Understanding Impact Scores

How Glint's AI classifies signal severity and what each level means for your trading.

Impact Levels

Level	Meaning	Typical Response
Critical	Major breaking event with immediate market impact	Act within seconds
High	Significant development with clear implications	Evaluate within minutes
Medium	Notable signal worth monitoring	Watch and wait
Low	Minor development, background noise	Ignore unless relevant

How Impact Is Calculated

Glint's AI considers several factors when scoring impact:

1. Source reliability — Higher-tier sources get higher base scores

2. Content severity — Keywords and phrases indicating urgency (e.g., "BREAKING", "emergency", "immediately")

3. Entity significance — Signals involving major entities (heads of state, central banks, major companies) score higher

4. Historical precedent — Similar past signals and their actual market impact inform the model

5. Market sensitivity — Some market categories are more volatile than others

Impact Distribution

In a typical day, Glint processes thousands of signals. The distribution roughly follows:

• Critical: 1-5% of signals — These are rare and almost always actionable

• High: 10-15% of signals — Worth reviewing, many are tradeable

• Medium: 25-35% of signals — Good for context, occasionally tradeable

• Low: 50-60% of signals — Filtered out by default in most views

Recommended Filter Settings

Trading Style	Filters
Active scalper	Critical + High
Swing trader	Critical + High + Medium
Researcher	All levels
Alert-only	Critical only

False Positives

No AI system is perfect. Occasionally a signal may be scored higher or lower than warranted. Common false positive scenarios:

• Engagement farming — Sensational tweets from accounts with large followings can trigger High impact despite low substance

• Satire/sarcasm — The AI may misinterpret satirical content as genuine breaking news

• Duplicate signals — The same event reported by multiple sources can create the appearance of higher impact

Glint continuously refines its classification pipeline to minimize these issues. If you notice a misclassified signal, feedback helps improve the system.