Paper: arXiv 2502.09079
Authors: Francesco Puoti, Fabrizio Pittorino, Manuel Roveri
Abstract
This paper offers a thorough examination of the univariate predictability in cryptocurrency time-series. By exploiting a combination of complexity measure and model predictions we explore the cryptocurrencies time-series forecasting task focusing on the exchange rate in USD of Litecoin, Binance Coin, Bitcoin, Ethereum, and XRP. On one hand, to assess the complexity and the randomness of these time-series, a comparative analysis has been performed using Brownian and colored noises as a benchmark. The results obtained from the Complexity-Entropy causality plane and power density spectrum analysis reveal that cryptocurrency time-series exhibit characteristics closely resembling those of Brownian noise when analyzed in a univariate context. On the other hand, the application of a wide range of statistical, machine and deep learning models for time-series forecasting demonstrates the low predictability of cryptocurrencies. Notably, our analysis reveals that simpler models such as Naive models consistently outperform the more complex machine and deep learning ones in terms of forecasting accuracy across different forecast horizons and time windows. The combined study of complexity and forecasting accuracies highlights the difficulty of predicting the cryptocurrency market. These findings provide valuable insights into the inherent characteristics of the cryptocurrency data and highlight the need to reassess the challenges associated with predicting cryptocurrency’s price movements.
Complexity vs Empirical Score
- Math Complexity: 6.0/10
- Empirical Rigor: 8.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced mathematical methods like Permutation Entropy and the Complexity-Entropy causality plane (CH-plane) with Jensen-Shannon complexity, increasing math complexity. It demonstrates high empirical rigor through a multi-cryptocurrency study, use of multiple time windows, and rigorous comparison of statistical, ML, and DL models against naive baselines in a univariate forecasting task.
Research Flowchart
flowchart TD
A["Research Goal:<br>Quantify Crypto Predictability"] --> B["Data Input:<br>USD Exchange Rates of 5 Cryptos"]
B --> C{"Complexity Analysis<br>vs. Noise Benchmarks"}
C --> D["Complexity-Entropy<br>Causality Plane"]
C --> E["Power Density<br>Spectrum"]
D & E --> F{"Forecasting Analysis<br>Statistical, ML, and DL Models"}
F --> G["Key Findings:<br>Cryptos resemble Brownian Noise"]
F --> H["Key Findings:<br>Simpler models (Naive) outperform<br>complex ML/DL models"]
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 a combined study of time‑series complexity analysis and forecasting using a wide range of statistical, machine‑learning and deep‑learning models on cryptocurrency price data.
Source sentence
By exploiting a combination of complexity measure and model predictions we explore the cryptocurrencies time-series forecasting task focusing on the exchange rate in USD of Litecoin, Binance Coin, Bitcoin, Ethereum, and XRP.
Assumes 4
- The analysis is performed in a purely univariate context, without incorporating multivariate information. model
Source sentence
These findings provide valuable insights into the inherent characteristics of the cryptocurrency data and highlight the need to reassess the challenges associated with predicting cryptocurrency's price movements.
- No data preprocessing is applied to the price series. data
Source sentence
Each cryptocurrency time-series ranges from 2020-07-03 to 2023-12-21, and no data preprocessing was applied.
- The dataset is split at a fixed date, with the last 180 observations used as the test set. data
Source sentence
The split date index is 2023-07-04, meaning that the last 180 data points represent
- Forecasting performance is evaluated using mean absolute percentage error (MAPE) across multiple horizons. statistical
Source sentence
Table 4: Aggregated view of mean MAPE ± standard deviation across all cryptocurrencies and time windows...
Data
- Yahoo Finance daily pricing data market-prices public
Source sentence
The daily pricing data of the cryptocurrencies used in this study are collected from the Yahoo Finance database. Each cryptocurrency time-series ranges from 2020-07-03 to 2023-12-21...
Costs
Not applicable
Finds 4
- Cryptocurrency price series exhibit complexity and randomness characteristics closely resembling Brownian noise in a univariate analysis. statistical
Source sentence
The results obtained from the Complexity-Entropy causality plane and power density spectrum analysis reveal that cryptocurrency time-series exhibit characteristics closely resembling those of Brownian noise when analyzed in a univariate context.
