QSD AI Power Barbell Timing Sensitivity Sweep
Strategy ID: qsd_ai_power_barbell_steve_live_v1 · Generated: 2026-06-05T18:42:45.139858+00:00 · Data: Alpaca IEX adjusted daily close, 2022-02-28 to 2026-05-22
Recommendation
Use a semi-automatic Friday rebalance policy: compute the signal from the prior completed close, prepare target weights before the Friday decision window, and require an intraday/downside guard review before execution. The daily sweep supports Friday as acceptable, but Friday ranks 3/5 by Net Sharpe Ratio in this close-only study; intraday timing remains unproven.
Full automation vs semi-automation: prefer semi-automatic execution with an intraday guard for now. The weekday conclusion is based on causal daily bars, but open, open+30m, mid-day, and 15:30 ET execution cannot be resolved from this close-only cache. A fully automatic Friday job is reasonable only after adding an intraday/open-bar timing sweep and explicit downside veto rules.
Weekday Rebalance Sweep
Net Sharpe Ratio means Sharpe after the 5 bps turnover cost. Gross Sharpe Ratio is before that cost. Compound Annual Growth Rate (CAGR), Maximum Drawdown (MaxDD), turnover, and stress returns are shown below.
Execution Timing Sweep
Measured rows use daily close proxies only. Unavailable rows are intentionally visible because the current strategy research dataset does not contain intraday bars or open prices.
Risk By Timing Option
Key Table
| Variant | Status | Net Sharpe Ratio | Gross Sharpe Ratio | Compound Annual Growth Rate (CAGR) | Maximum Drawdown (MaxDD) | Avg turnover/day | Annualized turnover | Avg cost bps/day | Stress avg return | Stress worst return | Notes |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Monday | measured | 1.33 | 1.37 | 40.7% | -33.2% | 8.0% | 2020.3% | 0.40 | -2.7% | -12.2% | Prior close signal applied to same trading day's close-to-close return proxy; 5 bps turnover cost. |
| Tuesday | measured | 1.37 | 1.41 | 42.1% | -36.1% | 8.4% | 2119.9% | 0.42 | -2.8% | -12.9% | Prior close signal applied to same trading day's close-to-close return proxy; 5 bps turnover cost. |
| Wednesday | measured | 1.52 | 1.56 | 48.0% | -34.1% | 8.3% | 2087.9% | 0.41 | -2.7% | -12.4% | Prior close signal applied to same trading day's close-to-close return proxy; 5 bps turnover cost. |
| Thursday | measured | 1.46 | 1.50 | 46.0% | -31.3% | 8.3% | 2090.4% | 0.41 | -2.7% | -12.4% | Prior close signal applied to same trading day's close-to-close return proxy; 5 bps turnover cost. |
| Friday | measured | 1.40 | 1.43 | 43.2% | -32.9% | 8.1% | 2046.7% | 0.41 | -2.7% | -12.2% | Prior close signal applied to same trading day's close-to-close return proxy; 5 bps turnover cost. |
| Friday prior close signal -> close-to-close proxy | measured | 1.40 | 1.43 | 43.2% | -32.9% | 8.1% | 2046.7% | 0.41 | -2.7% | -12.2% | Prior close signal applied to same trading day's close-to-close return proxy; 5 bps turnover cost. |
| Friday same-day close signal -> next trading-day close proxy | measured | 1.32 | 1.36 | 40.4% | -31.8% | 8.4% | 2107.7% | 0.42 | -2.8% | -12.2% | Uses Friday close signal only after the close, then applies the next trading day's close-to-close return proxy. |
| Friday + daily risk guard proxy | measured | 1.44 | 1.48 | 42.0% | -30.5% | 8.5% | 2148.5% | 0.43 | -2.5% | -12.2% | Prior close signal applied to same trading day's close-to-close return proxy; 5 bps turnover cost. |
| Next open | unavailable | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | Unavailable in this artifact: the available cached strategy dataset is adjusted daily close only, with no open bars. |
| Near open / open+30m | unavailable | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | Unavailable without intraday minute bars or an open+30m execution price proxy. |
| Mid-day | unavailable | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | Unavailable without intraday bars. |
| 15:30 ET / close-30m | unavailable | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | Unavailable without intraday bars. |
| Close / Market-On-Close exact | unavailable | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | Only a close-to-close proxy is available; exact MOC fill quality is not measured. |
| Next trading-day open | unavailable | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | n/a | Unavailable without open bars. Same-day close signal to next trading-day close is shown as the closest daily proxy. |
Method And Limitations
Signals use prices available through the prior completed close for the weekday sweep. The portfolio is rebalanced weekly on the selected weekday and applied to the same trading day's close-to-close return as a daily execution proxy. The stress subset is days where SOXX close-to-close return is at or below the lower of -2.5% or the 10th percentile in the available sample.
The current artifact cannot answer true open, open+30m, mid-day, 15:30 ET, exact Market-On-Close, or next-open fill questions. Those need Alpaca open/high/low/close daily bars at minimum for open proxies and minute bars for intraday timing.
Strategy code link in this branch: research/2026-05-26_candidate_strategies/experiments/EXP-qsd-inspired-strategy-generation-v2/code/run_generation.py. Timing sweep tool: code/timing_sweep.py.