The Challenge
Professional poker players need to compute Nash equilibrium strategies across millions of possible game states. Existing tools required expensive local hardware, large databases, and hours of computation — making them impractical for real-time study and in-session use.
The Chartopus team needed a cloud-native solver that could deliver institutional-quality game theory computations instantly, without any local setup.
The Solution
1D.works designed and built the core technology stack behind Chartopus — a proprietary cloud neural network solver that computes game-theoretically optimal (GTO) poker strategies in real time.
Key technical components:
- Counter Factual Regret (CFR) Algorithms: Custom implementation of state-of-the-art CFR solvers that converge on Nash equilibrium strategies across complex decision trees.
- Neural Network Approximation: Multiple neural networks trained to approximate solver outputs, enabling near-instant lookups for preflop and postflop spots without re-solving from scratch.
- Cloud-First Architecture: Fully hosted on Google Cloud and Azure — no database, server, or local setup required by end users. All computation runs server-side.
- Three Integrated Products: The solver powers a Chart Viewer (strategy lookup), Trainer (practice against AI profiles), and Analyzer (hand history review with GTO feedback).
Why This Matters
This work was completed years before AI became mainstream in the poker and gaming industry. The same techniques — game tree search, regret minimization, neural network function approximation — are now foundational to modern AI research (cf. DeepMind's work on imperfect information games).
Building Chartopus required solving problems that directly transfer to finance and trading:
- Decision-making under uncertainty with incomplete information
- Real-time inference from pre-trained neural network models
- Scalable cloud compute for computationally intensive workloads
- Proprietary model IP protection in a competitive market
The Result
- 30+ years of combined domain expertise encoded into algorithmic strategy
- Cloud SaaS platform serving professional poker players globally
- Sub-second strategy lookups across millions of game states
- Zero local infrastructure required — browser-only access
Explore our Institutional Trading Infrastructure service or read about Infrastructure is Alpha.