A Proof of Concept for predicting Stake.com game outcomes (Mines & Coinflip) using a TensorFlow.js
model and provably fair seed pattern analysis.
⚠ Educational Use Only
This does not break cryptography. Results are guesses from pattern recognition and not reliable for gambling. You will likely lose money if used for real betting.
Requirements:
- Node.js 18+
- npm
# 1. Clone repo
git clone https://github.com/your-username/stake-predictor-poc.git
cd stake-predictor-poc
# 2. Install dependencies
npm install
🔮 Usage When you run the predictor, it will ask you for seeds via CLI: Client Seed → Get from Stake (Game → Fairness → Copy) Server Seed Hash → Same place as above Nonce → The number of bets you've made with current seed pair (next bet’s nonce)
Example: Mines Prediction
node index.js
Select Game: 1 ( 1 for mines 2 for coinflip )
Enter Client Seed: Your_Active_Client_Seed
Enter Server Seed Hash: Hash_Of_The_Current_Server_Seed
Enter Nonce: 0
[ ✓ ] [ ✓ ] [ X ] [ ✓ ] [ ✓ ]
...
Recommendation: Pick ✓, avoid X
Example: Coinflip Prediction
node index.js
Select Game: 2 ( 1 for mine 2 for coinflip )
Enter Client Seed: Your_Active_Client_Seed
Enter Server Seed Hash: Hash_Of_The_Current_Server_Seed
Enter Nonce: 0
Nonce 0: Heads (82%)
Nonce 1: Heads (71%)
...
Example:
```python-repl
Nonce 0: Heads (82%)
Nonce 1: Heads (71%)
🛠 Tech
- Node.js
- TensorFlow.js
- crypto (HMAC-SHA256)
📌 Contribute Fork → Improve → PR. Ideas:
-
Larger training dataset
-
Add more games (Limbo, Plinko)
-
Better visualizations
-
📜 License MIT — see LICENSE.