Cryptocurrency Trading: A Comprehensive Survey

Background

• More than 85% of cryptocurrency trading papers appeared since 2018, underscoring the field’s recent emergence; the survey samples work from 2013 through June 2021.Figure 1 reports cumulative cryptocurrency trading publications during 2013–2021 (June 2021).Source: §1 · p. 1, Figure 1 · p. 2

Survey scope

• The literature is organised into six areas: software systems, systematic trading, emergent technologies, portfolios and assets, market conditions, and miscellaneous research.The categories include real-time, turtle, and arbitrage systems; technical and pairs trading; econometric and machine-learning methods; co-movements; bubbles, crashes, and extreme conditions.Source: §1 · pp. 1–2

Contributions

• The paper analyses research objectives, technologies, datasets, and trends, distinguishing statistical methods from machine-learning technologies to guide researchers.It also identifies potential research directions and aims to provide sufficient material for researchers interested in cryptocurrency trading.Source: §1 · pp. 2–3

Related work

• Unlike prior surveys focused on market efficiency, cryptocurrencies, systems, or bitcoin bibliometrics, this work claims the first comprehensive survey specifically focused on cryptocurrency trading.Related work includes Kyriazis et al., Ahamad et al., Sharma et al., Ujan et al., and Ignasi et al..Source: §1 · p. 2

Contributions

• It defines cryptocurrency trading through cryptocurrency markets, trading models, and trading strategies, with strategies as the survey’s core focus.The survey nonetheless covers all aspects of cryptocurrency trading.Source: §1 · p. 3
• The survey covers 146 cryptocurrency trading papers spanning finance, economics, artificial intelligence, and computer science.Papers addressing multiple aspects are surveyed in each relevant category.Source: §1 · p. 3

• The section frames cryptocurrency trading around blockchain, cryptocurrency markets, and cryptocurrency trading strategies.These topics provide the section’s introductory scope.Source: §2 · p. 3

2.1.1 Blockchain Technology Introduction

• Blockchain records immutable, time-stamped data in cryptographically linked blocks managed by a distributed network, as illustrated by its transaction workflow.It can record financial transactions and other objects with intrinsic value [8].Source: §2.1.1 · p. 3, Figure 3 · p. 4

2.1.2 From Blockchain to cryptocurrencies

• Cryptocurrency transactions use peer-to-peer verification and public-key signatures, while miners confirm valid transactions by adding hashed blocks that subsequent blocks further confirm.Cryptocurrencies are Blockchain-based tokens used to transfer value, pay transaction fees, and sometimes digitize assets or provide utility.Source: §2.1.1 · p. 4, §2.1.2 · p. 4

2.2.1 What is cryptocurrency?

• Cryptocurrencies are decentralized media of exchange that use cryptographic functions and blockchain technology to provide decentralization, transparency, and immutability.Their security is based on cryptography rather than people or trust.Source: §2.2.1 · p. 4
• Excluding financial intermediaries lowers transaction costs, while blockchain decentralization allows transaction verification across the network and theoretically limits central-authority and government control.Cryptocurrencies also control token supply through blockchain-encoded timetables, with Bitcoin expected to reach its final quantity around 2,140.Source: §2.2.1 · p. 5
• As of December 20, 2019, 4,950 cryptocurrencies and 20,325 cryptocurrency markets represented a market capitalization of around 190 billion dollars, followed by exponential growth in 2017, a large bubble burst in early 2018, and dramatic pandemic-era appreciation.In 2021, cryptocurrency market value remained highly volatile but consistently at historically high levels.Source: §2.2.1 · p. 5, Figure 4 · p. 6

2.2.2 Cryptocurrency Exchanges

• Cryptocurrency exchanges let customers trade cryptocurrencies as market makers using bid-ask spreads or as matching platforms charging fees; listed venues include regulated exchanges such as CME, CBOE, and BAKKT.Table 1 lists classical exchanges by volume, with regulatory authority and supported currencies collected from official sources.Source: §2.2.2 · p. 6, Table 1 · p. 7

2.3.1 Definition

• Cryptocurrency trading is buying and selling cryptocurrencies for profit through exchange transactions or Contracts for Differences, guided by algorithmic rules.Its three aspects are the traded cryptocurrency, transaction mode, and predefined trading strategy.Source: §2.3.1 · p. 7

