Discussing the article: "Nonlinear OHLC Features Derived from Elliptic Curves"

[MetaQuotes]

Check out the new article: Nonlinear OHLC Features Derived from Elliptic Curves.

The article discusses the projection of EURUSD daily candlesticks onto the secp256k1 elliptic curve and the extraction of 96 features (EC+TA) to predict the direction of the next candlestick in CatBoost. The price mapping to the curve and the training pipeline on 2,000 D1 bars are shown; the full model achieves a test AUC of 0.6508, with EC features contributing 60.6%. The materials are suitable for reproduction in Python/MetaTrader 5.

You are building a model to predict the direction of the next daily EURUSD candlestick and face a practical problem: the classic set of TA features (returns, MA/RSI, volatility, candlestick proportions) yields limited improvement, while attempts to make it more complex lead to overfitting and poor portability across market regimes. A new feature source is needed, deterministically computed from OHLC data, that provides a nonlinear representation of price, integrates easily into the existing ML pipeline, and is validated by a fair temporal split using AUC.

In this article, we address exactly this task. The method is simple to formulate: we map OHLC values to points on the secp256k1 elliptic curve using SHA-256 + try-and-increment, extract metadata from the points and the mapping process (nonce, x bits, Open+Close/High+Low group arithmetic), combine them with classic TA features, and test their added value in CatBoost on a time-based split. An experiment on 2,000 EURUSD D1 candlesticks yields a measurable result (test AUC = 0.6508), while the signal turns out to be regime-dependent—the key practical conclusion of the article.

Author: Yevgeniy Koshtenko