A bot that plays poker while you pretend to work
An open-source computer vision pokerbot that scrapes PartyPoker, PokerStars, and GGPoker tables via screenshot analysis, then bets using Monte Carlo equity simulation and tweakable strategies.

What it does
DeeperMind is a Windows-only pokerbot that watches your screen, reads cards and buttons via OpenCV image recognition, and physically moves your mouse to click bets, folds, or calls. It targets “Zoom” and “Fast Forward” tables on the three major poker sites. You can run pre-built strategies or edit your own in a Qt-based GUI, then analyze results with stacked bar charts showing which decisions won or lost money at each betting stage.
The interesting bit
The project sits in an awkward limbo between open source and commercial product. The repo builds from source, but the README funnels you toward paid binaries at a separate website; full strategy editing requires a password you buy with Bitcoin or earn by contributing table mappings. Meanwhile the actual decision engine is surprisingly transparent—Monte Carlo equity calculation, genetic algorithm tuning, and a strategy editor where you drag curves to set minimum call thresholds.
Key highlights
- Table-agnostic scraper: teach new poker clients by marking UI regions in a visual editor, or train a neural network with data augmentation
- Monte Carlo equity engine in both slow Python and faster-but-broken NumPy variants
- Strategy analyzer breaks down profitability by preflop/flop/turn/river and individual hands
- Docker compose setup exists, though the bot only officially supports Windows
- VirtualBox integration so the bot controls a VM’s mouse without hijacking your main desktop
Caveats
- README warns explicitly: don’t test with real money until you’ve verified correct functioning, or you risk losing your account and bankroll
- The fast NumPy Monte Carlo implementation “is not yet working correctly” with failing tests
- Windows 10+, 4GB RAM, 4-core CPU, and 1920x1080 resolution are listed as minimums; anything less is unverified
- Some tests are outdated and the authors acknowledge more are needed
Verdict
Worth a look if you’re researching game-playing bots, computer vision for legacy desktop apps, or the grey-market economics of poker automation. Avoid if you want something that just works out of the box, or if you value your gambling accounts.