Executive finding
The model organized attention better than it proved alpha
Several experiments produced positive average raw returns at longer horizons. The same experiments generally showed negative average excess returns versus QQQ. That distinction matters because a rising market can make a weak selector look profitable.
Current conclusion: useful research triage and lifecycle observation, but no robust evidence that the broad screen beats a passive QQQ benchmark after a next-session-open entry.
Strict AExpanded E
Strict AExpanded E
Controlled ablation
What changed when one rule was relaxed
| Experiment | Rule profile | Rows | 5d raw | 5d excess | 20d raw | 20d excess |
|---|---|---|---|---|---|---|
| A | strict_filter_control | 3,881 | 0.19% | -0.20% | 0.14% | -0.91% |
| B | relax_rsi_upper_to_78_only | 4,366 | 0.18% | -0.23% | 0.02% | -0.94% |
| C | relax_rvol_to_070_only | 4,868 | 0.16% | -0.22% | 0.25% | -0.83% |
| D | relax_trend_legacy | 7,533 | 0.25% | -0.23% | 0.42% | -1.28% |
| E | expanded_v4671_all | 19,788 | 0.25% | -0.20% | 0.65% | -0.97% |
| F | relax_price_cap_only | 7,857 | 0.24% | -0.16% | 0.28% | -0.81% |
| G | relax_r60_only | 4,484 | 0.19% | -0.22% | 0.32% | -0.84% |
Returns are arithmetic averages of overlapping daily observations. They are not a portfolio backtest or audited performance record.
Ticker evidence lookup
Inspect coverage before telling a story
The tool reports V4.67.1 historical sample observations. It does not generate a buy or sell instruction.
Why confidence remains limited
Five constraints the headline cannot omit
- Daily signal rows overlap and are not independent trades.
- The research rules evolved while the same historical period was being studied, creating overfitting risk.
- The replay assumes next-session-open entry but does not model every spread, tax, slippage, or liquidity constraint.
- Sector metadata was neutralized defensively where point-in-time classifications were unavailable.
- A separate untouched out-of-sample period is still required before any claim of persistent alpha.
Primary sources first
Evidence trail
Documents and market data change. Verify the latest filing and state the observation date before relying on a conclusion.