Information, Computing and Intelligent Systems, 2024, No.

5, 112 – 124
UDC 004.94 https://doi.org/10.20535.2786-8729.5.2024/316563

THE AUTOMATIC CRYPTOCURRENCY TRADING SYSTEM

USING A SCALPING STRATEGY
Elisa Beraudo *
National Technical University of Ukraine
“Igor Sikorsky Kyiv Polytechnic Institute”, Kyiv, Ukraine
https://orcid.org/0000-0001-7550-3620

Yurii Oliinyk
National Technical University of Ukraine
“Igor Sikorsky Kyiv Polytechnic Institute”, Kyiv, Ukraine
https://orcid.org/ 0000-0002-7408-4927

*Corresponding author: [email protected]

The study focuses on the development and implementation of an automated system for scalping
strategies in cryptocurrency markets. Scalping, a high-frequency trading strategy, aims to generate
profits from small price fluctuations. The primary goal of the research is to create an automated trading
bot that addresses critical issues such as latency, risk management, scalability, and reliability in
real-world market conditions. To achieve this, the following objectives were defined: develop a novel
scalping method, implement a software solution to integrate the method into an automated trading
system, and evaluate its effectiveness through experimental testing.
The research methodology utilized technical indicators, including the Exponential Moving
Average (EMA) and Volume Weighted Average Price (VWAP). Pseudocode was created to illustrate
the decision-making process, incorporating key parameters such as smoothing factors, time periods, and
thresholds for trade execution. The software architecture consists of modules: Binance exchange
integration, data collection and management, strategy analysis, trade execution, and historical data
storage. Technologies such as PostgreSQL, Redis, WebSocket, and Python libraries (Pandas, NumPy,
TA-Lib) were employed to ensure the robustness and efficiency of the system.
Experiments were conducted using the BTC/USDT trading pair, known for its high liquidity and
volatility. The system was tested on hardware featuring an Intel Core i7-10700K processor, 32 GB of
RAM, and a 1 Gbps network connection. A comparative analysis between the scalping strategy and a
trend-following strategy demonstrated the advantages of scalping in volatile markets. The scalping bot
executed 15 trades (13 successful) within two hours, achieving a total profit of 120 USDT.
Performance metrics, including latency (15–50 ms), signal processing time, CPU utilization
(5–55%), and memory usage (120–2100 MB), were measured. The results confirmed the system's
modular architecture and its ability to scale linearly with increasing trading volumes.
The findings validate the effectiveness of the proposed method and the reliability of the developed
system in real-world conditions. Future research may focus on optimizing algorithms to reduce resource
consumption and integrating advanced risk management techniques to enhance performance.
Key words: automated trading, scalping, cryptocurrency, Binance API, algorithmic trading.

1. Introduction
The rapid development of the cryptocurrency market has created new challenges and
opportunities in the field of algorithmic trading. Among these challenges are the high volatility of
cryptocurrencies, the necessity of real-time data analysis, and the need for efficient strategies to
maximize profit in highly dynamic environments. Scalping strategies, which focus on generating
profit from small price fluctuations, are among the most promising approaches. However, their
implementation requires advanced algorithm optimization and seamless integration with trading
platforms.
This study addresses the scientific problem of developing efficient methods and tools for
automated cryptocurrency trading, focusing on improving the speed and accuracy of executing
© The Author(s) 2024. Published by Igor Sikorsky Kyiv Polytechnic Institute.
This is an Open Access article distributed under the terms of the license CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), which
permits re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
The automatic cryptocurrency trading system using a scalping strategy 113

trading operations. The relevance of this topic lies in its potential to enhance the effectiveness of
trading systems in a market that is increasingly dominated by automation and algorithmic strategies.

