A strong hand and a high percentage are different facts
Poker presents several useful numbers, but they do not all answer the same question. Understanding the label and assumptions matters more than comparing percentages without context.
Explore poker educationMade hand: what your known cards form
A made-hand category such as pair, straight or flush describes the cards already available. In Hold’em, the best five-card hand is chosen from your hole cards and the board. Other variants have different card-use rules.
A pair can be the best hand now and still lose after later cards. A made-hand label therefore describes a current fact, not the final result.
Heads-up win rate: a model of unseen cards
The dedicated Hold’em insight calculation compares against one uniformly random unseen opponent holding. On earlier streets it also samples remaining board cards. On the river it can count all possible unseen opponent pairs.
The panel’s win rate counts outright wins. Ties are tracked separately in the calculation and are not automatically equivalent to a full win. This number does not model the range suggested by a particular opponent’s betting.
Pot share: include split outcomes
Expected pot share can award a fraction when a pot is split. For example, winning half a pot is different from winning all of it. Public all-in share and supported multiway estimates must be read using their own labels, opponent counts and assumptions.
A sampled multiway single-pot share does not establish a prediction for side pots, future folds or an opponent’s actual range. A heads-up number should not be copied into a multiway situation.
Historical observations: what happened in a sample
Where an empirical summary is shown, its observed-hand count and matching starting-card context describe completed observations. A historical mean pot share is not the current hand’s win probability.
Ask the assistant which measure it is explaining. Good questions include “Does this include ties?” and “What opponent assumption is used?”. Neither a strong category nor a model estimate guarantees the next result.