Topic: How do Crypto quantitative fund managers obtain Alpha?
host
Zheng Naiqian @ZnQ_626
- LUCIDA Founder
- 2019 Bgain Digital Asset Trading League Season 1 Mixed Strategy Group Champion;
- 2020 TokenInsight Global Asset Quantitative Competition, Compound Strategy Group: April runner-up, May champion, and season third place;
- 2021 TokenInsight x KuCoin Global Asset Quantitative Competition, third place in the compound strategy group season;
Guests
Ruiqi @ShadowLabsorg
- Founder of ShadowLabs & Investment Director of DC Capital
- Quantitative product management scale exceeds US$300 million
- Market making consultant for many exchanges and well-known projects
Wizwu @wuxiaodong10
- RIVENDELL CAPITAL Multi-factor & subjective strategy fund manager
- Computer + Finance background
- 20M non-traditional crypto strategies
- Focus on on-chain and off-chain data mining and neutral multi-factor strategies
What does the framework of a fund manager’s Alpha strategy look like?
Naiqian Zheng@LUCIDA:
LUCIDA is a multi-strategy hedge fund. We develop various low-correlation diversified strategies to ensure that our performance can survive both bull and bear markets.
Let me take proprietary funds as an example. Our profit goal is to outperform the spot price increase of Bitcoin in a bull market, so we will first do a macro market timing, that is, determine whether the market is at the bottom of a bear market and the top of a bull market. This judgment is very low frequency, probably in years.
If we believe that the current market is at the bottom of a bear market, we will convert all our funds into full positions in Bitcoin and hold them throughout the bull market. On this basis, we will use quantitative strategies such as CTA, multi-factor strategies, and statistical arbitrage strategies to enhance returns, which are also the core sources of Alpha in a bull market. At the same time, we will dynamically adjust the capital allocation between these strategies according to the current market environment to ensure the utilization rate of funds.
If we think the market has reached the top of the bull market, we will sell all our Bitcoins and convert them into US dollars to survive the bear market. We will also use strategies such as CTA and option volatility arbitrage in the bear market to increase the amount of US dollars until the next market cycle.
Therefore, all Alpha contributions include two categories: 1. Macro-timing judgment of bull and bear markets, which is also one of our core competitiveness. 2. Enhanced returns from quantitative strategies. For example, if Bitcoin rises from 10,000 to 50,000, it is unrealistic to accurately buy the bottom at 10,000 and sell at 50,000. Then we will use quantitative strategies to enhance returns and ensure that we can outperform the rise of Bitcoin.
Wizwu:
Speaking of Alpha strategy, it is related to the capital attributes of our fund. We have received a lot of native funds from the cryptocurrency circle, all of which are cryptocurrencies, so we have to passively earn Alpha, which is essentially an index-added strategy. In this index-added strategy, we have multi-factor strategies and some subjective strategies.
As an institution, we need to consider many things when doing subjective trading, including holding period, liquidity of small coins, etc. These factors mean that we have fewer targets to choose from. If we hold too many positions, it will be easy to be dispersed and we will not be able to outperform the market; if we hold too few positions, we will have to compete with the project investors, so our framework is to do everything.
For example, if we find a factor, different people have different ways of dealing with it, some are neutral, some are subjective, and some are quantitative, which represent different trading ideas, so I count subjective and multi-factor together. Because there is no such precedent in the Crypto market, for us, there are both factor strategies in the stock market, which are data-driven; there are also value-based ones, but we have not found them; there are also some futures, especially the analysis ideas of inventory and supply and demand in futures. So all of this depends on our ability to understand data and trading clues.
But we don’t have the investment research department of the first-tier funds native to the cryptocurrency circle, because we don’t have as many resources and such a broad vision as they do. We focus on being flexible and data-driven. So different people in the market make different money, which is a bit like the futures market. The industry makes industry money, the quantitative makes quantitative money, and the subjective makes subjective money. The methodologies are different, and the money earned in the end is also different.
Generally speaking, we mainly use the currency standard. If we use the currency standard, we hope that our strategy can reach a Sharpe level of 3-4, and the annualized return should reach more than 10%. Macro timing will be done less or very low frequency. Based on this, we derive factors through some insights into the market. These factors can be applied to various strategies, including subjective, multi-factor, etc.
In the process of digging out factors, we like to move some factors from the futures or stock markets for testing, and we also have our own trading experience.
Ruiqi:
We are a purely quantitative and automated team, so when we first designed the Alpha structure framework, we followed the principles of high engineering and high automation, so we rely heavily on data-driven and execution. We internally divide our Alpha framework into execution Alpha and prediction Alpha.