2. Literature review and problem statement

The implementation of automated trading systems using scalping strategies has been widely
studied in recent years, as the cryptocurrency market presents unique challenges and opportunities.
Existing literature provides valuable insights into the tools, algorithms, and methodologies employed
to address the core challenges of latency, risk management, scalability, and reliability in high-
frequency trading.
One of the foundational areas of research involves addressing latency in automated trading
systems. Studies such as those by Miles and Cliff [2] highlight the importance of real-time data
streaming technologies, like WebSocket, for minimizing delays in data retrieval and trade execution.
WebSocket's ability to facilitate low-latency, bidirectional communication between clients and
servers has made it a popular choice for trading platforms. Additionally, the use of in-memory data
structures, such as those provided by Redis, has been identified as a key mechanism for speeding up
temporary data storage and access during high-frequency trading operations. Yuxin Fan et al. [3]
extend this discussion by exploring machine learning-based approaches for optimizing real-time data
processing in high-frequency trading algorithms. Their study highlights the use of dynamic feature
selection mechanisms and lightweight neural networks, which not only reduce computational
complexity but also improve adaptability and efficiency in varying market conditions.
The challenge of risk management in scalping strategies has been explored extensively.
Research by Liu et al. [4] emphasizes the importance of leveraging technical indicators, such as the
Simple Moving Average (SMA) and Relative Strength Index (RSI), to make informed trading
decisions. These indicators allow for the evaluation of market trends and momentum, enabling traders
to identify optimal entry and exit points. However, the literature also notes the limitations of
traditional indicators in highly volatile markets. Abdelatif Hafid et al. [5] address these limitations by
presenting a machine learning-based classification model that integrates enhanced technical
indicators such as MACD, RSI, and Bollinger Bands. Their study demonstrates a buy/sell signal
accuracy of over 92%, providing a significant advantage in decision-making under volatile
conditions. Similarly, Anika Tahsin Meem [6] proposes a novel deep learning approach that applies
Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM), Bidirectional Long
Short-Term Memory (BiLSTM), and Gated Recurrent Unit (GRU) architectures to predict market
risk (volatility). This model demonstrates improved accuracy and generalization, making it
particularly useful for navigating unpredictable market scenarios.
Another critical aspect discussed in the literature is scalability, particularly in systems designed
to handle multiple cryptocurrency pairs and large volumes of data. Modular software architectures
have been a recurring theme in studies, with researchers advocating for the use of distributed systems
to improve scalability. For instance, Jericevic et al. [7] propose a modular approach where separate
components handle data collection, analysis, and trade execution, ensuring efficient distribution of
computational workloads. The use of cloud-based infrastructure is also frequently cited as a solution
for scaling trading systems to accommodate growing market demands.
Reliability is a recurring concern in automated trading systems, as technical disruptions can
lead to significant financial losses. Existing research underscores the importance of implementing
robust error-handling and recovery mechanisms. For example, the work of Byrd et al. [8] focuses on
strategies for ensuring system continuity during API outages or connection losses. Techniques such
as retry mechanisms and failover protocols have been highlighted as critical for maintaining
uninterrupted operations.
Blockchain metrics have also gained attention as valuable indicators in cryptocurrency trading.
Juan C. King et al. [9] propose the use of blockchain-derived metrics such as hash rate and mining
difficulty to identify trading signals. These metrics, inspired by the "Hash Ribbon," provide
statistically significant advantages in highly volatile markets, further enriching the decision-making
process.
114 Information, Computing and Intelligent Systems № 5, 2024

The comparison of scalping strategies with other trading approaches, such as trend-following,
has also been a focus of academic and industry research. Scalping is often described as more
profitable in volatile markets due to its high-frequency nature, as demonstrated by experiments
conducted by Sood et al. [10]. However, these studies also acknowledge the trade-offs, noting that
scalping requires higher computational resources and carries increased exposure to short-term market
risks compared to trend-following strategies. In this context, Shamima Nasrin Tumpa et al. [11]
explore Recurrent Neural Networks (RNN) for real-time price prediction and trading strategy
optimization, demonstrating their effectiveness in addressing the challenges posed by high market
volatility.
Despite the extensive research available, there are notable gaps in the literature. For instance,
while many studies address latency and risk management independently, there is limited work on
integrated solutions that simultaneously optimize multiple system aspects. Additionally, while the
advantages of using specific technical indicators are well-documented, there is a need for comparative
studies to evaluate their relative effectiveness in scalping versus other strategies.
This study builds upon the existing body of knowledge by addressing these gaps and proposing
a comprehensive solution that integrates real-time data processing, advanced risk management
techniques, modular architecture for scalability and reliability, and innovative blockchain metrics. By
combining insights from previous research with advanced methodologies, this study aims to advance
the field of automated cryptocurrency trading and provide a practical framework for implementing
efficient scalping strategies.
The core challenge in implementing scalping strategies lies in addressing latency issues.
Scalping requires systems to process market data and execute trades within milliseconds. Delays in
data collection, processing, or trade execution can result in significant losses. Furthermore, the high
frequency of trades exposes the system to increased risks due to rapid market fluctuations. Effective
mechanisms are necessary to mitigate these risks while maintaining the profitability of the strategy.
Another significant challenge is ensuring scalability. Scalping systems must be capable of
handling large volumes of market data and supporting multiple cryptocurrency pairs without
compromising performance. This necessitates a modular software architecture that can efficiently
distribute processing tasks and scale as trading volumes increase. Scalability is critical for
maintaining system reliability and ensuring that it can adapt to the growing demands of the
cryptocurrency market.
Reliability is also a key consideration. Automatic trading systems must operate continuously,
even during technical disruptions such as connection losses or API failures. Ensuring uninterrupted
operation requires robust error-handling mechanisms and automatic recovery features. Without these
capabilities, the system risks downtime, missed trading opportunities, and potential financial losses.
To address these challenges, this study focuses on developing an automatic trading system
optimized for scalping strategies. The proposed system utilizes advanced technologies such as
WebSocket for real-time data streaming, Redis for temporary data caching, and PostgreSQL for
reliable storage of historical market data. The system integrates technical indicators, including SMA,
RSI, and Bollinger Bands [12], to enable informed decision-making and improve trade execution
efficiency.
A review of existing literature reveals that while significant advancements have been made in
each of these areas, few studies have addressed the need for a comprehensive system that integrates
low-latency data processing, advanced risk management techniques, scalability, and reliability in the
context of cryptocurrency scalping.
This gap in the literature highlights the need for research dedicated to the development of an
integrated automatic cryptocurrency trading system optimized for scalping strategies. This study aims
to fill this gap by proposing a system that combines real-time data processing, advanced risk
management techniques, and modular scalability to address the unique challenges of high-frequency
trading in volatile cryptocurrency markets.
The automatic cryptocurrency trading system using a scalping strategy 115