2.3.2 Advantages of Trading Cryptocurrency

• Cryptocurrency trading offers intraday profit opportunities from volatility, operates continuously worldwide, can preserve user privacy under exchange-specific KYC, and enables peer-to-peer transactions with lower intermediation costs.Some cryptocurrencies also provide programmable capabilities, including limited ownership, voting rights, or partial interests in physical assets.Source: §2.3.2 · pp. 7–8

2.3.3 Disadvantages of Trading Cryptocurrency

• Cryptocurrency trading faces transaction backlogs from limited scalability, cybersecurity breaches, and uncertain or divergent regulatory systems across countries.The Poly Network breach reportedly involved over $600 million in stolen Ethereum and other cryptocurrencies, while security requires ongoing enhanced infrastructure.Source: §2.3.3 · p. 8

3 Cryptocurrency Trading Strategy

• Cryptocurrency trading strategies are grouped into technical and fundamental approaches, with programmatic trading emerging as a third category that executes quantitative decisions automatically.Programmatic strategies primarily use price, volume, technical indicators, or ratios to exploit market inefficiencies; cryptocurrency markets’ arbitrage opportunities, fluctuation, and transparency encourage their use.Source: §3 · pp. 8–9

3.1 Cryptocurrency Trading Software System

• Cryptocurrency trading systems pre-program principles and procedures for cryptocurrency and fiat transactions while addressing manipulation, cybercrime, and transaction delays.Commercial examples include Capfolio, 3Commas, CCXT, Freqtrade, and Ctubio, and system design considers the capital market, base asset, investment plan, and strategies.Source: §3.1 · p. 9

3.2 Systematic Trading

• Systematic cryptocurrency trading covers technical analysis, pairs trading, and other strategies, with technical rules reported to generate excess returns and pairs trading profiting when spreads return toward equilibrium.Pairs trading shorts the higher-priced similar asset and buys the lower-priced one when spreads widen, then closes positions as the spread narrows.Source: §3.2 · p. 9

3.3 Tools for building automated trading systems

• Tools for automated cryptocurrency trading include emergent strategies based on econometrics and machine learning technologies.These approaches provide the main technology categories reviewed for building automated trading systems.Source: §3.3 · p. 9

3.3.1 Econometrics on Cryptocurrency

• Econometric cryptocurrency research applies statistical and economic models to time-series data, including GARCH, BEKK, linear models, and ARMA, to evaluate price fluctuations and predict volatility.Volatility prediction helps assess cryptocurrency price fluctuation and supports pricing cryptocurrency-related derivatives.Source: §3.3.1 · pp. 9–10

3.3.2 Machine Learning Technology

• Machine learning develops cryptocurrency trading signals by mapping economic, social, technical, and seasonal inputs to predictive objectives and generalizing historical patterns to unseen data.Model validation through statistical approaches such as cross-validation is necessary because generalization error is especially important in financial applications; methods include supervised, unsupervised, and reinforcement learning.Source: §3.3.2 · pp. 10–11, Table 2 · p. 10

3.4 Portfolio Research

• Crypto-asset portfolio research applies diversification across cryptocurrencies, market sectors, and global assets such as stocks and futures to allocate investments for targeted risk and return.Mean-variance optimisation is presented as a prominent portfolio-theory approach.Source: §3.4 · p. 11

3.5 Market Condition Research

• Market-condition research focuses on cryptocurrency bubbles, crashes, and extreme volatility, including Bitcoin’s approximately +600% rise from $10,645 to $63,346 between October 7, 2020 and April 15, 2021.The 2017 bubble and 2018 Bitcoin collapse motivated research on bubbles and extreme conditions, while some experts recommend limiting cryptocurrency exposure to 1% or 2%.Source: §3.5 · pp. 11–12

4.1 Survey Scope

• The survey uses a bottom-up scope from trading-system platforms and technologies through prediction, trading signals, technical methods, market conditions, portfolios, and risk management.Additional covered topics include pricing rules, dynamic market analysis, and regulatory implications.Source: §4.1 · p. 12

4.2 Paper Collection Methodology

• Papers were collected using six cryptocurrency-focused keyword searches across Google Scholar and arXiv, with SSRN treated as duplicate Google Scholar coverage and snowballing continued to closure.The searches ran until July 1, 2021, and arXiv papers had not undergone formal peer review.Source: §4.2 · p. 13

4.3 Collection Results

• 146 papers were identified through keyword searches and snowballing across six cryptocurrency-trading research areas.The collection used Google Scholar and arXiv searches supplemented by snowballing until closure.Source: §4.3 · p. 13, §4.2 · p. 13
• 48.63% of collected papers appeared in Finance and Economics venues, while the remainder spanned Science, engineering, data mining, physics, AI, other venues, and arXiv.The venue distribution indicates that cryptocurrency-trading research is concentrated in Finance and Economics but broadly interdisciplinary.Source: §4.3 · pp. 13–14, Figure 7 · p. 14

