Evaluation of Dynamic Cointegration-Based Pairs.pdf

(827 KB) Pobierz
Evaluation of Dynamic Cointegration-Based Pairs
Trading Strategy in the Cryptocurrency Market
M
ASOOD
T
ADI
CY Tech, CY Cergy Paris University, Cergy, France
tadimasood@cy-tech.fr
I
RINA
K
ORTCHEMSKI
arXiv:2109.10662v1 [q-fin.TR] 22 Sep 2021
CY Tech, CY Cergy Paris University, Cergy, France
ik@cy-tech.fr
Abstract
Purpose
— This research aims to demonstrate a dynamic cointegration-based pairs
trading strategy, including an optimal look-back window framework in the
cryptocurrency market, and evaluate its return and risk by applying three different
scenarios.
Design/methodology/approach
— We employ the Engle-Granger methodology, the
Kapetanios-Snell-Shin (KSS) test, and the Johansen test as cointegration tests in
different scenarios. We calibrate the mean-reversion speed of the Ornstein-Uhlenbeck
process to obtain the half-life used for the asset selection phase and look-back window
estimation.
Findings
— By considering the main limitations in the market microstructure, our
strategy exceeds the naive buy-and-hold approach in the Bitmex exchange. Another
significant finding is that we implement a numerous collection of cryptocurrency coins
to formulate the model’s spread, which improves the risk-adjusted profitability of the
pairs trading strategy. Besides, the strategy’s maximum drawdown level is reasonably
low, which makes it useful to be deployed. The results also indicate that a class of
coins has better potential arbitrage opportunities than others.
Originality/value
— This research has some noticeable advantages, making it stand out
from similar studies in the cryptocurrency market. First is the accuracy of data in which
minute-binned data create the signals in the formation period. Besides, to backtest the
strategy during the trading period, we simulate the trading signals using best bid/ask
quotes and market trades. We exclusively take the order execution into account when
the asset size is already available at its quoted price (with one or more period gaps after
signal generation). This action makes the backtesting much more realistic.
1
Keywords
— Arbitrage opportunity, Pairs trading strategy, Basket trading,
Cointegration, Mean reversion, Cryptocurrency market
Paper type
— Research paper
2
1
Introduction
A cryptocurrency is volatile digital security designed to operate as a tool of exchange
that uses secure cryptography (Gandal and Halaburda
2016).
Bitcoin is the first
cryptocurrency that is issued in 2009. After the bitcoin announcement, a lot of
alternative cryptocurrencies (altcoins) have been created.
The total market
capitalization of the cryptocurrency market is more than
341
billion dollars on
30/09/2020 (See Coinmarketcap.com). Deploying algorithmic trading strategies is
developing over time in the Cryptocurrency market, accelerating tradings to maximize
profits.
One of the most well-known strategies among different algorithmic trading methods is
the statistical arbitrage strategy. Statistical arbitrage is a profitable situation stemming
from pricing inefficiencies among financial markets. Statistical arbitrage is not a real
arbitrage opportunity, but it is merely possible to obtain profit applying past statistics
(Aldridge
2013).
In fact, there are two different potential arbitrage opportunities in the
cryptocurrency market; the exchange to exchange arbitrage and the statistical arbitrage.
The exchange to exchange arbitrage has a potential profit, but it is quite risky, and there
are many challenges to deploy it. Instead, the statistical arbitrage opportunities have the
same potential profits without the same risks as the former one (Pritchard
2018).
Statistical arbitrage strategies are often deployed based on mean reversion property, but
they can also be designed using other factors. Pairs trading is the commonly recognized
statistical arbitrage strategy that involves identifying pairs of securities whose prices
tend to move mutually. Whenever the relationship between financial securities behaves
abnormally, the pair would be traded. Then the open positions will be closed when the
unusual behavior of pairs reverts to their normal mode (Vidyamurthy
2004).
According to Krauss (2017), pairs trading strategy is a two-step process. The first step,
which is called the formation period, attempts to find two or more securities whose
prices move together historically. In the second step, which is the trading period, we
seek abnormalities with their price movement to profit from statistical arbitrage
opportunities.
There are two general approaches to find appropriate pairs of assets in the formation
period: the heuristic approach and the statistical approach. The heuristic approach,
which is regularly called the distance approach, is more straightforward than the
statistical approach, which the latter is based on the cointegration concept. In the
trading period, we can combine our strategy with different mathematical tools such as
stochastic processes, stochastic control, machine learning, and other methods to
improve the results (ibid.).
Compared to the leading financial markets, such as the stock market and fixed income
market, limited research has been conducted on statistical arbitrage strategies in the
3
cryptocurrency market. In the next section, we review some papers concerning these
strategies frequently studied in the cryptocurrency market and explain some of their
practical weaknesses and drawbacks.
2
Literature Review