3. The aim and objectives of the study

The cryptocurrency market is rapidly evolving, offering unique opportunities for implementing
innovative trading strategies. Scalping, a high-frequency trading method focused on small price
fluctuations, is one of the most effective approaches for maximizing profits in volatile markets.
However, its implementation poses significant challenges, including latency, risk management,
scalability, and reliability, which require advanced automatic trading systems.
The aim of the study is to develop and implement an automatic cryptocurrency trading system
optimized for scalping strategies. The proposed system will address key limitations of existing
solutions, such as latency and scalability issues, while ensuring robust risk management and reliable
operation. Additionally, the study aims to evaluate the performance of the developed system through
experimental testing and comparison with alternative trading approaches.
To achieve this aim, the following objectives are set:
1. Develop a novel scalping method that optimizes trade execution, risk management, and
scalability in dynamic cryptocurrency market conditions.
2. Implement a software solution integrating the proposed scalping method into an automatic
trading system, ensuring real-time data processing and seamless operation within cryptocurrency
exchanges.
3. Evaluate the effectiveness of the developed system through experimental testing, comparing
its performance with other trading strategies to provide practical recommendations for its application.

4. The study materials and development of a scalping method and software solution
4.1. Development of a scalping method
To better understand the decision-making process, pseudocode was developed to implement
the method. Below is an example of pseudocode for implementing a strategy based on Exponential
Moving Average (EMA) and Volume Weighted Average Price (VWAP):

Initialize parameters: EMA_period, VWAP_period, smoothing_factor, threshold

While market is open:

Get current price P_t and volume V_t
Calculate EMA using smoothing_factor and previous EMA value
Calculate VWAP based on cumulative price and volume

If (P_t > EMA and VWAP indicates strong buy signal):

Place Buy Order
Else If (P_t < EMA and VWAP indicates strong sell signal):
Place Sell Order
End If

Update historical price and volume data

End While

When modeling trading strategies, it is crucial to identify and adjust key parameters that
influence algorithm efficiency. The number of periods for the EMA determines the indicator's
sensitivity to price changes, with higher periods reducing sensitivity and lower periods increasing it.
The time period for VWAP calculation affects the accuracy of average price estimation, where shorter
periods are typically used in scalping to enhance entry and exit signal precision. The threshold for
entering or exiting positions defines the levels at which the bot executes buy or sell decisions,
typically fine-tuned through historical data analysis and backtesting. Additionally, the smoothing
factor (α) controls how quickly the EMA responds to price changes, enabling the bot to adapt to
varying market conditions effectively.
All these parameters require tuning based on backtesting results to ensure optimal bot
performance. Selecting the optimal values for each parameter can significantly impact trading
efficiency, risk reduction, and profit maximization.
116 Information, Computing and Intelligent Systems № 5, 2024

4.2. Development of a scalping method and software solution

The software architecture is based on a modular approach, enabling effective distribution of
functionality across various components as shown in Figure 1.