Crypto exchanges are very scattered, and there are many investment tools. For example, if I want to obtain the risk exposure of a transaction, I can choose to trade futures or spot, or choose to trade on different exchanges. So we will compare the funding costs of different markets at the execution level, such as the price of futures and spot, basis, handling fees, transaction slippage, borrowing costs, etc. After comprehensively comparing different costs, we will try to choose the lowest cost tool. In this part, we can achieve an annualized level of about 5% to 20% through comparison and selection, and we regard this category as the executed Alpha.
The second part is the predicted Alpha, which mainly refers to the predictions at different levels, different cycles and different targets, including time series and cross-sectional ones. We will adjust our risk exposure on different titles based on the predictions.
However, there is a special situation. The predicted Alpha will be somewhat coupled with the executed Alpha in processing. For example, I predict a direction now, but the prediction I made may only solve 20% of the problem. The remaining 80% comes from whether I can actually execute it. This part includes order placement techniques, transaction probability analysis, conditional probability of capital cost, etc. These all have certain execution factors and certain prediction factors. Generally speaking, we achieve our Alpha breakthrough under such a system.
When we attribute performance, the proportions contributed by these two parts of Alpha are also different.
For example, the execution alpha mentioned just now, our goal is to outperform the benchmark by 5% to 20%, so this part is relatively certain, but the upper limit of the return is relatively limited. The prediction alpha is different. For example, some of our high-frequency predictions have very small profits per transaction, and they are mixed with a lot of execution alpha. However, for some medium and low-frequency predictions, their proportion in prediction alpha may be relatively high.
What is your view on the Crypto market? What kind of market do you think Crypto is?
WizWu：
As mentioned earlier, we should make different money in different markets. We make money from logical analysis in futures, and the same is true for Crypto. The characteristic of the Crypto market itself is high volatility. For example, the return of the U-based fund rate is at least 20% annualized in a bull market. So if we want to make money, we have to think about how we make money based on these characteristics. If we come in with U, we may use it for arbitrage first. This is the risk-free return of arbitrage.
Crypto is in a bull market now, and the risk-free rate of return on Pendle is 30% to 40%. Assuming we calculate the most accurate Sortino ratio, the final minus is the expected minimum return. After the minus, as a risk strategy, the remaining return is actually not much, so this is one of the reasons why we do coin-based Alpha.
My view of the market is hot money. I will make money wherever there is money to be made and wherever the logic is clear.
The market rotation rhythm of the Crypto market this year is very similar to that of the A-shares. In the past five or six years, the A-shares have had a main line every year. For example, it was carbon neutrality at the beginning, and this year it is AI. But in history, among the bull and bear markets in the cryptocurrency market that I have experienced and reviewed, there is only this year that has such a main line. This year, there is one AI and one Meme. Before this, there was no main line in the cryptocurrency market. It was really a very boring market. This is also the difference between this year and the previous ones. So this year, if you can catch AI and Meme in the cryptocurrency market, you can make a lot of money.
When it comes to capturing the hot spots and rotation patterns of the Crypto industry, momentum itself is the most important part. In addition to data, we also pay attention to public opinion on Twitter. However, if there are few targets, the data we can focus on is still the value of the target itself.
We have a tool internally that is somewhat similar to Wind. We have been working on factors for almost two years, and we store the market and Twitter sentiment. However, we do not pay much attention to sectors, because we do not catch sector rotation in this way. Our factors will select the coins with better elasticity in the sector and buy these assets.
Ruiqi:
We believe that Crypto is a highly speculative market, mainly composed of a large number of continuous transactions and occasional event transactions. This is also the reason why we continue to participate in the market.
Compared with other financial targets or markets, it has more emotional trading and event trading, which is more suitable for quantitative capture, so it is also consistent with our trading advantages.
Today, market competition has intensified. Whether in transaction execution or forecasting, there are now a hundred flowers blooming and a hundred boats vying for the stream, but there are still some highly structured opportunities. The sources of these structural trading opportunities are still full of emotions and events.
The market has begun to undergo structural differentiation. First, in terms of market predictability, the effectiveness of pricing on old assets has been further improved.
Specifically, we can see that in the past, if a trend wanted to ferment, it might take several hours or even a day or two, but now the trend may end in 10 minutes. The huge errors caused by different factors will be corrected quickly.
However, we found that there is still a good alpha in new assets. If we also participate in some Altcoins, we will find that there are some new assets in everyone's narrative, whether it is competition, entrepreneurship, or new trends, you will find that the factors used before are still effective on these assets.
However, it is difficult to acquire new assets. For example, your technical implementation, data access, and the stability of the trading model are somewhat lacking.
