a9876543245/DEEPFOLD-SOLVER

What it solves

DEEPFOLD-SOLVER is a Game Theory Optimal (GTO) poker solver designed to calculate equilibrium strategies for poker hands. It eliminates the need for manual trial-and-error by providing mathematically precise strategies, while solving the common performance bottlenecks of solver software through GPU acceleration and efficient multi-core CPU utilization.

How it works

The project uses a Discounted Counterfactual Regret Minimization (DCFR) engine. It features a dual-backend system that automatically detects hardware: it uses native CUDA kernels for NVIDIA GPUs (Turing, Ampere, Ada, Hopper, and Blackwell architectures) or falls back to a multi-core CPU backend using AVX2 kernels for modern processors. To ensure accuracy, the software employs a "parity gate" that verifies the fast GPU/AVX2 paths produce bit-identical results to a slower, correct reference oracle.

Who it’s for

It is built for serious poker players and students of the game who want to analyze hand histories, study GTO strategies, and train their decision-making using a professional-grade solver on Windows 10/11.

Highlights

• Hybrid Compute Engine: Seamlessly switches between GPU (CUDA) and CPU (AVX2/Scalar) backends with linear scaling across all available CPU cores.
• Advanced Post-Solve Analysis: Includes a Runout Report to group turn cards by texture (e.g., Flush, Straight) and a 1326 Combo Drill for per-combo blocker analysis.
• Visual Insight Tools: Heatmaps for Expected Value (EV) and aggression frequencies, alongside a strategy grid for action mixes.
• Bespoke Memory Management: Pre-defined memory profiles (safe, balanced, performance) to prevent out-of-memory (OOM) crashes.
• Integrated Content: Comes with a library of over 2,550 preflop scenarios and 120+ pre-solved flop spots.
• Training Mode: Includes 10-question drills to test user decisions against the calculated equilibrium.