Fig.1. Software Architecture

The main modules of the system include: the Binance interaction module, the data collection
module, the data management module, the analysis and strategy module, the trading execution
module, and the historical data storage module. Each module performs a specific task, ensuring high
flexibility, scalability, and adaptability of the system.
– The Binance interaction module manages integration with the cryptocurrency exchange,
providing access to streaming and historical data via WebSocket and REST API.
– The data collection module structures the received information for further use in analytical
and trading processes.
– The data management module implements temporary storage via Redis and processes current
data using Pandas.
– The analysis and strategy module calculates technical indicators such as SMA, RSI, and
Bollinger Bands, making decisions on trading operations.
– The trading execution module interacts with the Binance API to place orders and monitor
trades.
– The historical data storage module uses PostgreSQL for reliable and structured data storage.
The bot's lifecycle consists of the stages of initialization, data collection, analysis, strategy
execution, and error handling as shown in Figure 2. During the initialization stage, API keys are
configured, a WebSocket connection to Binance is established, and necessary components like Redis
are prepared. Data collection ensures the retrieval of real-time and historical market data for analysis.
Market analysis is performed using indicators, enabling well-reasoned decisions on trade entry or
exit.
The strategy execution stage involves interaction with the exchange API for order placement
and logging of all operations. In the event of technical failures, such as connection loss, the bot
automatically resumes its operation, ensuring system stability.
The automatic cryptocurrency trading system using a scalping strategy 117

Fig.2. UML Activity Diagram of the Bot's Lifecycle

The technologies employed during development include PostgreSQL for historical data storage,
Redis for fast access to temporary data, WebSocket and REST API Binance for exchange integration,
and Python libraries (Pandas, NumPy, TA-Lib) ensures a robust and efficient architecture. for market
data processing. PostgreSQL provides reliable storage of large volumes of information necessary for
strategy analysis and optimization. Redis is used for data caching, significantly improving access
speed.
118 Information, Computing and Intelligent Systems № 5, 2024

This modular, technologically advanced approach ensures the scalability, reliability, and
efficiency of the trading system under real market conditions.
The proposed trading method based on a scalping strategy was implemented through software
that enables real-time market data analysis, decision-making regarding trade entry and exit, and
automatic trade execution. The bot operates on WebSocket technology, allowing data to be received
with minimal delay, which is critical for the successful application of scalping strategies. These
strategies rely on frequent trades with small profits, and their success depends on rapid responses to
market fluctuations.

5. Results of investigating the proposed method and software – optimized Scalping method
using EMA and VWAP Indicators
The experiment was conducted using the Bitcoin to Tether trading pair (BTC/USDT)
cryptocurrency pair, selected for its high liquidity, popularity, and significant volatility, which created
favorable conditions for testing the scalping bot. The collected data was analyzed using key indicators
such as the SMA, RSI, and Bollinger Bands. This ensured comprehensive market condition analysis
and allowed well-reasoned decisions regarding trade entry and exit.
The experiments were executed on a computer system with the following specifications:
Processor: Intel Core i7-10700K (8 cores, 16 threads, 3.8 GHz base clock, 5.1 GHz turbo boost).
– Memory: 32 GB DDR4 RAM (3200 MHz).
– Storage: 1 TB NVMe SSD (read/write speed: 3500/3000 MB/s).
– Operating System: Ubuntu 22.04 LTS.
– Network: Wired Ethernet connection with a speed of 1 Gbps.
This configuration ensured the system's ability to handle high-frequency trading operations,
process large volumes of market data, and execute trades with minimal latency.
To assess the effectiveness of the scalping method, it was compared with a trend-following
strategy. The trend bot focused on long-term trades and used indicators like SMA and Moving
Average Convergence Divergence (MACD). Unlike the scalping bot, the trend bot conducted fewer
trades, focusing on stable market trends. Experiments showed that the scalping bot achieved higher
overall profits due to more frequent trades, while the trend bot provided more stable profits per trade
under steady market conditions as shown in Table 1.

Table 1. Comparison of Scalping and Trend Bot Performance

Parameter Scalping Bot Trend Bot

Test Duration 2 hours 2 hours

Number of Trades 15 7
Number of Successful Trades 13 5
Total Profit (USDT) 120 85
Average Profit (%) 1.2% 0.85%
Maximum Profit (USDT) 25 40
Minimum Profit (USDT) 5 10
Maximum Loss (USDT) –8 –12
Position Holding Time (min) 4 20

The scalping bot’s higher trade frequency allowed it to capitalize on rapid price fluctuations,
demonstrating its strength in volatile markets. Its ability to adapt quickly to changing conditions was
evident in its low average position holding time and high success rate. The trend bot, on the other
hand, was more conservative, aiming for fewer but more significant trades, which is reflected in its
higher maximum profit per trade.
The automatic cryptocurrency trading system using a scalping strategy 119

An evaluation of memory usage during the experiment showed that the scalping bot’s resource
requirements scaled linearly with the number of cryptocurrency pairs processed. This was an essential
observation as it underlined the system's scalability. For instance, the bot consumed 120 MB of
memory for a single pair and scaled up to 2100 MB when handling 20 pairs, as shown in Table 2.

