Análisis de agentes automatizados de trading aplicados al mercado del oro mediante backtesting en tradingview

Abstract

This paper analyzes the performance of five automated trading agents applied to the gold market (XAU/USD) through backtesting on TradingView, from a systems engineering perspective. The agents, representing trend-following, mean-reversion, volatility breakout, momentum, and sideways-range strategies, were implemented in Pine Script (v6) with predefined parameters and a common risk management layer based on the ATR, and were run under uniform conditions (capital of USD 10,000, commission of 0.02%, slippage of 5 ticks) over a period of approximately five and a half months using a 15-minute timeframe. No agent was profitable during the period (profit factors ranging from 0.634 to 0.926), with the MACD agent being the best relative performer, though its advantage was largely offset by transaction costs. However, all five agents executed without errors throughout the entire window, maintained drawdowns between 2.35% and 3.26%, and produced traceable metrics, demonstrating execution consistent with the programmed rules. The main limitation is the length of the simulation window, which is confined to a single market regime (a high-volatility bearish correction following gold’s all-time high in January 2026), thereby limiting the generalizability of the results. The main contribution is a documented and replicable procedure for evaluating trading agents as software artifacts in simulated environments.

Keywords: automated trading agents, algorithmic trading, back testing, XAU/USD, software quality.