Poker Shark
Poker Statistics & Research
Original data from 33 million live poker hands. Population-level poker statistics showing what players actually do — not what solvers say they should.
Training-data methodology
How Poker Shark turns hand histories into training opponents
The training foundation contains 33 million live poker hands. Hands are aggregated from publicly available records and licensed data providers, then analyzed as population behavior rather than personal player profiles.
- 01
Source and normalize
Aggregate cash-game hand histories from publicly available records and licensed data providers, then normalize positions, actions, streets, and bet sizes into one comparable format.
- 02
Measure behavior
Calculate repeated tendencies such as VPIP, PFR, aggression, continuation betting, folding under pressure, sizing, and river behavior across meaningful samples.
- 03
Build explainable profiles
Group statistically distinct patterns into opponent archetypes whose decisions remain tied to visible tendencies rather than generated personality labels.
- 04
Validate the output
Compare simulated decisions and aggregate profile stats with the target ranges, then audit where behavior diverges by position, street, and board texture.
Scope and Limitations
Published studies disclose their own eligible subset and definitions. The c-bet study uses 21 million hands from the larger foundation because only hands meeting its continuation-bet criteria enter that analysis.
Population statistics are starting priors, not predictions about one person. Online and live environments differ, and an opponent can change during a session. The trainer therefore exposes confidence and repeated evidence so players learn when to trust a read and when to update it.
Player data handling is described in the data policy.
See this data in action
Train against opponents whose tendencies are built from these exact hand histories.