Table 2. Memory Usage During Scalping Bot Operation

Memory
Number of Base Memory Cache Memory Total Memory
During Peak
Pairs Usage (MB) Usage (MB) (MB)
Load (MB)
1 100 80 20 120
5 450 350 100 650
10 900 700 200 1300
15 1400 1100 300 1700
20 1900 1400 500 2100

For a single cryptocurrency pair, the scalping bot requires a total of 120 MB of memory. This
includes 80 MB allocated to core operations, such as maintaining the bot’s functionality and
processing real-time data streams, and 20 MB for cache memory, which is utilized for storing
frequently accessed data temporarily. The efficient allocation of memory resources ensures the bot
operates smoothly without unnecessary overhead, making it well-suited for low-volume trading
scenarios where only a single pair is analyzed.
As the number of cryptocurrency pairs being processed increases, the bot’s memory
consumption grows linearly. This consistent scaling demonstrates the modular design of the system,
allowing for predictable increases in resource usage with each additional pair. For example, when
managing 5 pairs, the total memory usage rises to 650 MB, comprising 350 MB for base memory and
100 MB for cache. At 10 pairs, memory consumption doubles to 1300 MB, with 700 MB dedicated
to base operations and 200 MB allocated for cache. At its maximum tested capacity of 20 pairs, the
bot requires 2100 MB of memory, including 1400 MB for base functions and 500 MB for cache. This
linear growth pattern highlights the bot's ability to handle increasingly complex workloads while
maintaining stable performance.
The predictable nature of memory consumption is a key factor in the bot’s scalability. Linear
scaling allows traders to effectively plan their system resources based on expected trading volumes,
ensuring smooth operations across varying scales. Whether used for single-pair trading or high-
volume scenarios involving multiple pairs, the bot demonstrates an ability to adapt without significant
performance degradation.
To further evaluate the scalability of the bot, several key performance metrics were measured,
including latency, average signal processing time, Computational tasks were executed using a Central
Processing Unit (CPU) usage, and memory consumption. Latency measures the delay between
receiving a market signal and the bot’s response, a critical metric for scalping strategies that rely on
rapid execution to capture small price fluctuations. Average Signal Processing Time reflects the bot’s
efficiency in analyzing incoming data and making trading decisions, ensuring responsiveness to
dynamic market conditions.
CPU usage was another important factor, as it indicates the computational load on the processor
during operation. Efficient CPU utilization ensures the bot can maintain high performance without
overburdening the system, even when processing multiple pairs. Finally, memory consumption
provides insight into the bot’s resource requirements, highlighting its capacity to scale smoothly as
trading activity intensifies.
These metrics, as summarized in Table 3, underscore the bot's robust architecture and its ability
to maintain performance under varying workloads. By effectively managing memory, processing
time, and computational resources, the scalping bot proves to be a scalable and reliable tool for
120 Information, Computing and Intelligent Systems № 5, 2024

traders, whether operating in low or high-volume trading environments. This adaptability ensures that
the bot is suitable for real-world applications, providing both efficiency and stability in volatile
cryptocurrency markets.

Table 3. Scalability Metrics of the Scalping Bot

Average Signal
Number of
Latency (ms) Processing CPU Usage (%) Memory (MB)
Pairs
Time (ms)
1 15 30 5 120
5 20 40 15 650
10 25 50 25 1300
15 35 60 40 1700
20 50 75 55 2100