- Simpler Naive forecasting models consistently outperform more complex machine‑learning and deep‑learning models across all horizons and time windows. predictive-accuracy
Source sentence
Notably, our analysis reveals that simpler models such as Naive models consistently outperform the more complex machine and deep learning ones in terms of forecasting accuracy across different forecast horizons and time windows.
- Advanced statistical models (AutoARIMA, AutoETS, AutoCES, TBATS) do not provide statistically significant improvements over Naive models. performance
Source sentence
This view illustrates that while statistical approaches generally outperform ML and DL methods, the complex statistical techniques (AutoARIMA, AutoETS, AutoCES, and TBATS) show no statistically significant improvement over Naive models, as evidenced by the overlapping standard deviations.
- Deep‑learning models (VanillaRNN, LSTM) suffer large prediction errors that increase as the forecasting window shrinks, indicating difficulty handling noise‑dominated cryptocurrency data. negative
Source sentence
As the time window shrinks, prediction errors for VanillaRNN and LSTM models increase across all forecast horizons.
Checked by others
Contradicted by
- A Time Series Approach to Explainability for Neural Nets with Applications to Risk-Management and Fraud Detection
Compare the findings
This paper: Cryptocurrency price series exhibit complexity and randomness characteristics closely resembling Brownian noise in a univariate analysis.
That paper: LPD sensitivities on price data show non‑stationary patterns that are difficult to reconcile with a random‑walk hypothesis.
2 papers in the archive : Cryptocurrency Time Series on the Binary Complexity-Entropy …
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (33)
- Bitcoin: A Peer-to-Peer Electronic Cash System (2008) — link
- Bitcoin: Medium of exchange or speculative assets? (2018) — link
- A brief survey of Cryptocurrency systems (2016) — link
- Prediction of cryptocurrency returns using machine learning (2021) — link
- Stock price prediction using geometric Brownian motion (2018) — link
- The dynamics of returns predictability in cryptocurrency markets (2023) — link
- A survey of deep learning applications in cryptocurrency (2024) — link
- Complexity analysis and forecasting of variations in cryptocurrency trading volume with support vector regression tuned by Bayesian optimization under different kernels: An empirical comparison from a large dataset (2022) — link
and 25 more
- Cryptocurrency forecasting with deep learning chaotic neural networks (2019) — link
- Complexity-entropy causality plane: A useful approach to quantify the stock market inefficiency (2010) — link
- Cryptocurrency price forecasting – A comparative analysis of ensemble learning and deep learning methods (2024) — link
- Forecasting and trading cryptocurrencies with machine learning under changing market conditions (2021) — link
- Permutation Jensen-Shannon distance: A versatile and fast symbolic tool for complex time-series analysis (2022) — link
- A Hybrid Model for Bitcoin Prices Prediction using Hidden Markov Models and Optimized LSTM Networks (2019) — link
- Predicting the Price of Bitcoin Using Machine Learning (2018) — link
- Forecasting cryptocurrencies under model and parameter instability (2019) — link
- Darts: User-Friendly Modern Machine Learning for Time Series (2022) — link
- Forecasting: Principles and Practice (3rd ed) (2021)
- Complex exponential smoothing (2022) — link
- Forecasting time series with complex seasonal patterns using exponential smoothing (2011)
- Forecasting at scale (2018)
- Random forests (2001)
- Xgboost: A scalable tree boosting system (2016)
- Long short-term memory (1997)
- Finding structure in time (1990)
- N-BEATS: Neural basis expansion analysis for time series forecasting (2019)
- Permutation entropy: a natural complexity measure for time series (2002)
- Distinguishing Noise from Chaos (2007) — link
- Intensive entropic non-triviality measure (2004) — link
- The Efficient Market Hypothesis: Review of Specialized Literature and Empirical Research (2015) — link
- Exploring disorder and complexity in the cryptocurrency space (2019) — link
- Spectral Analysis of Signals (2005) — link
- Statistical and Machine Learning forecasting methods: Concerns and ways forward (2018) — link
0 of 33 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.