Our platform is rooted in rigorous academic research, utilizing advanced quantitative methods and statistical validation to process cryptocurrency market microstructure.
We isolate and quantify the predictive power of each data layer independently. By removing features layer by layer, we statistically prove the value added by order book and derivatives metrics over pure spot price data.
Initially conceived as an HSE Thesis project, the pipeline incorporates formal hypothesis testing, out-of-sample cross-validation, and peer-reviewed deep learning architectures including causal GRUs with attention blocks.
45 distinct features mathematically derived from Level 2 order books, continuous futures contracts, and macro indicators, rigorously normalized to prevent look-ahead bias and non-stationarity issues.