ACADEMIC RESEARCH

1. BITCOIN MARKET MICROSTRUCTURE

Hong Kee Sul

Joon Chae

Hyoung-Goo Kang

Kibeom Seong

Kyoung-hun Bae

Assistant Professor, College of Business Administration, Hanyang University

Professor, College of Business Administration, Seoul National University

Associate Professor, College of Business Administration, Hanyang University

Co-Founder / Portfolio Manager, Entropy Trading Group

Assistant Professor, Chung-Ang University

1-1. How does it differ from stock market?

Full Report

This study analyzes the relationship among bitcoin order imbalances, liquidity, and returns using the BitMEX intraday trading data from April 2017 to July 2018. The bitcoin order imbalances are strongly and positively correlated with both liquidity measures and bitcoin returns. The kurtosis and skewness of bitcoin returns are extremely large compared to those of stock returns, implying large intraday volatility. Variance of bitcoin returns are very high compared to their mean and correlations between microstructure variables also tend to be high. Such patterns in the bitcoin market are much more salient than those in the stock market. However, in contrast to the stock market, autocorrelations in order imbalance measures are nearly zero in the bitcoin market.

1-2. 24 hours of Bitcoin

Full Report

This study examines the intraday patterns of the bitcoin market using BitMEX trading and quote data from April 2017 to July 2018. The largest magnitude of order imbalance measures is observed at 11:00 Korea Standard Time (KST) and at 13:00 KST, respectively. The correlations between returns and order imbalance measures are significant and positive at all times. The return and volatility of bitcoin price tend to increase from 10:00 KST to 16:00 KST while decreasing from 00:00 KST to 06:00 KST. Total number and dollar volume of transactions move similarly. Market returns show strong rising tendency during the lunch time in Korea between 11:00 and 13:00. Especially return and other trading activities surge during 11:00-13:00, the Korean lunch time. In sum, global bitcoin trading activities are largely in sync with the Korean stock market.

2. BITCOIN RETURNS AND ORDER IMBALANCE

Hong Kee Sul

Joon Chae

Hyoung-Goo Kang

Kibeom Seong

Kyoung-hun Bae

Assistant Professor, College of Business Administration, Hanyang University

Professor, College of Business Administration, Seoul National University

Associate Professor, College of Business Administration, Hanyang University

Co-Founder / Portfolio Manager, Entropy Trading Group

Assistant Professor, Chung-Ang University

2-1. What determines bitcoin order imbalance?

Full Report

This study examines the relationship between bitcoin order imbalance and returns using BitMEX trading and quote data from April 2017 to July 2018. The analysis shows that order imbalances are negatively correlated with lagged returns, implying that bitcoin investors are contrarian during the sample period. This pattern is more salient when bitcoin returns decline than when the returns rise. The subperiod analysis is more nuanced. In period 1 (25 April 2017 to 15 September 2017), investors have a strong tendency to buy when returns decline. After period 1 (16 September 2017 to 12 July 2018) in which volatility is high, such tendency becomes weakened. If we normalize order imbalance with the number of transactions, investors are contrarians in period 1 and period 4 (8 February 2018 to 12 July 2018). The hourly analysis finds that order imbalances are particularly high from 11:00 to 12:00 and 21:00 to 22:00 (KST).

2-2. Does bitcoin order imbalance predict returns?

3. CRYPTO-ASSET, NLP AND MACHINE LEARNING

Kibeom Seong

Juntae Yoon

R&D Header / Vice President, Daumsoft Inc.

Co-Founder / Portfolio Manager, Entropy Trading Group

3-1. Investor sentiment in Crypto message boards

As an extension of Koo et al. (2018), we investigate the relationship between Bitcoin price and natural language processed emotion data extracted from an online chat board at BitMEX, one of the biggest cryptocurrency exchanges. Combining emotion data with high-frequency quote and trade data from BitMEX, Binance, and Coinone between June 18, 2018 and November 27, 2018, vector autoregression (VAR) is performed with five minute sampling period to analyze complex interactions of emotions and Bitcoin prices. Our results suggest that Bitcoin return, volume, Korean premium, and volatility over the next five minute can be predicted by certain types of emotions expressed in the BitMEX chat board. We also develop a simple investment strategy utilizing some emotions significantly correlated with Bitcoin returns and show that this strategy is profitable.