Pokerbros Winning AI Bot Development
Budget: £250 – £750 GBP
I need a robust A-I solution that can sit in my PokerBros games and play to win consistently. The target performance is expert-level: it should recognise table states, size bets optimally, vary its timing so it feels natural to opponents, and steadily grow the bankroll over the long run.
To achieve that edge, I want a hybrid approach—hard-coded ranges and exploitative rules for common spots, backed by a machine-learning module that keeps adapting to opponents’ patterns as hands accumulate. Pre-flop charts, ICM awareness, GTO post-flop lines, and real-time opponent modelling all have to work together without slowing decision time beyond the site’s action clock.
Key expectations
• Runs reliably on Windows with whichever Android emulator best suits PokerBros (LDPlayer, BlueStacks, etc.).
• Captures hole cards, board, stack sizes and betting history via screen reading or any stable hook you prefer, converts them to structured data, and feeds the decision engine.
• Produces bet, fold or raise outputs that an input-automation layer can click back to the table.
• Includes a configuration panel where I can tweak rule parameters, turn the learning module on/off, and review performance logs.
• Ships with a pre-trained model plus the training pipeline so I can continue refining it on new hand histories.
I’ll sign off the project once the bot:
1. Beats a representative pool over a 20-k hand trial with a positive BB/100, and
2. Maintains an average response time under the site deadline while remaining undetected.
Source code, model weights, installation guide and a quick demo video make up the final deliverable.
To achieve that edge, I want a hybrid approach—hard-coded ranges and exploitative rules for common spots, backed by a machine-learning module that keeps adapting to opponents’ patterns as hands accumulate. Pre-flop charts, ICM awareness, GTO post-flop lines, and real-time opponent modelling all have to work together without slowing decision time beyond the site’s action clock.
Key expectations
• Runs reliably on Windows with whichever Android emulator best suits PokerBros (LDPlayer, BlueStacks, etc.).
• Captures hole cards, board, stack sizes and betting history via screen reading or any stable hook you prefer, converts them to structured data, and feeds the decision engine.
• Produces bet, fold or raise outputs that an input-automation layer can click back to the table.
• Includes a configuration panel where I can tweak rule parameters, turn the learning module on/off, and review performance logs.
• Ships with a pre-trained model plus the training pipeline so I can continue refining it on new hand histories.
I’ll sign off the project once the bot:
1. Beats a representative pool over a 20-k hand trial with a positive BB/100, and
2. Maintains an average response time under the site deadline while remaining undetected.
Source code, model weights, installation guide and a quick demo video make up the final deliverable.