Research
A backtested hybrid trading system combining technical indicators, FinBERT sentiment and XGBoost claims a 135% two-year return beating the S&P 500, the kind of result that is easy to publish and hard to trust
The preprint, accepted for the ComSIA 2026 conference, layers trend and mean-reversion signals with machine-learning trade calls and a volatility-based regime filter, reporting a final portfolio value of $235,492.83 over 24 months against a starting stake; the number is real, the question is whether the backtest methodology behind it would survive contact with live capital and an untested market regime.
The system design is unremarkable in isolation. EMA and MACD trend detection, RSI and Bollinger Band mean-reversion, FinBERT sentiment scoring and XGBoost signal generation are all individually well-established techniques. What the paper actually tests is whether stacking them, with a volatility-based regime switch reweighting the signals depending on detected market conditions, beats running any single approach alone. On the reported numbers, the combination does.
The figure that deserves scrutiny before admiration is the 135.49% return itself. A 24-month backtest covering 2024 to 2026 lived through one of the most AI-driven bull markets in market history. Any regime-adaptive system tuned on that window has effectively been trained to recognise and ride an AI rally. Whether these architectures generalise to a regime the training data never saw is the open question this site keeps returning to, and a paper reporting strong results entirely within a single directional macro period cannot answer it by construction.
This sits next to a pattern documented elsewhere on this site: the academic literature broadly still provides little clarity on the underlying predictive structures that drive market reactions, even as individual papers report strong backtested returns. That gap, strong point-in-time results without a clear structural explanation for why they should persist, is exactly the failure mode that turns a promising backtest into a live-capital disappointment.
Watch for a walk-forward or out-of-sample test spanning a different macro regime, ideally one including a genuine drawdown rather than a mostly-rising market, before treating a 135% two-year figure as evidence the architecture itself is sound rather than evidence the underlying market regime happened to favour trend-following, sentiment-augmented systems.
Read the original: arXiv preprint - A backtested hybrid trading system combining technical indicators, FinBERT sentiment and XGBoost claims a 135% two-year return beating the S&P 500, the kind of result that is easy to publish and hard to trust. Commentary is the independent editorial view of Share Trading; the original article is credited to its publisher.