4.4 Survey Organisation

• The review organises papers across six angles covering software systems, systematic trading, emergent technologies, portfolios and market conditions, and other cryptocurrency-trading research.It later compares research properties and technologies, summarises datasets and timelines, and identifies future research opportunities.Source: §4.4 · pp. 14–15

5.1 Trading Infrastructure Systems

• Existing cryptocurrency trading systems include proprietary platforms, unified-API libraries, and open-source platforms supporting exchange connectivity, analysis, strategy development, backtesting, or automated trading.Capfolio supports five exchanges and advanced backtesting, 3 Commas supports twelve exchanges, CCXT provides unified exchange access, and StockSharp supports 48 exchanges.Source: §5.1 · pp. 15–16

5.2 Real-time Cryptocurrency Trading Systems

• A real-time cryptocurrency trading system combined web clients, servers, Coinmarket API data collection, and databases for balances, trades, and order books, demonstrating user-friendly and secure trading.Traders used a web application to log in and buy or sell crypto assets.Source: §5.2 · p. 17

5.3 Turtle trading system in Cryptocurrency market

• Turtle trading produced 18.59% average net profit margin and 35.94% average profitability, whereas Extended Turtle reached 114.41% and 52.75%, respectively, across eight cryptocurrencies.The comparison used 87 trades for Original Turtle and 41 trades for Extended Turtle over nearly one year.Source: §5.3 · pp. 17–18

5.4 Arbitrage Trading Systems for Cryptocurrencies

• The arbitrage system found 186 signals among 787 cryptocurrencies across seven exchanges, including a BTG-BTC opportunity yielding up to 495.44%.The system bought on Cryptopia and sold on Binance, and the paper reported three additional well-traded signals with profit expectations around 20%.Source: §5.4 · p. 18

5.5 Characteristics of three cryptocurrency trading systems

• Turtle trading delivered high returns with high risk, while arbitrage earned less with lower risk; both systems performed well in capturing alpha.These findings characterize the contrasting profit and risk behavior of the two systems.Source: §5.5 · p. 18

6.1 Technical Analysis

• Technical-analysis studies examine five classical strategies using breakout, momentum, reversal, bottom-rotation, chart-pattern, and support-resistance signals.The strategies include Turtle Soup, Nem (XEM), Amazing Gann Box, Busted Double Top Pattern, and Bottom Rotation Trading.Source: §6.1 · pp. 18–19, Table 7 · p. 19
• Variable-length moving-average rules generated the most useful high-frequency signals, while broader evidence found technical rules predictive, profitable, and capable of producing an 8.76% annualised excess return excluding Bitcoin.The findings draw on nearly 15,000 trading rules, one-minute Bitcoin data, and daily data for 11 major cryptocurrencies.Source: §6.1 · p. 19
• Cryptocurrency markets exhibited long-term memory, multiple fractals, and time-varying inefficiency, with high liquidity and low volatility facilitating arbitrage for active traders.These properties were assessed using Hurst exponents, time-rolling MF-DFA, and quantile regression.Source: §6.1 · p. 19

6.2 Pairs Trading

• Pairs-trading models exploited stable long-run relationships and mean reversion, achieving a 3% monthly profit and finding 31 significantly cointegrated pairs across and within sectors.Extensions included intraday high-frequency trading, stochastic-control portfolio weights, and time-varying volatility models.Source: §6.2 · pp. 19–20

6.3 Others

• Order-size imbalances before large positive or negative events provided evidence of informed trading in Bitcoin, suggesting profits from private information obtained before wider dissemination.The evidence used USD/BTC exchange-rate trading data and a volume-imbalance-inspired indicator.Source: §6.3 · p. 20

7 Emergent Trading Technologies

• Econometric studies find cryptocurrency markets exhibit tail causality, time-varying integration, regime-dependent volatility, spillovers, persistence, and non-normal, heteroskedastic returns.Trading volume Granger-causes returns in seven large cryptocurrencies at both tails, MSGARCH outperforms single-regime GARCH for VaR forecasting, and volatility correlations and shock transmission vary across assets and regimes.Source: §7.1 · pp. 20–22

