Keep separate Candidate and Confirmed pattern lists — promote only after validation
Maintain two separate lists—'Pattern Candidates' and 'Confirmed Patterns'—promoting candidates only after they survive three independent observations, one alternative-explanation check, and one successful prediction.
Why This Is a Rule
Without a staging system, perceived patterns jump straight from "I noticed something" to "this is how things work" — skipping validation entirely. The two-list system creates a holding area (Pattern Candidates) where suspected patterns accumulate evidence before being promoted to the operational list (Confirmed Patterns) that you actually act on.
The promotion criteria prevent premature pattern commitment: three independent observations (the pattern recurred in different contexts, not just the context where you first noticed it), one alternative-explanation check (you generated at least one plausible alternative and couldn't rule it out immediately), and one successful prediction (you used the pattern to predict an outcome before it happened, and the prediction was correct).
The prediction criterion is the strongest filter. Patterns that explain past observations are easy to construct — your brain can always find a narrative that fits historical data. Patterns that successfully predict future outcomes are genuinely useful. A pattern that predicted once is more trustworthy than one that explained ten times retrospectively.
When This Fires
- When you notice a potential pattern and want to track it without committing to it
- During periodic reviews of your pattern dictionary
- When deciding whether to act on a perceived pattern
- Any time you're building a personal pattern recognition practice
Common Failure Mode
Treating the Candidate list as the Confirmed list — acting on unvalidated patterns because they feel right. The two lists exist specifically to prevent this. Candidates are hypotheses; Confirmed patterns are validated tools. Only Confirmed patterns should inform decisions.
The Protocol
(1) When you notice a potential pattern: add it to Pattern Candidates with the date and first observation. (2) As new observations occur, log them against the candidate. (3) Promotion checklist: three independent observations? ✓ Alternative explanation generated and tested? ✓ Successful prediction made? ✓ (4) All three criteria met → promote to Confirmed Patterns. (5) Any criterion failed → keep in Candidates and continue gathering evidence.