The scalability of the scalping bot was evaluated using key performance metrics, including
latency, average signal processing time, CPU usage, and memory consumption. These metrics,
detailed in Table 3, were measured for varying loads, starting from a single cryptocurrency pair up to
a maximum of 20 pairs. This assessment provided valuable insights into the bot's ability to maintain
stable performance as the operational load increased.
Latency, a critical metric for scalping strategies, measures the time delay between receiving a
market signal and the bot's response. For a single pair, latency was recorded at 15 ms. As the number
of pairs increased, latency grew steadily, reaching 50 ms when processing 20 pairs. Despite the
incremental increase, latency remained within acceptable limits, ensuring the bot could respond
quickly to market changes—a necessity for high-frequency trading.
Average signal processing time reflects the time taken to analyze incoming market data and
make trading decisions. For a single pair, the bot required 30 ms, and this time increased
proportionally to 75 ms when handling 20 pairs. The gradual rise in processing time demonstrates
efficient computational resource allocation, enabling the bot to handle complex data streams without
significant delays. This consistent performance underscores the bot’s ability to operate effectively in
fast-paced trading environments.
CPU usage was also analyzed to understand the bot’s computational demands. At minimal load,
the bot used only 5% of the CPU, demonstrating its optimized design for low-volume trading. As the
number of pairs increased, CPU usage rose proportionally, peaking at 55% for 20 pairs. This efficient
scaling ensured that the bot remained within safe operational thresholds, avoiding system overload
while maintaining stable performance.
Memory consumption increased linearly with the number of pairs, as previously noted, growing
from 120 MB for a single pair to 2100 MB for 20 pairs. This predictable scaling of memory usage
highlights the bot's modular architecture, allowing for effective resource management and planning
as trading volumes expand. The bot’s ability to maintain linear growth in resource consumption
further validates its scalability.
To analyze real-time performance, Table 4 breaks down response times into signal receiving,
data processing, and order placement. For a single pair, the total response time was 50 ms, divided
into 10 ms for signal receiving, 20 ms for processing, and 20 ms for order placement. At maximum
capacity with 20 pairs, the total response time increased to 105 ms, with 20 ms for signal receiving,
50 ms for processing, and 35 ms for order placement. These results indicate that the bot maintained
rapid response times even under maximum load, ensuring minimal delays in trade execution.
In summary, the scalping bot demonstrated strong scalability and responsiveness, maintaining
efficient performance across varying levels of trading activity. The latency, processing time, CPU
usage, and memory consumption metrics all showed proportional scaling, ensuring the bot’s
adaptability to different trading demands. This robust performance makes the bot a reliable tool for
The automatic cryptocurrency trading system using a scalping strategy 121

traders, capable of handling both small-scale and large-scale operations in volatile cryptocurrency
markets.

Table 4. Response Time for Scalping Bot Operations

Signal Order
Number of Processing Total Response
Receiving Time Placement Time
Pairs Time (ms) Time (ms)
(ms) (ms)
1 10 20 20 50
5 12 28 20 60
10 15 35 25 75
15 18 40 30 88
20 20 50 35 105

Overall data processing speed remains stable even under maximum load conditions.
Thus, the scalping bot demonstrates high efficiency and stability, even under significant loads.
This makes it a versatile tool suitable for real-world market conditions.

6. Analysis of the obtained results

The results of the study provide strong evidence of the scalability, efficiency, and adaptability
of the scalping bot in handling high-frequency cryptocurrency trading. These findings are explained
by the system's modular architecture and its ability to efficiently allocate computational resources,
even under significant workloads. The analysis of performance metrics, such as memory
consumption, CPU usage, latency, and response times, highlights the bot's ability to maintain stable
operation and reliable performance across various trading volumes.
One of the critical findings is the bot's scalability, which is explained by the linear nature of
resource consumption observed during the experiments. The memory usage, for example, increased
predictably from 120 MB for a single cryptocurrency pair to 2100 MB for 20 pairs. This predictable
scaling underscores the effectiveness of the bot’s design in managing growing workloads without
overwhelming system resources. Similarly, CPU usage remained proportional to the number of pairs
processed, peaking at 55% under maximum load. This efficient resource utilization ensures that the
bot can operate smoothly even during high trading volumes, making it highly suitable for real-world
applications.
The latency and response time metrics further validate the bot's adaptability to dynamic market
conditions. The experiments demonstrated that latency increased from 15 ms for a single pair to 50
ms for 20 pairs, while the total response time remained under 105 ms at maximum load. These results
suggest that the bot can execute trades quickly, an essential feature for scalping strategies that rely on
capturing small price fluctuations within volatile markets. These outcomes are attributed to the bot’s
streamlined decision-making processes and its use of real-time data streams.
While the scalping bot's rapid trade execution resulted in higher overall profitability compared
to a trend-following strategy, this strength also introduced higher risks. The increased trade frequency
amplified the overall risk exposure, despite the lower maximum loss per trade for the scalping bot (–
8 USDT) compared to the trend bot (–12 USDT). This trade-off between profitability and risk
underscores the importance of robust risk management mechanisms in automated trading systems.
Future improvements could include the integration of machine learning models to dynamically adjust
risk parameters based on market conditions.
The study also revealed potential limitations of the scalping bot. The reliance on linear resource
scaling, while effective for moderate workloads, may pose challenges when operating under extreme
conditions or with limited hardware capacity. Optimizing algorithms to reduce memory and CPU
demands would enable the bot to process more pairs or operate on less powerful hardware.
Additionally, the experiments focused primarily on the BTC/USDT pair, which, while highly liquid
122 Information, Computing and Intelligent Systems № 5, 2024

