EXP-012 verdict — Connors RSI(2) daily
- Run date: 2026-09-17
- Execution model:
v2 - Signal verdict:
signal_positive - Standalone operation verdict:
standalone_hold - Overall verdict:
hold - Nature: a historical robustness recheck. Not new OOS.
The first run was discarded after a leak was found: the control-group candidate filter used signals after the candidate date. A rerun of the same contract, fixed to exclude only signals before the candidate date, was made the canonical version (D-036).
Grounds for the verdict
The generalization group of 23 symbols, historical_recheck (2023-01-01~2026-08-26) and normalized_actual_proxy were used as the 1st (primary) cell.
| Item | Result |
|---|---|
| Actual strategy trades | 253 |
| Distinct actual entry dates | 253 days |
| Matched signal events | 489 |
| Mean event effect (signal − control) | +2.99%p |
| Monthly block bootstrap 95% interval | [+1.53%p, +4.36%p] |
| Median CAGR per symbol | 1.71% |
| Median excess CAGR per symbol | -11.58%p |
| Median MDD per symbol | -14.15% |
| Median net expectancy per trade per symbol | 1.06% |
| Median holding-day exposure per symbol | 7.42% |
Capital-starved (starved) cells | 0 |
The sample sizes passed the preregistered gate (100 trades, 50 entry days, 80% match success rate). The result is therefore recorded as decidable, not as hold.
Signal verdict
signal_positive. The 5-trading-day net return after an RSI oversold signal was higher than on ordinary uptrend trading days for the same symbol and period, and the entire monthly block bootstrap interval is above 0.
The generalization group's mean net return per signal event was about +1%, versus about -2% for the control. The event effect is positive, but the limitation of a historical recheck that repeatedly viewed the same period remains.
Standalone operation verdict
standalone_hold. The median net expectancy per trade per symbol was +1.06%, but median holding-day exposure was below the interim operating gate of 10%, and the median excess CAGR was -11.58%p. Even frictionless, at 2x costs and at 5bp one-way slippage, recent-period median exposure stayed around 7%, and removing costs did not change the negative direction of the standalone verdict.
Decision
connors_rsi2 is not raised to a live-trading candidate. However, since the signal verdict is positive, grounds remain to test capital use and simultaneous signal handling in the shared-capital portfolio (EXP-013).
Once the rules are frozen and enough newly observed data accumulate, a separate forward validation can be designed. Until then, further RSI parameter searches and 5-minute expansion are stopped. ETF dual momentum must be registered again as a separate strategy contract.
The ledgers and reproduction information are kept in the 0 , EXP-012 run.json (original supporting file), and the 2 .
Source and migration evidence
Source: reports/verdicts/EXP-012-connors-rsi2.md
Original SHA-256: 13501df2b43497a0221b959492fd9ae9a9f65466e489a18a4ccab541bf166918
Raw data and account information are not part of the public package. The original results do not replace statistical or scientific revalidation. The English text is a translation; numbers, tables and charts are carried over unchanged.