Scanner Tuning
The Hypothesis
The MorningEdge scanner classifies every gap-up candidate as either NEWS (confirmed catalyst found) or SUSPICIOUS (no catalyst detected). The thesis: gaps backed by real catalysts hold their gains; gaps without catalysts fade. Gemini's principle: "Volume without catalyst = mean reversion. Volume with catalyst = price discovery."
Day 0 Catalyst Breakdown
| Ticker | Label | Actual Catalyst | P/L | Close vs Open |
|---|---|---|---|---|
| NVAX | NEWS | Q4 earnings beat + guidance raise | +$187 | +14.11% |
| STLA | NEWS | Chinese EV tech / 22B loss | -$92 | -0.74% |
| INTU | NEWS | Earnings after bell (pre-positioning) | -$56 | +1.53% |
| OPEN | SUSPICIOUS | None (broad market sympathy) | +$287 | +7.87% |
| RGTI | SUSPICIOUS | $8.4M C-DAC contract (missed) | +$8 | +1.94% |
Finding 1: Not All NEWS Catalysts Are Equal
- Earnings beat + guidance raise (NVAX): Strongest catalyst type. +21.9% max extension. Price discovery.
- Bad-news gap (STLA): The 22B loss was the real story. Peaked in first 5 minutes, faded all day. Bad news with a silver lining is still bad news.
- Pre-earnings positioning (INTU): Gap was anticipation, not reaction. As Gemini warned: "Pre-earnings gaps are positioning, not conviction."
This led to tiered catalyst multipliers:
| Tier | Types | Multiplier |
|---|---|---|
| 1 (Highest) | EARNINGS, M_AND_A, FDA | 1.5x |
| 2 | ANALYST, CONTRACT, PARTNERSHIP | 1.2x |
| 3 | SECTOR, OTHER | 1.1x |
| None | SUSPICIOUS, TECHNICAL, UNKNOWN | 1.0x |
Finding 2: The RGTI Blind Spot
RGTI had a legitimate $8.4M C-DAC India contract that the Alpaca News API missed. This false negative motivated five supplementary news sources:
- Finviz — sector, short float, insider ownership
- Yahoo Finance — cross-reference RVOL + discovery
- Perplexity Sonar — web search for suspicious tickers (~$0.001/search)
- SEC EDGAR — 8-K filing detection
- xAI Grok — X/Twitter sentiment analysis
Implications
- Binary NEWS/SUSPICIOUS is insufficient. Catalyst quality matters as much as presence. Tiered multipliers now weight the composite score by catalyst type.
- Single-source news is unreliable. The RGTI blind spot proved that Alpaca's API alone misses catalysts. Multi-source verification reduces false negatives.
- Bad-news gaps should be flagged. Gemini Flash now classifies sentiment as BULLISH or BEARISH. Future iterations may apply negative multipliers.
- SUSPICIOUS ≠ exclude. OPEN (no catalyst) was the best performer. The 1.0x multiplier correctly deprioritizes without eliminating.