and volatile, may not represent the bot's performance across less active trading pairs. Expanding the
scope of testing to include a broader range of cryptocurrencies and market scenarios would provide
a more comprehensive understanding of the bot’s capabilities.
The bot’s reliability is another key aspect highlighted by the results. Error-handling
mechanisms ensured continuous operation, even during peak loads, preventing disruptions that could
have resulted in missed trading opportunities. However, the system’s dependence on stable network
connections and API availability introduces potential vulnerabilities. Future iterations could
incorporate redundancy features, such as fallback mechanisms or integration with multiple
exchanges, to enhance the system’s robustness in real-world trading environments.
In conclusion, the findings of this study demonstrate the practicality and reliability of the
scalping bot for professional trading applications. Its ability to handle multiple cryptocurrency pairs,
maintain rapid response times, and scale efficiently makes it a valuable tool for institutional investors
and high-frequency traders. Future research should focus on optimizing resource utilization,
integrating advanced risk management techniques, and testing the bot in live trading scenarios to
further validate its performance and identify additional areas for improvement. These enhancements
would ensure that the scalping bot remains competitive in the fast-evolving landscape of
cryptocurrency trading.

Conclusions
The study successfully developed an innovative scalping strategy optimized for trade execution,
risk management, and scalability in dynamic cryptocurrency market conditions. The implemented
automated trading system demonstrated the ability to process real-time market data with minimal
latency, ranging from 15 ms to 50 ms, enabling rapid decision-making and high-frequency trading.
The system’s modular design ensured linear scalability of resource usage, with memory consumption
increasing predictably from 120 MB for a single pair to 2100 MB when handling 20 pairs. This
confirmed the system’s capability to maintain stable performance under varying workloads.
Experimental testing revealed the advantages of the scalping strategy compared to a trend-
following approach. Over a two-hour test period, the scalping bot achieved higher overall
profitability, conducting 15 trades (13 successful) with a total profit of 120 USDT. In contrast, the
trend bot, focused on fewer trades, generated a profit of 85 USDT. The scalping bot’s higher trade
frequency allowed it to capitalize on short-term price fluctuations, while its low average position
holding time of 4 minutes highlighted its adaptability to volatile market conditions. However, the
trend bot achieved a higher maximum profit per trade under steady market conditions, demonstrating
its more conservative approach.
The scalability of the system was validated through key performance metrics, including
memory usage, latency, CPU utilization, and signal processing times. Results showed predictable and
proportional increases in resource usage as the number of cryptocurrency pairs increased, ensuring
stable performance without significant degradation. The system maintained acceptable response
times, growing from 50 ms for a single pair to 105 ms for 20 pairs, while CPU usage scaled efficiently
from 5% to 55%. These findings highlight the bot’s ability to handle both small-scale and high-
volume trading operations effectively.
The practical value of the developed system lies in its efficiency, reliability, and scalability,
making it a versatile tool for high-frequency traders operating in volatile markets. Its modular
architecture allows for smooth integration with real-world trading platforms, while its ability to adapt
to dynamic market conditions ensures robust performance. The experimental results confirm that the
system provides a competitive advantage in cryptocurrency trading, offering both high profitability
and resource efficiency.

References
[1] B. Miles and D. Cliff, “A cloud-native globally distributed financial exchange simulator for
studying real-world trading-latency issues at planetary scale,” arXiv preprint, arXiv:1909.12926,
2019. [Online]. Available: https://doi.org/10.48550/arXiv.1909.12926
The automatic cryptocurrency trading system using a scalping strategy 123

