Crypto with Style!
Part2 - Daily Study
AlphabetIM Research, April 2023
In Crypto with Style! Part1, we discussed the application of factor investing to cryptocurrency markets. Literature [1,2,3] suggests the use of a three to five-factor model to describe the returns of cryptocurrencies, including styles such as size, momentum, network adoption, and value. We chose to focus on top market cap and attempt to identify potential factors on a weekly basis, investigating their individual investing interest and isolating the most relevant ones through PCA and feature selection methods. We finally suggested a three-factor model that includes market, value, and liquidity factors as the best fit for our universe.
We apply here the same process on daily returns to examine if factors have the same explanatory power in returns decompositions. Obviously, we expect notable difference with the weekly portfolio and it will help to highlight importance of faster factors that drive the coins on the shorter daily horizon.
1. Data and Factors definition
We use the same data set as the weekly study coming from the public Binance api (1st November 2020 to 31st January 2023) for our universe of 62 top market caps*. And we use IntoTheBlock analytics to constitute Flow, Sentiment, Network and Value factors.
We define now 11 potential daily factors: Market, Momentum, Reversal (note that we add this factor compared to the weekly study as reversion effects are more important on shorter term), LowVol, Size, Liquidity, Positioning, Flow, Sentiment, Network and Value as follow:
- Market: we take our universe and weight the coins by their market capitalizations.
- Momentum: we choose a mix of 7d-reurn, 21d-return, 7d-return/7d-volatility, 7d-ewma return of residuals (residuals come from the regression of daily return against bitcoin with a 21d-beta).
- Reversal: we select 1d-return, 14d-RSI and 7d-skewness of distribution. We use this factor in descending order.
- LowVol: a mix of 7d-realized volatility and 21d-Beta.
- Size: market capitalization.
- Liquidity: a mix of 7d-volume and (7d-volume — 7d-volume of previous week)/30d-volume.
- Positioning: a mix of premium defined as (perpetual price-spot price)/spot price and premium speed = premium — premium of previous day.
- Flow: we take both log of whales net flow addressing accumulation or reduction of on-chain holding by whales; and log of exchange net flow, which gives an indicator of on-chain flows in and out centralized exchanges.
- Sentiment: we use social media sentiment namely Twitter and Telegram with their relative speed as provided by IntoTheBlock.
- Network: as suggested by Cong & al in [3] we will use first daily difference of log values of total addresses with balance and the first daily difference of log values of total transaction volume on chain in USD.
- Value: we specify a Value factor like [3] where we compare the coin price with its network size through a user-to-market ratio (where user is approximated by the total addresses with balance) and its network activity via a transaction-to-market ratio (where the transaction is the aggregate volume of transactions recorded on-chain).
We then windsorize all the features by removing 2-%iles extreme values and normalize the remaining features cross-coins daily. We give below the daily performance mean for each quintile and the t-stat associated:
Note that Momentum, Reversal, Positioning and Flow factors exhibit monotonous behaviors with a good differentiation between mean performances. Value, Sentiment and Network are not that linear but seem interesting too.
We can now form the High-Low portfolio for every feature (and this portfolio will then be called factor). We are exhibiting below the mean and its t-stat for all our factors:
We find again that Momentum, Reversal and Positioning stand out clearly as factors with the highest t-stat, with Flow, Value, Sentiment and Network lagging. LowVol, Size and Liquidity seem irrelevant. We illustrate below cumulative performances of the different factors:
And plot a bubble scatterplot on (volatility, return) axis with the size of the bubble proportional to the Sharpe Ratio of a given factor risk premium:
Compared to the weekly study, factors are more heterogeneous with a dominance from technical factors like Momentum and Reversal, which drive more the returns on the short term compared to Value and Network which are more explicative on the weekly basis.
2. PCA and Factors selection
The PCA on daily returns on our universe gives the following variance decomposition:
With 66% of the variance explained by the first five factors, daily returns display more noise than weekly ones, where equivalent PCA explained 85% of the variance. Nonetheless, the PCA highlights the preponderance of Reversal factor on the second component, Flow factor is the most correlated to the third component. On their side, Momentum and Sentiment styles seem involved in the fourth component.
We also perform the same features selection method to isolate most contributive factors. The scikit-learn mutual_info_regression to estimate entropy from k-nearest neighbors’ distances gives the following selection (refer to Part1 for the code):
bench_mktCap FLOW_style POS_style MOM_style
0 -0.001331 0.0 -0.008964 0.015483
1 0.068994 0.0 0.024559 -0.004473
2 0.023944 0.0 0.042313 0.009388
3 -0.004287 0.0 0.003335 0.003645
4 0.013639 0.0 -0.026038 0.004395
... ... ... ... ... ...
Beyond market, we find again Flow, Positioning and Momentum. We then conduct a Lasso penalized regression embedded selection:
Total factors: 11
Selected factors: 4
factors with coefficients shrank to zero: 7
# Retrieving selected Factors
list(selected_feat)
['bench_mktCap', 'FLOW_style', 'POS_style', 'MOM_style']This method finally retains the same factors in its selection.
