Price-Momentum Divergence Mapping with a Temporal Attention Proxy

Summary

The article proposes a trading signal that compares normalized price slope with an indicator slope over a lookback window, then treats their difference as structural divergence. It combines this measure with a three-horizon attention proxy intended to weight recent, intermediate, and older signals. Positive and negative divergence are scored separately, and a directional signal requires one side to exceed a threshold while the other remains below it. RSI and DeMarker are discussed as complementary momentum inputs, and the implementation is framed as a customizable MQL5 trading robot.

The reported forward walk test covers one forex pair, GBPJPY, on a two-hour timeframe over four months. Although the approach is presented as a pairing with a Temporal Fusion Transformer, the optimized settings did not use the attention proxy; the divergence calculation alone generated the reported 268 trades, $2,600 net profit, and drawdown slightly above 8%. The author calls for testing on more symbols and longer windows, and suggests more complete neural network integration as future work. The limited test does not establish general or live trading performance.

Key ideas

• The divergence engine compares normalized price and indicator slopes over a fixed lookback window.
• A three-horizon attention proxy weights recent, intermediate, and older divergence signals.
• Buy and sell signals use separate directional scores and a threshold with a mutual exclusion condition.
• In the reported GBPJPY test, optimized settings relied on divergence alone rather than the attention proxy.
• The single pair and limited forward test call for broader symbol and time period evaluation.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.