The live solver utilizes Pure MCCFR with depth-limited solving to compute local strategies for the current public state (board + action history). By using reach solving and accounting for the "gift" (EV already conceded by prior deviations), the solver avoids overspecializing to a fixed opponent range, keeping real-time refinements stable and non-exploitable.

Subgame solving is anchored to a blueprint generated via Discounted CFR. Following the bias approach, the blueprint is perturbed into a small set of intelligent strategies, which serves as the depth-limit continuation we evaluate against. To maintain a compact storage footprint, the blueprint employs robust, sparse betting trees designed to minimize exploitation when play deviates from the primary abstractions.

In Progress: Extending strategic stability logic to 3+ player environments. This research focuses on refining equilibrium nodes in multi-way pots, mitigating the strategic instability often triggered by multi-agent dynamics and irrational opponent play.