Evaluates every legal arrangement in milliseconds, ranked by expected value, and returns the recommended front / middle / back split plus alternatives.

Automatically enforces the back ≥ middle ≥ front ordering, eliminating mis-sets (fouls) that forfeit the whole round.

Detects and weighs special hands — Dragon, three flushes, three straights, six pairs, all-one-color, all-high / all-low — trading off 'go special' vs. a normal split.

Computes the probability of a sweep (shoot), being swept, and a home run (winning all three), balancing a safe sweep against contesting all rows.

Performs range inference on each opponent, optimizing for total EV across the whole table rather than beating just one player.

Deep learning + search give every arrangement an explainable expected return and win/lose distribution — so every point is accounted for.

Recognizes your 13 cards and the current situation (player count, rules).

Deep-learning model + real-time search evaluate the EV of every legal arrangement in parallel.

Outputs the EV-optimal arrangement and alternatives, flagging sweep probability and special-hand tips.