The distance approach is the basic researched framework introduced by Gatev,
Goetzmann, and Rouwenhorst (2006). This method is based on the minimum squared
distance of the normalized price of assets. The distance between normalized prices is
called spread. In order to construct a measurable scale, assets should become normal
first. To this aim, asset prices are divided into their initial value, and then the spread is
obtained by taking the difference between normalized prices. In the formation period,
the top first pairs with the minimum historic sum of squared distances between
normalized prices are considered in a subsequent trading period (ibid.).
According to Perlin (2009), the price series can be normalized first based on their
historical mean and historical volatility, and then the spread of each pair can be
constructed in the same way. Finally, in the trading period, the pairs arranged in their
ascending orders can be picked using the top first pairs of the list for pairs formation.
Simple non-parametric threshold rules are used to trigger trading signals. This
threshold and can be two historical standard deviations of normalized spread price.
The main findings establish distance pairs trading as profitable across different markets,
asset classes, and time frames. Driaunys et al. (2014) studied a pairs trading strategy
at the natural gas futures market based on the distance approach. In their research,
pairs of two futures contracts of the same underlying asset are selected, i.e., natural gas
with different maturities, which are the most liquid, close to expiration, and therefore
they are correlated contracts of natural gas futures. The contract with bid prices is
shorted through the trading period, Whenever the distance of the given
d
is reached.
Besides, the long position of the contract with the asking price is taken. They backtest
their model with different moving windows and different thresholds and realize that the
higher values of
d
works better and generate fewer trades and make the model more
stable.
Driaunys et al. (ibid.)’s research has some weaknesses in practice. First, the assumption
of no transaction cost makes their result unreliable. Second, they did not calculate the
risk-adjusted return of different scenarios. So, the performance of the research’s strategy
is questionable. Furthermore, they did not compare their method with a naive strategy,
i.e., buy-and-hold strategy, and besides, the required investment to deploy the strategy
is not determined.
Other researches are based on the statistical approach. This approach identifies two
4
or more time-series combined to form a long-term equilibrium relationship, although
the time series themselves may have a non-stationary trend. There are some research
papers based on the cointegration concept in the cryptocurrency market. Broek and
Sharif (2018) selected a set of cryptocurrency coins and split them into four main sectors
that depend on the coins’ fundamental features. Then, by examining unit-root tests,
they achieved a cointegration relationship among coins in each group. Finally, they
concluded that implementing a pairs trading strategy could be profitable due to arbitrage
opportunities. Nevertheless, ignoring the transaction costs and having a bias selection
make this paper’s profitability results unreliable.
Leung and Nguyen (2019) constructed cointegrated cryptocurrency pairs using three
different unit-root tests. Using the daily prices of four cryptocurrency coins from
Coinbase exchange, they introduced the ordinary least squares model to build a
cointegrated combination of coins and set the p-values of the estimated coefficients
less than 1%. The authors backtested the strategy with five different entry/exit
threshold levels. They realized that with the threshold set at
1.5
standard deviation, the
strategy is optimal. Furthermore, they incorporated the stop-loss exit and trailing
stop-loss exit possibilities, which lowered the profit return.
Kakushadze and Yu (2019) proposed the momentum factor statistical arbitrage
methodology based on a dollar-neutral mean-reversion strategy. Their method is to
short a fixed level of Bitcoin and keep it throughout the trading period and long
multiple altcoins, which varies every day depending on their momentum values. They
realized that low liquid altcoins have a better mean-reverting feature. Furthermore,
they found that if yesterday’s momentum of an altcoin is positive, it is expected to
trade it higher today and vise versa.
Pritchard (2018) built the strategy based on statistical tests as well as technical analysis
indicators. He applied different statistical tests such as the augmented Dickey-Fuller
test, the Hurst exponent, and the Johansen test. He achieved that each coin is not
mean-reverting, but a linear mean-reverting strategy can be implemented by
constructing the normalized deviation of price from its moving average. In his study,
the volume is considered the most critical barrier to utilize arbitrage opportunities in
the cryptocurrency market.
3
Market Structure
Numerous cryptocurrency exchanges allow their customers to trade cryptocurrencies
against other assets, such as conventional fiat money or other digital currencies.
BitMEX is one of the most well-known cryptocurrency exchanges with a Peer-to-Peer
Trading Platform, which offers leveraged contracts bought and sold in Bitcoin. This
5

Plik z chomika:

Inne pliki z tego folderu:

Inne foldery tego chomika:

Zgłoś jeśli naruszono regulamin