Models
Rules-based stock pickers, each tested on 2022–2026 data it was never tuned on. Every one beat SPY in that test.
3 live models10 bps costsOOS: Jan 2022 – Sep 2026
Out-of-sample annual return vs SPY
Backtest only counts stocks after they joined the S&P 500.
Momentum Top 15+26.6%Quality Growth+14.7%Momentum Blend+16.0%
SPY+12.0%
Momentum Top 15
FlagshipLive
Owns the 15 S&P 500 stocks with the strongest 12-month gains (skipping the latest month), rebalanced monthly.
Out-of-sample / yr
+26.6%
SPY +12.0%
Max drawdown
-25.3%
SPY -23.9%
Holdings
15
Monthly
1D +3.1% · SPY +0.4% · today's holdings
Prices only: the backtest uses the full rules.
Quality Growth
Live
Owns the 15 S&P 500 companies with high returns on capital whose revenue growth is speeding up, from their latest SEC filings. Rebalanced monthly.
Out-of-sample / yr
+14.7%
SPY +12.0%
Max drawdown
-18.8%
SPY -23.9%
Holdings
15
Monthly
1D +2.0% · SPY +0.4% · today's holdings
Point-in-time SEC quarterly filings: every number is used only from the day it was filed, as first reported.
Momentum Blend
Live
Half smooth-trend momentum stocks (Core), half breakout growth leaders (up to 10), rebuilt monthly from the S&P 1500.
Out-of-sample / yr
+16.0%
SPY +12.0%
Max drawdown
-28.1%
SPY -23.9%
Holdings
21
Monthly
1D +0.3% · SPY +0.4% · today's holdings
Prices only: the backtest uses the full rules (no fundamentals in this model).
Verified track record. Every signal is frozen and timestamped when it's made, and live returns count only from the next close. See the live record →
Backtests are hypothetical: after 10 bps trading costs per side, counting a stock only while it was actually in the index. Companies that were later acquired or delisted are still missing, so some survivorship bias remains. Past results don't predict future returns.
What we tested that didn't work
Shown for honesty. Same costs and caveats.
M6 Multi-factor 25, buffer 4N, small caps onlyOOS +2.4% vs SPY +12.0%
Best clean-sheet multi-factor in-sample; collapsed out-of-sample. In-sample +17.4% (SPY +18.3%); out-of-sample max drawdown -22.7%.
coteau