Paper: arXiv 2401.10931

Authors: Sauren Gupta, Apoorva Hathi Katharaki, Yifan Xu, Bhaskar Krishnamachari, Rajarshi Gupta

Abstract

This research explores a relatively unexplored area of predicting cryptocurrency staking rewards, offering potential insights to researchers and investors. We investigate two predictive methodologies: a) a straightforward sliding-window average, and b) linear regression models predicated on historical data. The findings reveal that ETH staking rewards can be forecasted with an RMSE within 0.7% and 1.1% of the mean value for 1-day and 7-day look-aheads respectively, using a 7-day sliding-window average approach. Additionally, we discern diverse prediction accuracies across various cryptocurrencies, including SOL, XTZ, ATOM, and MATIC. Linear regression is identified as superior to the moving-window average for perdicting in the short term for XTZ and ATOM. The results underscore the generally stable and predictable nature of staking rewards for most assets, with MATIC presenting a noteworthy exception.

Complexity vs Empirical Score

  • Math Complexity: 2.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper employs straightforward predictive models like sliding-window averages and linear regression, with minimal advanced mathematical formalism. However, it demonstrates high empirical rigor through specific error metrics (RMSE) across multiple assets, backtesting protocols with 1- and 7-day forecasts, and detailed discussion of data sources and implementation challenges.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Predict Cryptocurrency Staking Rewards"] --> B["Data Input:<br>Historical Staking Reward Data"]
  B --> C{"Methodology Selection"}
  C --> D["Sliding Window<br>Average Model"]
  C --> E["Linear Regression<br>Model"]
  D --> F["Computation:<br>RMSE Evaluation"]
  E --> F
  F --> G["Key Findings & Outcomes"]
  
  subgraph G
      direction LR
      G1["ETH: High Accuracy<br>RMSE 0.7% - 1.1%"]
      G2["XTZ/ATOM: Linear Regression<br>Better for Short Term"]
      G3["MATIC: Exception<br>Less Predictable"]
  end

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 using a sliding‑window average and linear regression models to forecast cryptocurrency staking rewards.

Source sentence
The solution we propose in the paper involves the application of predictive models to forecast the staking rewards of a variety of cryptocurrencies.

Assumes 6

  • The sliding‑window average assumes that future rewards will mirror the recent past. model
    Source sentence
    Sliding-window average essentially involves computing the mean rewards over a specified time window, which is a straightforward and intuitive method that can provide a quick snapshot of the trend in staking rewards.
  • Linear regression assumes a linear relationship between the independent and dependent variables. model
    Source sentence
    Linear regression is an effective predictive model that can capture more complex patterns in the data. It assumes a linear relationship between the independent and dependent variables, which may not always hold true.
  • Historical trends and patterns in staking rewards contain useful information for future performance. data
    Source sentence
    Our approach is predicated on the assumption that historical trends and patterns in staking rewards can provide valuable insights into future performance.
  • The selected features (rewards, price, trends) are sufficient to model staking reward dynamics. data
    Source sentence
    Table 1 below delineates the selected features and their corresponding data sources that were utilized in this study.
Show 2 more
  • Training on three previous months and testing on the next month yields good performance. execution
    Source sentence
    We found that training for three previous months and using it to test for the next month gives good performance.
  • Root mean square error (RMSE) is an appropriate metric for evaluating prediction accuracy. statistical
    Source sentence
    The metrics considered is root mean square error (RMSE), which measures the average difference between a statistical model’s predicted values and the actual values.

Data

  • Time series of rewards · Coinbase data storage crypto-onchain commercial
    2021-06-23 to 2022-08-06 · ETH
    Source sentence
    Taking the example of ETH, a collection of time-series data for about one year from 2021-06-23 to 2022-08-06 was diligently gathered in Figure 1.
  • Price feed · Coinbase data storage market-prices commercial
    ETH, SOL, XTZ, MATIC, ATOM
    Source sentence
    ... Price feed Real-time price information of the asset
  • Trends · Google Trends alternative public
    ETH, SOL, XTZ, MATIC, ATOM
    Source sentence
    Trends The frequency of searches for the cryptocurrency asset
  • Additional features from YFinance · YFinance market-prices public
    ETH, SOL, XTZ, MATIC, ATOM
    Source sentence
    These features were extracted from a variety of sources, including Coinbase data storage, Google Trends and YFinance.

Recognised sources: Coinbase Yahoo Finance

Costs

Not applicable The paper reports no trading or execution result.

Finds 2

  • Across tokens, the moving‑window average generally yields lower RMSE/Mean than linear regression for multi‑day forecasts, except for MATIC. performance
    Source sentence
    The moving window average performs better overall for all of the tokens except MATIC.
  • Simple linear regression using only reward history outperforms multiple regression with price and trends. predictive-accuracy
    Source sentence
    Simple linear regression based on the staking rewards time series alone was found to outperform multiple linear regression using other features.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 9 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 (21)

  1. Blockchain without waste: Proof-of-stake (2021)
  2. Evaluation of performance and security of proof of work and proof of stake using blockchain (2021)
  3. Proof-of-stake consensus mechanisms for future blockchain networks: fundamentals, applications and opportunities (2019)
  4. What proof of stake is and why it matters (2013)
  5. Ethereum white paper (2013)
  6. Cosmos whitepaper (2019)
  7. Equilibrium staking levels in a proof-of-stake blockchain (2021)
  8. Long short-term memory (1997)
and 13 more
  1. Optimal Staking and Liquid Token Holding Decisions in Cryptocurrency Markets (2023)
  2. Reap the harvest on blockchain: A survey of yield farming protocols (2022)
  3. Staking pool centralization in proof-of-stake blockchain network (2020)
  4. Staking pools on blockchains (2022) — link
  5. A survey on machine learning models for financial time series forecasting (2022)
  6. Machine Learning approaches in stock price prediction: A systematic review (2022)
  7. Comparative performance of machine learning ensemble algorithms for forecasting cryptocurrency prices (2021)
  8. Deep learning in predicting cryptocurrency volatility (2022)
  9. A systematic review of the bubble dynamics of cryptocurrency prices (2020)
  10. Cryptocurrency price prediction using traditional statistical and machine-learning techniques: A survey (2021)
  11. Applying Artificial Intelligence in Cryptocurrency Markets: A Survey (2022)
  12. Time-series forecasting of Bitcoin prices using high-dimensional features: a machine learning approach (2020)
  13. A novel cryptocurrency price prediction model using GRU, LSTM and bi-LSTM machine learning algorithms (2021)

0 of 21 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.