Simple construction methods remained competitive across the historical sample.
Independent equity research · Historical analysis · Read only
From factor signals to portfolio decisions.
Explore how equity factors identify opportunities, how portfolio construction affects risk and returns, and what historical results can tell us.
Historical researchProject 01 v1.0.2Project 02 v0.6.0Research only · No live trading
01 · Rank
Combine Momentum and Sales Yield rankings
02 · Construct
Compare portfolio weighting and risk controls
03 · Evaluate
Measure performance, risk and trading costs
04 · Validate
Check calculations and historical consistency
01 / Validated factor model
Historical factor performance.
Annualized return20.82%
Annualized volatility16.46%
Sharpe · no risk-free subtraction1.265
Maximum drawdown-20.92%
Growth of $1 invested · factor strategy vs SPY
Monthly, net of modeled costsSelected factor strategySPY2017–2026
The strategy combines Momentum and Sales Yield equally, selects the highest-ranked 20% of eligible stocks, and rebalances monthly. Returns include modeled transaction costs and use the same forward-return periods as the benchmark.
02 / Portfolio construction
Portfolio construction involves tradeoffs.
Across the tested methods, lower volatility did not eliminate severe historical drawdowns. Constraints helped manage concentration, with tradeoffs in return and turnover.
| Construction method | CAGR | Volatility | Sharpe | Max drawdown |
|---|---|---|---|---|
| Equal weight | 24.74% | 20.62% | 1.176 | -39.35% |
| Signal weight | 24.57% | 20.79% | 1.162 | -39.47% |
| Inverse volatility | 21.09% | 19.23% | 1.092 | -39.42% |
| Minimum variance | 17.57% | 17.97% | 0.992 | -39.74% |
| Signal-aware | 19.87% | 20.12% | 1.002 | -39.76% |
Project 02 typically uses a top-30 ranked candidate universe with daily naturally drifting weights, unlike Project 01's exact top-quintile baseline. These are not interchangeable portfolio results.
Position and sector constraints reduced concentration, often at a return or turnover cost.
Lower realized volatility did not necessarily prevent severe drawdowns.
03 / Validation and evidence
Validation and research inputs.
Baseline reconciliation
117 / 117monthly portfolio observations matched
Gross returns and net returns matched exactly; turnover differences were floating-point noise. The later 2023–2026 comparison is a Later Historical Evaluation, not a live-forward out-of-sample test.
Inputs & assumptions
- Market data
- Yahoo Finance through yfinance
- Fundamentals
- SEC EDGAR Companyfacts (annual 10-K/10-K/A)
- Benchmark
- SPY
- Modeled transaction cost
- 10 basis points
- Historical monthly returns through
- 2026-09-30
04 / Limitations
Research limitations.
Historical analysis onlyHistorical data
The historical universe uses current S&P 500 constituents, introducing survivorship bias. Sector labels are not historical, and Sales Yield uses annual diluted shares as a market-capitalization proxy. SEC fundamentals have uneven historical coverage.
Performance interpretation
Model selection and evaluation use historical data, not an untouched live-forward sample. Modeled transaction costs do not capture taxes, market impact, slippage, or execution constraints.
Research use and scope
Stress tests are deterministic sensitivities, not forecasts. Results are for historical research, not live trading or investment recommendations.