7.2.1 Common Machine Learning Technology in this survey

• The survey classifies common machine-learning technologies by objective into classification, clustering, regression, reinforcement learning, and deep learning.Classifiers include NB, SVM, KNN, DT, RF, and GB; regression includes linear regression and scatterplot smoothing; reinforcement learning includes DQN and DBM; and deep learning includes CNN, RNN, GRU, MLP, LSTM, and Seq2seq.Source: §7.2.1 · pp. 23–24

7.2.2 Research on Machine Learning Models

• In cryptocurrency prediction, SVM achieved 62.31% accuracy and 0.77 precision, while LSTM improved F1-score about 7% over GB for 10-minute prediction and suited high-volatility data.NARX achieved nearly 52% prediction accuracy at 10-second intervals. VAR augmentation allowed R2N2 to be shallower, quicker, and better-predicting than an RNN.Source: §7.2.2 · pp. 26–27

7.2.3 Deep Neural Network

• Deep neural-network studies use CNN-LSTM, autoencoder-LSTM, recurrent, and ensemble architectures to improve cryptocurrency forecasting, with performance shaped by complexity, retraining, dataset size, and model architecture.CNN-LSTM achieved the highest accuracy among the compared models on MAE, RMSE, and R2, while an autoencoder-augmented LSTM achieved 78% accuracy for tick-level prediction of eight cryptocurrency pairs; LSTM was more robust than MLP for long-term dependencies.Source: §7.2.3 · pp. 27–28

7.2.4 Sentiment Analysis

• Sentiment and attention data provide trading signals: LR correctly predicted 43.9% of price increases and 61.9% of decreases, Smuts’s LSTM reached 76% hourly-price accuracy, and community activity directly affected price and volume fluctuations.Google searches significantly influenced short-term Bitcoin returns, search trends were connected with cryptocurrency prices, and accumulated data and animated community activity were associated with market fluctuations.Source: §7.2.4 · pp. 28–29
• Sentiment signals from Twitter, news, internet searches, and GitHub comments showed predictive or investment relevance for cryptocurrency prices and returns.Topic-price relationships, social-media sentiment, news attention, economic-uncertainty queries, and repository-comment sentiment were investigated across cryptocurrency markets.Source: §7.2.4 · p. 29
• 0.806 Pearson correlation on test data was obtained by extreme-gradient-boosting regression using hourly Twitter sentiment and trading volume to predict ZClassic prices.The correlation was statistically significant at p < 0.0001; related work also found cryptocurrency social-media sentiment capable of generating positive return on investment and GitHub-comment metrics predictive of prices.Source: §7.2.4 · p. 29

7.2.5 Reinforcement Learning

• 24% profit during a 64% cryptocurrency-market drop was achieved by a double-Q network with unsupervised pre-training, outperforming LSTM, CNN, and MLP models in extreme conditions.The architecture combines spectral and temporal feature extraction with a double-Q network for action generation.Source: §7.2.5 · p. 30
• 14.4% monthly BTC profit, 74% LTC profit, and 41% ETH profit were obtained by a deep-reinforcement-learning model selecting buy, sell, or hold actions.Other reinforcement-learning studies found DQN-CNN superior to DQN-MLP and double Q-learning more profitable than dueling double Q-learning under their respective backtests.Source: §7.2.5 · p. 30
• 71.21% higher simulated investment return than buy-and-hold was achieved by the PATSOS hybrid neuro-fuzzy controller for forecasting Bitcoin price direction.Topological data analysis separated important dynamical patterns from sampling noise, while transaction-network edges yielded 60.05% price-movement classification accuracy using approximately 0.45% of unique edges [211].Source: §7.2.5 · pp. 30–31

8.1 Research among cryptocurrency pairs and related factors

• Bitcoin was dominant in cryptocurrency exchange-rate dynamics, while Litecoin and Bitcoin exerted the greatest effects on other cryptocurrencies through return and volatility spillovers.Connectedness was not necessarily explained by market size, and Bitcoin’s dynamics were at least as influential as the US dollar’s during the studied period.Source: §8.1 · pp. 31–32
• Bitcoin showed long-run hedging effects on other cryptocurrencies and strong hedging or safe-haven properties for energy commodities, while its relation to gold and currencies remained asset-like but contested.Studies reported Bitcoin’s returns and volatility leading other cryptocurrencies at low frequencies, alongside distinct speculative risk-return characteristics relative to gold and the US dollar.Source: §8.1 · p. 32

