BC.Game Crash AI Strategy

The project centers on building an AI-driven pattern-finding engine for the Crash game on BC.Game. The model’s sole objective is to analyze historical multipliers, detect statistically significant behaviors, and surface insights that can later inform strategy design. Data pipeline You will create an automated scraper or API routine that continuously gathers Crash game multipliers directly from the BC.Game website. No offline files will be supplied, so robust web-data handling, cleaning, and storage (CSV, SQL, or parquet) forms the foundation of the work. Modeling approach Once data is captured, the next phase is exploratory analysis followed by model development. I’m open to classic time-series methods, deep learning approaches with TensorFlow or PyTorch, or a hybrid ensemble—whichever best extracts repeatable patterns. Clear commentary on feature engineering, train/validation splits, and overfitting prevention is essential. Deliverables • Codebase with scraper, preprocessing routines, and the trained model • A concise report (not more than 5 pages) explaining methodology, key findings, and reproducibility steps • Demonstration notebook or script that loads the latest live data and outputs pattern metrics or visualizations in real time Acceptance criteria The final notebook must execute end-to-end on my machine, regenerate the model (or load saved weights), and produce pattern summaries within reasonable runtime. All external libraries should be listed with exact versions in a requirements file. Showcase any prior AI model building—particularly where noisy, stochastic data was involved—so I can gauge fit quickly. Budget: USD 30–250