Week of March 9, 2026 — First full week of paper trading, calibrating stops
Portfolio performance
ALFRED paper portfolio: +1.4% vs SPY +0.2%. Outperformed on a low-volume week. Not reading too much into it — one week of outperformance is noise. The more important thing is that the system executed trades cleanly with no errors.
Signals that fired
Three entries this week across AAPL, NVDA, and AMD. All three triggered on SMA stack alignment + CMF confirmation. NVDA was the standout — entered Monday, exited Wednesday, +1.1% on the position. AAPL was flat. AMD small loss (-0.3%).
What worked
Half-Kelly position sizing behaved exactly as intended. When two signals fired simultaneously (NVDA + AMD), it reduced position size appropriately rather than over-concentrating. The drawdown circuit breaker wasn't triggered, which is what you want in a normal week.
What I got wrong
The 95% VaR guard was too conservative on AAPL — it flagged the position as high-risk before entry because AAPL had elevated recent volatility, but the actual trade was fine. Need to review whether the lookback window for VaR is calibrated correctly (currently using 60-day, might try 20-day).
Market observation
Low-volume week with no major macro events. Market was in drift mode. These weeks tend to produce false signals from momentum-only systems. The CMF confirmation layer helped avoid two potential false entries.
What I'm changing
Experimenting with 20-day vs 60-day lookback for VaR calculation. Will run backtests on both configurations this weekend and log the results. Hypothesis: 20-day is more responsive but noisier; 60-day is smoother but lags.