What is the contribution of different factors in the Crypto market? What are the underlying sources of returns for these factors?
Wizwu:
The characteristic of the Crypto market is that the funding rate is high, that is, the basis is large. For futures, the basis can be understood as the monthly spread. Assuming that they are understood as the same thing, the volatility of the monthly spread in the Crypto market is very large. Features such as arbitrage are built around this kind of thing, and alternative factors may also be built around this logic.
In addition, due to the large market volatility, some of the coins in the differentiation are very elastic, so the real money you get still depends on timing. So we tried this momentum and found that the neutral momentum profit is the same as the Bitcoin level in the bull market. If you don't choose the timing, it is difficult to see a good excess return. This is also related to the trading mechanism of Crypto.
In addition, the data and some over-the-counter data that our exchange can provide are also different from those in the traditional market. Therefore, many of our excess returns are derived from these unique features and strategies that have been played out in the traditional market.
Ruiqi:
One of the representatives of sentiment factors is the momentum factor, which is essentially about chasing highs and selling lows.
The profits of this factor mainly come from the overreaction of the market. For example, when retail investors see a certain currency rising, they usually think that this upward trend will continue, so they follow up and buy. At this time, we can add fuel to the fire and profit from it.
In addition, you can also conduct momentum reversal trading, based on the judgment of market overreaction, ambush and reverse operation in advance. The core of these transactions is to take advantage of the market's overreaction to gain profits.
The profits of event factors mainly come from the repricing of assets, which requires a certain reaction time. For example, by monitoring data on Twitter or potential data of large market trends, you can react quickly after an event occurs. For example, when CPI data is released, the price of Bitcoin may fluctuate violently. In this case, reacting quickly and trading can make a profit.
From the perspective of high-frequency trading, many traders are insensitive to transaction costs, which leads to them often conducting all transactions on a single market when conducting large transactions. This behavior will have a greater impact on the market, thus bringing arbitrage opportunities.
The liquidity factor is effective in the high-frequency market in the long term and is one of the important tools for fund managers to obtain Alpha.
What do you think is the difference in methodology if we want to get some Alpha in the Crypto market compared to the traditional financial market? How can we get more Alpha in Crypto?
Naiqian Zheng@LUCIDA:
In recent years, I have clearly felt that people may be the most core element of Alpha.
Although the Crypto industry has developed a lot, compared with the A-share market, there is a clear generation gap in the average level of Crypto industry practitioners, especially secondary market participants.
The second point is that the data and infrastructure of this market are really poor. There is almost no complete data supplier like Wind and Bloomberg in the A-share market.
The data quality is poor and highly fragmented. Getting data is a headache for many teams, but how can you model without data?
I think if an institution has obvious advantages over its peers in terms of talent and data, it will be a very stable source of excess returns.
Wizwu:
Compared with traditional financial markets, the Crypto market has several distinctive features: high volatility, high elasticity of small currencies, and strong speculation. To gain Alpha in the Crypto market, you must develop strategies around these characteristics.
A core problem is that the risk-free arbitrage returns in the Crypto market are too high. This is devastating to the value factor of the Crypto market, because there are very few projects that can bring stable USDT dividends. So when we want to calculate the value, PE, and price-earnings ratio, we will find that no matter how we calculate it, it is far less than the arbitrage returns on a U-based basis. Therefore, it is not feasible to use the value factors in the traditional financial market to measure the Alpha of the Crypto market.
In the Crypto market, the core values we need to focus on are different from those in the traditional market. In the traditional stock market, factors such as value and price-to-earnings ratio are the core, while in the Crypto market, we may pay more attention to the price-to-dream ratio, that is, the optimistic estimate of future expectations and everything derived from achieving these expectations.
A specific example of a factor is a value factor. For example, in the Layer 2 (L2) solution, MATIC, the change in the number of native token addresses holding 10 to 100 U (USDT) can often indicate some market trends. When a public chain is about to usher in a hit application or large-scale adoption, the increase in these small holders is usually a positive signal. It is often more resonant with market sentiment and prices, and it is also relatively early. An address like this essentially represents one person, which is a question of how many people there are. From the perspective of this factor characterization, you think that an address with a balance of 10 to 100 US dollars is more like a real user.
Ruiqi:
I have summarized several differences:
Information asymmetry caused by market fragmentation
The decentralized nature of the Crypto market leads to information asymmetry. It is difficult for non-professional investors to understand the market situation, so arbitrage opportunities are particularly obvious.
Buying high and selling low and market volatility
Unlike traditional financial markets, assets in the Crypto market are usually traded in multiple regional markets. Therefore, this dispersion makes the phenomenon of chasing ups and downs and running around more common. Investors' frequent switching of attention and irrational transactions are more common in the Crypto market.
