Paper: arXiv 2412.15448
Authors: Akash Deep, Abootaleb Shirvani, Chris Monico, Svetlozar Rachev, Frank J. Fabozzi
Abstract
Because of the theoretical challenges posed by the Efficient Market Hypothesis to technical analysis, the effectiveness of technical indicators in high-frequency trading remains inadequately explored, particularly at the minute-level frequency, where effects of the microstructure of the market dominate. This study evaluates the integration of traditional technical indicators with random forest regression models using minute-level SPY data, analyzing 13 distinct model configurations. Our empirical results reveal a stark contrast between in-sample and out-of-sample performance, with \(R^2\) values deteriorating from 0.749–0.812 during training to negative values in testing. A feature importance analysis demonstrates that primary price-based features dominate the predictions made by the model, accounting for over 60% of the importance, while established technical indicators, such as RSI and Bollinger Bands, account for only 14%–15%. Although the indicator-enhanced models achieved superior risk-adjusted metrics, with Rachev ratios between 0.919 and 0.961, they consistently underperformed a simple buy-and-hold strategy, generating returns ranging from -2.4% to -3.9%. These findings challenge conventional assumptions about the usefulness of technical indicators in algorithmic trading, suggesting that in high-frequency contexts, they may be more relevant to risk management rather than to predicting returns. For practitioners and researchers, our findings indicate that successful high-frequency trading strategies should focus on adaptive feature selection and regime-specific modeling rather than relying on traditional technical indicators, as well as indicating the critical importance of robust out-of-sample testing in the development of a model.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 6.0/10 (capped from 8.0: transaction costs not modelled; see penalty flags)
- Quadrant: Street Traders — empirical and applied, lower math complexity (effective empirical rigor capped from 8.0)
Why this score: The paper relies on standard statistical metrics and machine learning implementation details (e.g., random forest regression on minute-level data) rather than advanced mathematical derivations, yet it is heavily data-driven, featuring specific backtest results, out-of-sample testing, and risk-adjusted performance metrics like the Rachev ratio.
Research Flowchart
flowchart TD
subgraph Goal
A["Research Goal<br>Evaluate RF integration with technical indicators<br>for minute-level HFT performance"]
end
subgraph Methodology
B["Data & Configuration<br>Minute-level SPY data<br>13 Random Forest model configurations"]
C["Computational Process<br>In-Sample Training vs Out-of-Sample Testing<br>Feature Importance Analysis<br>Risk-Adjusted Metric Calculation"]
end
subgraph Outcomes
D["Performance Results<br>R^2 drop 0.749-0.812 (Train) to Negative (Test)<br>Returns: -2.4% to -3.9%"]
E["Feature Analysis<br>Price-based features: >60% importance<br>Technical indicators (RSI/Bollinger): 14-15%"]
F["Key Findings<br>Technical indicators relate to risk management<br>Focus on adaptive feature selection & regime-specific modeling"]
end
A --> B
B --> C
C --> D
C --> E
D --> F
E --> F
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes integrating traditional technical indicators with random forest regression models for minute‑level high‑frequency stock price prediction.
Source sentence
This study evaluates the integration of traditional technical indicators with random forest regression models using minute-level SPY data
Assumes 6
- Trades are executed at minute‑end closing prices without accounting for bid‑ask spreads, commissions, or market impact. execution
Source sentence
All trades are executed at minute-end closing prices. While this implementation provides realistic control of the sizes of the positions, it is somewhat optimistic, as it does not take into account bid–ask spreads, commission costs, or potential price impact.
- A per‑minute turnover constraint of 0.4% of portfolio value is imposed to reflect realistic liquidity limits. execution
Source sentence
To approximate real‑world trading constraints, we impose a turnover constraint limiting position changes in any minute to 0.4% of the portfolio value
- The 10‑year US Treasury yield is used as a proxy for the risk‑free rate, converted to a per‑minute rate. data
Source sentence
The 10‑year US Treasury yield, provided on a daily basis, is used to compute a per‑minute risk‑free rate, which is necessary for calculating excess returns
- Data are filtered to regular market hours (10:00 AM–3:30 PM CT) to avoid low‑liquidity periods. data
Source sentence
The dataset is filtered to focus on regular market trading hours, between 10:00 AM and 3:30 PM CT, to avoid periods of low liquidity
Show 2 more
- Training and testing splits are time‑ordered (80% train, 20% test) to prevent data leakage. data
Source sentence
The filtered dataset is then split into training and testing sets, with 80% of the data allocated to training and 20% to testing. The splitting is time‑ordered to preserve the temporal nature of the stock price data and avoid data leakage
- Model performance is evaluated using risk‑adjusted metrics (Rachev, Sharpe, etc.) assuming returns are excess of the per‑minute risk‑free rate. statistical
Source sentence
where rp is the asset return, rf is the risk‑free rate
Data
2024-08 – 2024-10 · SPY (S&P 500 ETF) · equities · intraday
- Minute‑level SPY data market-prices commercial
Source sentence
The dataset used in this study consists of minute‑level historical stock data for the SPY (S&P 500 ETF), covering the period from April 2024 to September 2024
- 10‑year US Treasury yield macro commercial
Source sentence
Additionally, the 10‑year US Treasury yield is incorporated as a proxy for the risk‑free rate, a crucial factor in calculating excess returns
Costs
Not modeled
Source sentence
it does not take into account bid–ask spreads, commission costs, or potential price impact
Finds 5
Sharpe 0.00 · total return -2.4% · max drawdown 4.0%
- In‑sample R² values range from 0.749 to 0.812, but out‑of‑sample R² becomes negative, indicating severe over‑fitting. predictive-accuracy
Source sentence
Our empirical results reveal a stark contrast between in‑sample and out‑of‑sample performance, with $R^2$ values deteriorating from 0.749--0.812 during training to negative values in testing
- Indicator‑enhanced models achieve Rachev ratios between 0.919 and 0.961 but generate negative returns (‑2.4% to ‑3.9%), underperforming a simple buy‑and‑hold benchmark. performance
Source sentence
Although the indicator‑enhanced models achieved superior risk‑adjusted metrics, with Rachev ratios between 0.919 and 0.961, they consistently underperformed a simple buy‑and‑hold strategy, generating returns ranging from -2.4% to -3.9%
- Primary price‑based features account for over 60% of feature importance, while technical indicators contribute only 14%–15%. statistical
Source sentence
A feature importance analysis demonstrates that primary price‑based features dominate the predictions made by the model, accounting for over 60% of the importance, while established technical indicators, such as RSI and Bollinger Bands, account for only 14%--15%
- Sharpe ratios of the random‑forest strategies are near zero (0.00 to 0.0046), indicating negligible risk‑adjusted returns. performance
Source sentence
with Sharpe ratios near 0.00 to 0.0046
- Most trading losses occur during periods of heightened market volatility, suggesting static models struggle to adapt to shifting regimes. negative
Source sentence
most trading losses occurred during periods of heightened market volatility, suggesting that the models struggle to adapt dynamically to shifting volatility regimes
Checked by others
Qualified by
- Demystifying the trend of the healthcare index: Is historical price a key driver?
Compare the findings
This paper: Primary price‑based features account for over 60% of feature importance, while technical indicators contribute only 14%–15%.
That paper: Raw original price features provide moderate but inconsistent predictive utility compared to nowcasting features.
2 papers in the archive
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.