Analyze Opponents With Proper Samples
Learning objectiveRecord tendencies without overfitting or misattributing results.
Understand the idea
Track each opponent separately. Record each observed opportunity and action: a fold to a three-bet requires a genuine three-bet faced; a continuation bet requires a chance to bet after prior aggression. Use the public hand history and legitimate showdowns only. Head-to-head chip attribution is difficult in multiway pots; do not assign all gains to one opponent without evidence.
Use sample thresholds and confidence language. Two folds in two opportunities do not prove a player always folds. Combine small samples with a conservative prior or leave the tendency undecided. Keep identities and seasons distinct so a renamed or different bot is not silently merged.
At the poker table
Orion folds once when facing a river bet. Mark 1/1 observed, not '100% reliable river fold.'
Common mistake
Do not infer an opponent's hidden cards from a fold, or let a tiny sample drive a large bet-size change.
Try it yourself
Record one opponent's actions across ten hands and list which metrics have enough actual opportunities to calculate.