How PokerIQ works out the numbers

Equity is your share of the pot by the river. When every remaining board can be counted, PokerIQ counts them all and marks the figure exact. When there are too many, for example against several opponents before the flop, it samples random deals and shows a 95% margin, marked simulated. Saved hands and the AI analyst use 200,000 deals with a fixed seed, so the same spot always gives the same figure.

Against a player type, villain’s hand is drawn from that type’s range for what they did before the flop, and the figure is marked as being against an assumed range.

Outs are the next cards that put you ahead. PokerIQ counts them only against the hands that are currently beating you, so a card that only improves you against hands you already beat isn’t counted as an out.

The equity a call needs is the call divided by what you would win after rake. PokerIQ’s default rake is 5% of the pot, capped at 3 big blinds, with no rake when a hand ends before the flop. Amounts are in big blinds throughout.

Five player types describe the opponent: Unknown (the default when you have no read), Regular, Loose-passive, Nit and Maniac. Each has preflop ranges for the lines PokerIQ offers, such as raising first, calling or defending the big blind, and a curve saying how often it keeps going against each bet size.

When you bet, the type keeps going with the strongest part of its range, with each hand judged by its equity against a random hand, as a player sees their own cards. Every range is on show in the Range Explorer.

Each bet or raise size is valued from one split of the range: how often villain folds, your equity when called, and the expected value. These figures are estimates, because they rest on the type’s range and fold habits, and PokerIQ lists those assumptions beside them.

With several opponents, PokerIQ values bets when it’s checked to you, treating each opponent as holding the same range and deciding on their own. Raises into several players aren’t valued yet.

The action comes from fixed rules with test cases, never from the AI. Facing a bet, the rules compare calling with folding using your equity and the price, and allow a draw to call when what it needs to win later stays within your style’s limit: 15% of the stack behind for Tight, 25% for Balanced and 40% for Aggressive. Spots within 3 points of the equity needed are labelled close.

Where bets can be valued, PokerIQ bets or raises when the best size beats checking or calling by your style’s margin: 10% of the pot for Tight, 5% for Balanced and 2.5% for Aggressive. Sizes bigger than the pot are only considered when the stack is no more than three times the pot. Each recommendation names the rule that chose it.

Google’s Gemini writes the explanation, never the numbers. It places markers where figures go, and PokerIQ fills them in from the engine with their certainty badges. Any other figure it writes is withheld, and a sentence that argues for a different action from the engine’s is removed. If an answer breaks these rules badly, it is written again once; after that, the offending sentences are dropped.

• exactCounted over every remaining card or board, with no sampling.
• simulatedToo many boards to count, so PokerIQ samples random deals and shows a 95% margin.
• estimateRests on assumptions, such as a player type’s range or how often it folds, and lists them.
• vs assumed rangeWorked out against a range you chose for villain, not against a known hand.

More than 400 automated tests run on every change: the engine’s accuracy and speed against known results, every rule of the recommendation, the checks on AI answers, and the database rules that keep each account’s hands private.