Rahul S. P. — Quantitative Researcher

Quantitative Research

Research on market microstructure, neural architectures for trading, and cross-asset dynamics. All studies are empirical, tested on live data, and grounded in statistical rigor.

Empirical Studies

5

Experimental results validated on live market data with walk-forward testing.

Entry Speed vs Confirmation Quality in Tick-Level Scalping

We study the trade-off between entry speed and confirmation quality across 21,000 scalping signals over 90 days. Using 42.9 million ticks of XAUUSD data, we show that the edge in consecutive-bar reversal signals is maximal at the exact instant price crosses last_close and decays rapidly. Pending STOP orders at the break level outperform all confirmation-based entries by $12,646, with a profit factor of 1.59 vs 1.34. Every post-hoc sustain filter tested produced negative returns at all parameter combinations.

Read paper →

XAG Directional Disagreement as a Cross-Asset Lot Scaling Signal

We show that directional disagreement between XAUUSD and XAGUSD over a 20-bar window is the strongest single predictor of scalping signal quality, with Spearman rho between -0.23 and -0.29 (p approximately 0). Lower disagreement implies stronger co-movement and higher reversal reliability. We design a four-tier lot scaling system based on this metric, with the top tier (disagreement <= 8 plus XAG bar reversal) receiving 1.5x allocation.

Read paper →

Architecture & Models

1

Neural network architectures designed for financial time series.

GoldSSM: Selective State Space Models for Gold Price Forecasting

We present GoldSSM, a selective state space model for intraday gold price direction forecasting. The architecture combines a Variable Selection Network, a stack of Mamba blocks with selective scan, and temporal attention pooling. At 2.0M parameters, GoldSSM serves as a drop-in replacement for Transformer-based models with identical forward signatures, while offering linear-time sequence processing and improved handling of long-range dependencies in financial time series.

Read paper →

Feature Engineering

2

Construction, selection, and validation of predictive features.

Alpha101 on Intraday Gold: Why Most Equity Factors Fail

We evaluate all 101 formulaic alpha factors from Kakushadze (2016) on intraday XAUUSD data. Only 4 of 101 factors achieve AUC above 0.515 for direction prediction, and only two (alpha024 and alpha083) survive forward selection. The failure mode is structural: Alpha101 factors exploit cross-sectional dispersion across a stock universe, a mechanism that does not exist for a single instrument. We document which factor families fail and why.

Read paper →

107 Features for Gold: Building an Institutional Feature Pipeline

We describe the design and validation of a 107-feature pipeline for intraday gold trading. The pipeline spans six feature groups: price dynamics, cross-asset signals, volatility regimes, microstructure proxies, temporal patterns, and statistical complexity measures. We detail the engineering choices behind each group, the cache invalidation strategy, and the empirical AUC contribution of each feature family. The pipeline supports both batch backtesting and live execution with sub-second latency.

Read paper →