the return series. Real-world performance would be lower.
- Equal-weight is naive vs. cap- or signal-weighted alternatives.