Testing Your Patterns

Before trusting a pattern you have noticed, put it through these filters:

• Sample size — how many examples support the pattern? Three is not enough.
• Base rate — how common is this event regardless of the pattern? (Bad things "in threes" ignores all the non-groups)
• Feedback loop — have your pattern predictions been tested and confirmed?
• Alternative explanations — could randomness, selective memory, or confirmation bias explain the pattern?
• Prediction — if the pattern is real, what should happen next? Check.