In summary, our analysis reveals that while the Market factor remains the predominant driver of our coin portfolio, the Momentum, Reversal, Positioning and Flow styles are also important factors that we have to include in our feature selection. These factors are also clearly identified in the Principal Component Analysis (PCA) that we conducted on the entire portfolio. Additionally, the Sentiment factor emerges in the PCA and exhibits relatively good performance in section 1, and should be considered to a lesser extent. Note that Value and Liquidity, which were the most relevant for the weekly study, do not appear anywhere for the daily approach.
3. Multi-Factors Model
Having identified the significant factors, we shall now explore regression models to elucidate the cross-coins daily returns of our portfolio.
Over a period of more than two years, our model achieved an R-squared value of 0.423, with a significant Beta Market of 1.14 and a t-stat of 192. Other most important betas are Reversal and Momentum factors with t-stats of respectively 7.1 and 4.4. We also note that Sentiment, Value and Positioning bring a significative contribution to the model. We then test all possible models by combining different numbers of factors and ranked them according to their adjusted R-squared values. The best model is a 4-factors model that used the Reversal, Momentum and Positioning factors:
We note a great improvment in R-squared from 0.423 to 0.591 in this 4-factors model. With the market, we incorporate in this model the 3 factors clearly identified in the features selection method at section 2 (Note though that Flow factor is not represented in the best model). And it affirms the dominance of price related factors for daily returns. If we now apply the best model for weekly returns using Value and Liquidity factors, we come up with:
This model is clearly not as descriptive as our 4-factor models for this daily portfolio with a R-squared of 0.422 versus 0.591. Those factors are thus probably too slow to be good candidates to explain daily forward return and should be kept for longer term decomposition. But consistently with what we observed on weekly returns models, results show variability among coins. Sentiment is again prevalent for coins like DOGE and Gaming/Metaverse related tokens. DeFi representatives are more impacted with Network and Flow factors compared to the full sample even if all of them remain more driven by technical factors.
Final Thoughts
In this study, we have explored factors decomposition of our daily portfolio and compared it to our previous weekly study. We found that a 4-factor model driven by price-related factors are dominant in driving short-term returns. Reversal is particularly interesting on this horizon and form with Momentum and Positioning the best model to explain daily return, in sharp constrast with the best one for weekly returns dominated by Value and Liquidity. Some token specificities are though noticeable both in the weekly and daily analysis like the importance of Sentiment factor for subsets of coins. It invites particularly to a deeper Sentiment factor analysis for further exploration.
As usual, we would be happy to hear your thoughts.
Questions and comments can be addressed to: contact@alphabetim.io
*62 coins: BTC, ETH, BNB, XRP, SOL, ADA, DOT, DOGE, AVAX, TRX, MATIC, LTC, NEAR, UNI, BCH, XMR, LINK, XLM, ATOM, ETC, ALGO, HBAR, VET, EGLD, MANA, XTZ, FIL, SAND, ZEC, AAVE, EOS, MKR, THETA, GRT, RUNE, NEO, CHZ, KSM, DASH, ENJ, WAVES, BAT, CRV, ONE, KAVA, COMP, XEM, QTUM, 1INCH, OMG, ZRX, YFI, SNX, ONT, STORJ, SUSHI, REN, DENT, BAKE, REEF, BAND, TOMO
References
[1] Cong & al. Value Premium, Network Adoption,and Factor Pricing of Crypto Assets (2022). European Financial Management Association (EFMA) conference.
[2] Liu, Y. and Tsyvinski, A. Risks and returns of cryptocurrency (2021). The Review of Financial Studies, 34(6), pp.2689–2727.
[3] Liu, Y., Tsyvinski, A. and Wu, X. Common risk factors in cryptocurrency (2022). The Journal of Finance, 77(2).
Disclaimer:
No Investment Advice
The contents of this document are for informational purposes only and do not constitute an offer or solicitation to invest in units of a fund. They do not constitute investment advice or a proposal for financial advisory services and are subject to correction and modification. They do not constitute trading advice or any advice about cryptocurrencies or digital assets. AlphabetIM does not recommend that any cryptocurrency should be bought, sold, or held by you. You are strongly advised to conduct due diligence and consult your financial advisor before making investment decisions.
Accuracy of Information
AlphabetIM will strive to ensure the accuracy of the information in this report, although it will not be held responsible for any missing or inaccurate information.
Non Endorsement
The appearance of third-party advertisements and hyperlinks in this report or on AlphabetIM’s website do not constitute an endorsement, guarantee, warranty, or recommendation by AlphabetIM. You are advised to conduct your due diligence before using any third-party services.
About AlphabetIM
AlphabetIM designs institutional-grade quantitative investment solutions for professional investors on digital assets.
