🃏 GTO3 — PokerHack 2025 Winner

Hackathon-winning poker bot applying Game Theory Optimal (GTO) strategy via probability-based decision trees. Built for PokerHack 2025 hosted by HackMelbourne. Competed against bots from all participating teams and won.

🏆 Winner — PokerHack 2025 (HackMelbourne) Devpost submission

---

Strategy

The bot implements a layered decision framework combining hand strength evaluation, pot odds calculation, positional awareness, and controlled probabilistic bluffing:

• Preflop — evaluates hand strength (pairs, suited connectors, high-card equity) and adjusts opening range by position
• Postflop — reads board texture and hand potential (made hands vs draws) to determine bet sizing
• Position awareness — plays tighter from early position, exploits late-position advantage
• Pot odds — folds draws where EV is negative, calls and raises where pot odds justify it
• Aggression scaling — increases aggression as the round progresses to pressure passive opponents
• Probabilistic bluffing — tuned bluff frequency to remain unexploitable without being too passive

---

Why it won

Most bots in the competition played reactive strategies — call or fold based on hand strength alone. GTO3 won by combining position awareness with pot odds calculation, two factors most competitors ignored. The probabilistic bluffing layer made it impossible for opponent bots to build a static counter-strategy.

---

Running the bot

Install

pip install -r requirements.txt

Configure opponents in poker_conf.yaml

ai_players:
  - name: Opponent1
    path: sample_player/fish_player_setup.py
  - name: Opponent2
    path: sample_player/random_player_setup.py
  - name: GTO3
    path: submission/Team-Bots.py
initial_stack: 100
max_round: 10
small_blind: 10

Start the server

python -m pypokergui serve ./poker_conf.yaml --port 8000 --speed moderate

Open the browser tab → click Start Poker → watch it play.

If you get a port error, change --port 8000 to --port 8001

---

Built with

• PyPokerEngine — game engine and simulation framework
• Python probability utilities — hand equity calculation
• Custom decision tree — position + pot odds + bluff frequency logic