Can math predict who stays together?

Partially. Data science can identify probability factors (those with secure attachment have X% higher satisfaction; those showing contempt have X% higher breakup rate). But individual outcomes vary. Two couples with identical profiles might have very different futures based on choices, therapy, and luck.

What's the most predictive compatibility variable?

According to research: attachment style (single strongest predictor), followed by conflict resolution behavior (Four Horsemen or Golden Behaviors), then values alignment. Personality similarity ranks much lower. Most algorithms that claim 90%+ accuracy are overfitting—they work on their training data, fail on new couples.

Will AI eventually match us perfectly?

Unlikely. Relationships involve free will, choice, and human unpredictability that algorithms can't capture. An AI might predict "this match has 78% success probability," but that 22% chance covers people who choose growth, therapy, or sacrifice to stay together. Love isn't pure math.