[2] Y. Fan, Z. Hu, L. Fu, Y. Cheng, L. Wang, and Y. Wang, “Optimizing real-time data processing
in high-frequency trading algorithms using machine learning,” arXiv preprint, arXiv:2412.01062,
2024. [Online]. Available: https://doi.org/10.48550/arXiv.2412.01062
[3] X. Liu et al., “FinRL: A deep reinforcement learning library for automated trading in quantitative
finance,” arXiv preprint, arXiv:2011.09607, 2020. [Online]. Available:
https://doi.org/10.48550/arXiv.2011.09607
[4] A. Hafid, M. Rahouti, L. Kong, M. Ebrahim, and M. A. Serhani, “Predicting Bitcoin market trends
with enhanced technical indicator integration and classification models,” arXiv preprint,
arXiv:2410.06935, 2024. [Online]. Available: https://doi.org/10.48550/arXiv.2410.06935
[5] A. T. Meem, “A deep learning approach to predict the fall of cryptocurrency long before its actual
fall,” arXiv preprint, arXiv:2411.13615, 2024. [Online]. Available:
https://doi.org/10.48550/arXiv.2411.13615
[6] I. Jericevic, D. Sing, and T. Gebbie, “CoinTossX: An open-source low-latency high-throughput
matching engine,” arXiv preprint, arXiv:2102.10925, 2021. [Online]. Available:
https://doi.org/10.48550/arXiv.2102.10925
[7] D. Byrd et al., “The importance of low latency to order book imbalance trading strategies,” arXiv
preprint, arXiv:2006.08682, 2020. [Online]. Available:
https://doi.org/10.48550/arXiv.2006.08682
[8] J. C. King, R. Dale, and J. M. Amigó, “Blockchain metrics and indicators in cryptocurrency
trading,” arXiv preprint, arXiv:2403.00770, 2024. [Online]. Available:
https://doi.org/10.48550/arXiv.2403.00770
[9] S. Sood et al., “Evaluation of reinforcement learning techniques for trading on a simulated stock
market,” arXiv preprint, arXiv:2309.03202, 2023. [Online]. Available:
https://doi.org/10.48550/arXiv.2309.03202
[10] S. N. Tumpa and K. D. G. Maduranga, “Utilizing RNN for real-time cryptocurrency price
prediction and trading strategy optimization,” arXiv preprint, arXiv:2411.05829, 2024. [Online].
Available: https://doi.org/10.48550/arXiv.2411.05829
[11] J. Bollinger, “Bollinger on Bollinger bands,” McGraw Hill Professional, New York, 227 р.,
2002.
124 Information, Computing and Intelligent Systems № 5, 2024

УДК 004.94

АВТОМАТИЧНА СИСТЕМА ТОРГІВЛІ КРИПТОВАЛЮТОЮ

З ВИКОРИСТАННЯМ СКАЛЬПІНГОВОЇ СТРАТЕГІЇ

Бераудо Еліза
Національний технічний університет України
«Київський політехнічний інститут імені Ігоря Сікорського», Київ, Україна
https://orcid.org/0000-0001-7550-3620

Олійник Юрій
Національний технічний університет України
«Київський політехнічний інститут імені Ігоря Сікорського», Київ, Україна
https://orcid.org/ 0000-0002-7408-4927

У дослідженні розглянуто питання розробки та впровадження автоматизованої системи

для скальпінгових стратегій на криптовалютному ринку. Скальпінг є високочастотною
стратегією, що орієнтована на отримання прибутку від незначних цінових коливань. Основна
мета роботи – створення автоматизованого торгового бота, що вирішує ключові проблеми,
пов’язані з латентністю, управлінням ризиками, масштабованістю та надійністю у реальних
ринкових умовах. Для досягнення цієї мети було сформульовано завдання: розробити новий
метод скальпінгу, реалізувати програмне рішення для його інтеграції у систему
автоматизованої торгівлі та оцінити ефективність роботи системи за допомогою
експериментального тестування.
Методологія дослідження базується на використанні технічних індикаторів, таких як
експоненційне ковзне середнє (Exponential Moving Average, EMA) та об’ємнозважена середня
ціна (Volume Weighted Average Price, VWAP). Псевдокод методики було розроблено для
кращого розуміння процесу прийняття рішень, включаючи параметри згладжування, часові
періоди та порогові значення для входу та виходу з позицій. Реалізована архітектура системи
включає такі модулі: модуль інтеграції з біржею Binance, модуль збору та управління даними,
модуль аналізу та стратегії, модуль виконання торгових операцій і зберігання історичних
даних. Для розробки системи використовувалися технології PostgreSQL, Redis, WebSocket та
бібліотеки Python (Pandas, NumPy, TA-Lib).
Експерименти проводилися на парі BTC/USDT, яка характеризується високою
ліквідністю та волатильністю. Система тестувалася на обладнанні з процесором
Intel Core i7-10700 K, 32 ГБ оперативної пам’яті та підключенням до мережі зі швидкістю 1
Гбіт/с. Порівняльний аналіз скальпінгової стратегії та трендової стратегії показав перевагу
першої в умовах високої волатильності ринку: бот виконав 15 угод (13 успішних) із загальним
прибутком 120 USDT протягом двох годин.
Було оцінено ключові показники продуктивності, такі як затримка (15–50 мс), час
обробки сигналів, завантаженість CPU (5–55%) та використання пам’яті (120–2100 МБ).
Лінійне масштабування ресурсів підтвердило модульність і гнучкість архітектури системи, що
дозволяє ефективно керувати збільшенням обсягів торгівлі.
Отримані результати свідчать про ефективність запропонованого методу та надійність
розробленої системи у реальних умовах. Подальші дослідження можуть бути спрямовані на
оптимізацію алгоритмів для зменшення споживання ресурсів та інтеграцію передових методів
управління ризиками.
Ключові слова: автоматична торгівля, скальпінг, криптовалюта, Binance API, алгоритмічна
торгівля.