Page 04The classifieds — signals wait here until they earn the front page3 screeners
S-01Falling KnifeA risk-reducer, not a market-beater — and we say so.lower drawdown than SPY,
not higher returnOOS · every multi-yr window165turned + watching
not higher returnOOS · every multi-yr window165turned + watching
| Ticker | Sector | Sell-off depth | Days since break | State |
|---|---|---|---|---|
| ABNB | Consumer Discretionary | −18% | 39 | turned |
| ADP | Industrials | −29% | 9 | turned |
| A | Health Care | −28% | 11 | watching |
| ABBV | Health Care | −19% | 6 | watching |
Known blind spot: violent V-recoveries get filtered by the sell-off gate — documented, not hidden.open screener →
S-02PullbackUptrend retraces, ranked by the FINRA dark-pool selection model.+0.55R OOS at signal level ·
Score +0.145R vs −0.02R controlone test era · monitored for decay90candidates
Score +0.145R vs −0.02R controlone test era · monitored for decay90candidates
| Ticker | Sector | vs 50-day | vs 52-wk high | FINRA score | State |
|---|---|---|---|---|---|
| APA | Energy | -3.3% | -16.0% | 36 | candidate |
| MSCI | Financials | +2.5% | -6.9% | 32 | candidate |
| MPWR | Information Technology | +3.8% | -6.7% | 31 | candidate |
| CINF | Financials | +3.3% | -1.5% | 31 | candidate |
Honest constraint: thousands of signals chase ~50 portfolio slots — the Score column exists because selection, not signal, is the bottleneck.open screener →
S-03MomentumTop-decile 12-1 monthly book, equal-weight. The ranking layer failed OOS — so there isn't one.equal-weight by design ·
ranking layer rejectedOOS train/test split50held · as of 2026-06-30
ranking layer rejectedOOS train/test split50held · as of 2026-06-30
| Rank | Ticker | 12-1 momentum | Weight | Held | State |
|---|---|---|---|---|---|
| #1 | LITE | +799% | 2.0% | 12 mo | held |
| #2 | WDC | +732% | 2.0% | 11 mo | held |
| #3 | MU | +690% | 2.0% | 8 mo | held |
| #11 | DELL | +249% | 2.0% | 1 mo | new |
Why equal-weight: rank #1 and #50 carry the same weight because no ranking signal survived out-of-sample. Boring is the finding.open screener →
House rule:every stat above is out-of-sample — if a number is missing, we haven't earned it yet. Screeners feed the ticker dossiers and the front page; they don't lead the site, the market does.