8.2 Crypto-asset Portfolio Research

• 1/N portfolios outperformed single cryptocurrencies and over 75% of mean-variance optimal portfolios in Sharpe ratio and certainty equivalent returns, although adding crypto-assets increased risk exposure.Mean-variance analysis found that combining cryptocurrencies expands low-risk investment opportunities, while Omega-based optimization improved returns with greater risk exposure.Source: §8.2 · pp. 33–34
• Risk models for cryptocurrency portfolios achieved good VaR and Expected Shortfall estimation, with the t-copula/skewed-t GARCH model superior for market risk and ZPP-based models superior for credit risk.Cauchy-distribution modeling also yielded analytical VaR measures and optimized cryptocurrency portfolios.Source: §8.2 · p. 34
• Deep-learning portfolio methods produced positive results, with a CNN strategy ranking second to PAMR and a stacked recurrent model outperforming other regression techniques.The CNN learned asset weights by maximizing cumulative returns, whereas the recurrent model generated ex-ante portfolio return and risk estimates.Source: §8.2 · pp. 34–35

8.3 Bubbles and Crash Analysis

• Bubble-related research found no clear persistent bubble in Bitcoin or Ethereum, while price overreactions did not provide profit opportunities after transaction costs and permanent volatility jumps were important.Reddit usage was strongly related to cryptocurrency prices, and negative news during the 2017 boom helped explain Litecoin and Ripple returns during the 2018 crash.Source: §8.3 · pp. 35–36
• Cryptocurrency bubbles occurred in multiple explosive periods, with higher volatility and trading volume positively associated with bubbles and probit models outperforming linear models for prediction.Explosivity in one cryptocurrency generally depended on explosivity in others, indicating contemporaneous co-explosivity not necessarily determined by market size.Source: §8.3 · p. 35

8.4 Extreme condition

• Extreme-market studies found heavier-tailed and asymmetric return-volume dependence, greater tail than median interdependence, and higher shock persistence after the 2017/18 crash.Liquidity demand and dynamic herding can promote extreme price movements, while market size and momentum drive connectivity under both normal and extreme conditions.Source: §8.4 · p. 36

Market behavior

• Market-behavior studies examined GUI and API effects, behavioral asset-pricing anomalies, and trader heterogeneity, while an agent-based model reproduced Bitcoin’s unit roots, fat tails, volatility clustering, mining, hashing power, and energy use.Yang tested 20 market anomalies and highlighted speculators’ role in momentum and reversal research.Source: §9 · p. 37
• 85% of variations in bitcoin returns and idiosyncratic components of order flow help explain arbitrage spreads between exchanges, while herding and trend chasing appear in some cryptocurrency markets.Leclair and Vidal-Thomás et al. also found significant evidence of market herding.Source: §9 · p. 37
• Purchases with Tether were timed after market downturns and produced sizeable increases in Bitcoin prices, with one large Bitfinex player buying Bitcoin as prices fell.Griffin et al. mapped Bitcoin and Tether blockchains using algorithms to analyze these transactions.Source: §9 · p. 37

Benchmarks and market infrastructure

• Research benchmarked exchanges, wallets, payments, and mining, and examined algorithmic trading, decentralized exchanges, automated market makers, artificial markets, network effects, cryptocurrency money, and literature-review methods [286].Aspris et al. reported significant listing and trading differences between decentralized and centralized exchange tokens, while Cocco et al. reproduced some Bitcoin return statistics with heterogeneous agents.Source: §9 · pp. 37–38

Regulation and market analysis

• Analysis of global cryptocurrency regulations found stricter regulation undesirable, while exchange data showed no systematic evidence that regulatory measures caused traders to flee or enter affected regional jurisdictions.Uncertainty quantification and opinion mining analyzed market sentiment, while permutation entropy and statistical complexity found more than four hundred cryptocurrencies significantly complied with efficient-market assumptions [280, 281].Source: §9 · p. 38

Trading practice and tools

• Introductory resources explain cryptocurrency, wallets, exchanges, investment steps, automated systems, algorithmic strategies, and Bitcoin financial analysis through books, simulations, and software packages.Ward et al. discuss parameter adjustment, strategy combinations, and dynamic switching, while Fantazzini et al. introduce Bitcoin-Finance and bubble in R.Source: §9 · p. 38

10.1 Timeline

• The timeline progresses from Bitcoin’s invention in 2009 and the first cryptocurrency exchange in 2010 to search-query correlations, causality, bubble analysis, volatility modeling, machine-learning prediction, portfolios, networks, and regulation.Key later events include China’s 2021 transaction-related ban, which led to a sharp bitcoin-price drop, and El Salvador’s adoption of Bitcoin as legal tender.Source: §10.1 · p. 39

