Paper: arXiv 2511.06224

Authors: Anmar Kareem, Alexander Aue

Abstract

This paper evaluates the performance of classical time series models in forecasting Bitcoin prices, focusing on ARIMA, SARIMA, GARCH, and EGARCH. Daily price data from 2010 to 2020 were analyzed, with models trained on the first 90 percent and tested on the final 10 percent. Forecast accuracy was assessed using MAE, RMSE, AIC, and BIC. The results show that ARIMA provided the strongest forecasts for short-run log-price dynamics, while EGARCH offered the best fit for volatility by capturing asymmetry in responses to shocks. These findings suggest that despite Bitcoin’s extreme volatility, classical time series models remain valuable for short-run forecasting. The study contributes to understanding cryptocurrency predictability and sets the stage for future work integrating machine learning and macroeconomic variables.

Complexity vs Empirical Score

  • Math Complexity: 6.5/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper uses advanced time series models (ARIMA, SARIMA, GARCH, EGARCH) requiring statistical theory and model diagnostics, scoring high on math. It also demonstrates strong empirical rigor with a real dataset (Bitcoin 2010-2020), a clear train/test split, and standard forecasting metrics (MAE, RMSE, AIC, BIC).

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Forecast Bitcoin Prices & Volatility<br>using Classical Time Series Models"] --> B["Data Input:<br>Daily Bitcoin Prices (2010-2020)<br>Train: 90% | Test: 10%"]
  B --> C["Methodology:<br>ARIMA, SARIMA, GARCH, EGARCH"]
  C --> D["Evaluation Metrics:<br>MAE, RMSE, AIC, BIC"]
  D --> E["Key Findings / Outcomes"]
  E --> F["Price Forecasting:<br>ARIMA best for<br>short-run log-prices"]
  E --> G["Volatility Modeling:<br>EGARCH best fit<br>(captures asymmetry)"]
  E --> H["Conclusion:<br>Classical models remain<br>valuable despite volatility"]

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 evaluates classical time series models for Bitcoin price and volatility forecasting, focusing on ARIMA, SARIMA, GARCH, and EGARCH.

Source sentence
This paper evaluates the performance of classical time series models in forecasting Bitcoin prices, focusing on ARIMA, SARIMA, GARCH, and EGARCH.

Assumes 4

  • Log‑return series is stationary, justifying ARIMA‑type models. statistical
    Source sentence
    In Table 1, the ADF test produced a very low p‑value, which allows us to reject the null hypothesis of a unit root. This indicates that the log return series is stationary
  • Log‑return series exhibits conditional heteroskedasticity, justifying GARCH‑type models. statistical
    Source sentence
    The Engle’s ARCH test in Table 1 also reported a highly significant p‑value, providing strong evidence of conditional heteroskedasticity in the log returns.
  • Models are trained on the first 90 % of the data and tested on the final 10 % for out‑of‑sample evaluation. data
    Source sentence
    each model trained on the first 90% and tested on the final 10%.
  • Only Bitcoin price values ≥ $100 are retained to avoid distortion from early low‑price periods. data
    Source sentence
    the dataset was also filtered to include only Bitcoin values of $100 or higher.

Data

Backtest sample: 2010 – 2020 · Bitcoin · crypto · daily

  • btc data.csv · GitHub crypto-onchain public
    July 17, 2010 to February 2, 2020 · Bitcoin
    Source sentence
    The dataset used in this study is publicly available on GitHub (btc data.csv). It covers daily Bitcoin activity from July 17, 2010 to February 2, 2020, with 3,488 rows of observations.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • ARIMA provided the strongest forecasts for short‑run log‑price dynamics, outperforming SARIMA in MAE and RMSE. predictive-accuracy
    Source sentence
    ARIMA provided the strongest forecasts for short-run log-price dynamics, while EGARCH offered the best fit for volatility by capturing asymmetry in responses to shocks.
  • ARIMA(1,1,1) achieved lower MAE (0.1366) and RMSE (0.1634) than SARIMA(1,1,1)(1,1,1,7). performance
    Source sentence
    ARIMA(1,1,1) performed better with lower MAE and RMSE values, showing that a simpler specification captured the short-run dynamics of Bitcoin more effectively than SARIMA.
  • EGARCH(1,1) gave the best fit for volatility, with lower AIC/BIC than the basic GARCH model and captured asymmetric shock responses. predictive-accuracy
    Source sentence
    the EGARCH(1,1) model achieved the best fit, as reflected in lower information criteria values compared to the basic GARCH model, and successfully captured the asymmetric response of volatility to negative and positive shocks.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 41 related papers). Coverage is partial, so this is neither replication nor refutation.

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 (2)

  1. Bitcoin prices using high-dimensional features: A machine learning approach
  2. Bitcoin traded at extremely low values, which could distort the analysis and reduce the reliability of the forecasting results

0 of 2 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.