10.2 Research Distribution among Properties

• 37.67% of papers study return prediction, while bubbles/extreme conditions and relationships among cryptocurrency pairs and portfolios form another major research focus.Prediction of volatility, trading systems, technical trading, and other topics collectively account for roughly one-third of the research.Source: §10.2 · p. 40

10.3 Research Distribution among Categories and Technologies

• Among 146 papers, 102 (69.86%) cover statistical methods or machine learning; 88 of these (86.28%) use statistical methods, 13.72% use machine learning, and 17 (16.67%) compare multiple techniques.The analysis excludes papers focused purely on technical-indicator trading systems or industry and historical discussions.Source: §10.3 · p. 40

10.3.1 Research Distribution among Statistical methods

• Basic regression and time-series analysis are the most commonly used statistical methods in cryptocurrency-trading research.The statistical-method categories include regression, classifiers and clustering, time-series models, decision trees and probabilistic classifiers, modern portfolio theory, and other market analyses.Source: §10.3.1 · p. 41

10.3.2 Research Distribution among Machine Learning Categories

• LSTM, RNN, and GRU methods are the most popular machine-learning category, which accounts for 13.7% of the total papers.The review distinguishes ANNs, temporal-structure models such as LSTM/RNN/GRUs, and deep-learning or reinforcement-learning methods.Source: §10.3.2 · p. 41

10.4 Datasets used in Cryptocurrency Trading

• Datasets mainly contain exchange-derived prices, trading volumes, and order-level information, while sentiment datasets combine market data with media or Internet data and labels.Bitcoin is the most commonly used cryptocurrency; dataset descriptions distinguish resolution, time range, task, and sources such as exchanges, aggregated indices, and user forums.Source: §10.4 · pp. 41–42
• Less than half of cryptocurrency papers published since January 2017 used correct data, according to Alexander et al. ’s analysis of 152 published and SSRN discussion papers.This finding highlights a data-quality issue in the literature’s empirical foundation.Source: §10.4 · p. 42

Sentiment-based research

• Sentiment research could expand media inputs, update NLP baselines, improve text preprocessing and labeling, extend holding-period samples, incorporate transaction fees, and study opinion dynamics and user reputation.Suggested extensions include adding video sources, neural networks for label training, and more robust preprocessing.Source: §11 · pp. 42, 45

Long-and-short term trading research

• Long- and short-term cryptocurrency trading require different priorities: longer horizons emphasize risk control, whereas shorter horizons emphasize transaction costs, noise, and automated trading.Further opportunities include trading-signal extraction, time-series research, portfolio management, crash-related price movements, and cryptocurrency derivative pricing.Source: §11 · p. 45

Correlation between cryptocurrency and others

• Further research should examine cryptocurrency correlations with other assets and currencies, especially during extreme conditions such as financial collapse, potentially using principal component analysis.Existing studies discuss correlations with other financial markets as predictors of cryptocurrency-market direction.Source: §11 · p. 45

Bubbles and crash research

• Bubble and crash research could connect bubble formation to financial collapse and aftermaths, apply microeconomic, physical, or industrial models such as Omori law, and model supply and demand.Existing work examines cryptocurrency bubbles, links with the VIX, and spillover effects.Source: §11 · pp. 45–46

Game theory and agent-based analysis

• Game theory and agent-based modeling represent a promising research direction for cryptocurrency trading, following their use in traditional financial markets.The section explicitly proposes applying these methods to cryptocurrency markets.Source: §11 · p. 46

Public nature of Blockchain technology

• Blockchain transaction networks and user identification could provide new price-prediction features and improve understanding of financial bubbles.The proposed direction relies on deeper analysis of connections between transaction-network formation and currency prices.Source: §11 · p. 46

Balance between the opening of trading research literature and the fading of alphas

• As cryptocurrency research may reduce market predictability by exposing mispricing, future work could develop real-time pricing methods and account for the growing share of informed traders.This addresses the balance between alpha trading and the publication of trading research, as discussed by Mclean et al..Source: §11 · p. 46

• The survey covers 146 cryptocurrency trading papers and analyses the literature’s definitions, research distributions, datasets, trends, and opportunities.Research distributions are examined across properties, categories, and technologies; future directions are discussed in Section 11.Source: §12 · p. 46
• The survey offers academics and quantitative traders a concise way to become familiar with cryptocurrency trading research and pursue pressing problems.It is intended to motivate further contributions along the research opportunities identified by the survey.Source: §12 · p. 46