Market manipulation
In the Crypto market, market manipulation is more common than in traditional markets. For most ordinary investors, it is difficult to use this phenomenon to trade or design trading strategies. However, for some high-frequency trading companies, they can manipulate the market on a larger scale than in traditional markets to obtain Alpha. This behavior is illegal in traditional markets and will lead to jail time.
The difference between the asset management product structure of the Crypto market and the traditional financial market
Naiqian Zheng@LUCIDA:
I found that more than 80% of secondary teams use very neutral arbitrage strategies, so the homogeneity between strategies is very serious.
From the perspective of investment, the principle of the strategy itself is not complicated, and if you are doing it at a lower frequency, you don’t need to put too much energy into transaction execution. This leads to more than 80% of the products being in the arbitrage track. Then, when you go to do some CTA or options, or multi-factor strategies, the input-output ratio is particularly inappropriate compared to this kind of statistical arbitrage.
This also includes high frequency trading. Then you switch equipment and optimize all your transaction details, but in the end, your management scale is still significantly different from that of arbitrage. So do you think that arbitrage products will become the mainstream of the entire market in the future?
Wizwu:
Bond trading is also a big part of the traditional financial market, not just the Crypto market. The trading volume of bonds of different levels is not low, so arbitrage trading will always exist. As long as it can be operated under some semi-compliant premise, the arbitrage income of the Crypto market can be at least two to six times that of the traditional market, which provides a very high capacity and profit space for arbitrage trading, so this situation will continue to exist.
As for other strategies, such as CTA strategies, they are also a large-capacity option. Such strategies may not be truly recognized by the market until the arbitrage returns drop. At that time, the Sharpe ratio of our strategy will become very good.
Now the arbitrage income is calculated based on U. Thanks to the unified account of the current exchange, we can also run similar strategies through the currency standard. So our current direction is to use U to run arbitrage and use currency to run risks. This is the best allocation method.
Ruiqi:
I basically agree with Wiz. First of all, the market is highly fragmented, and there are so-called barriers to entry. These problems may be difficult to solve in the next two to three years. Therefore, in the foreseeable two to three years, arbitrage space will continue to exist. Even if the arbitrage space is reduced, the trading volume and capital capacity of arbitrage will still account for the majority of the market.
But by then, arbitrage may not exist in the form of asset management products. It will be more self-operated by high-frequency quantitative teams, mainly because the high-frequency teams will directly eat up the profits themselves, and there will be no additional profits to distribute to the market. This is probably the case.
For some asset management projects, they will settle for the second best and provide an adjusted risk-return ratio with a decent price-performance ratio, such as statistical arbitrage and CTA strategies. In the next two to three years, such soil may begin to emerge.
Naiqian Zheng@LUCIDA:
The structure of Crypto asset management products is also very different from that of A-shares, because I have observed that the most mainstream product of A-shares now is index increase. Whether it is benchmarked against broad-based indexes such as 300, 500 or 1000, products based on index increase should be the best-selling ones.
Most of the underlying index increases are implemented by multi-factor models. But I found that there are almost no such products in the Crypto market. I know that there are probably less than 10% of teams developing multi-factor strategies. Why are there so few teams developing multi-factor strategies?
Wizwu:
The reason is that the returns of USDT in the market are too high.
For example, I buy almost all of the USDT on PENDLE. In this case, I will not choose my own strategy. Because when my strategy deducts 30% of the risk and divides it by volatility, its performance is not even as good as the Sharpe ratio and other indicators of the traditional futures market. Therefore, I think that when the market risk-free return is so high, everyone will naturally choose the risk-free return. According to this calculation method, the proportion of the strategy standard should be subtracted from the risk-free return. When we use the real risk-free return of this market (annualized 30%) to calculate, everything becomes futile, and it is meaningless no matter how we calculate.
Our multi-factor strategy has become more diversified. When we first designed it, we did design it according to the neutral multi-factor strategy of A-shares or traditional futures. But later it gradually became more diversified and added more subjective factors.
I think the core reason is that the retracement cycle of this market is very short and the changes are very fast. In this case, there are some framework problems in implementing multi-factor strategies. We cannot only look at the market trends in the past two years to prove that a factor is effective in the long run. In the traditional market, we may dig out a factor and test it not only in A-shares but also in US stocks. If it is effective for 20 years in US stocks and 5 years in A-shares, we can say that this is an effective factor and can be used for large-scale fund operations.
However, in the Crypto market, it is difficult to have such a verification opportunity to use this factor to make a neutral strategy. Maybe it can only be tested in a one or two-year backtest cycle, which is not reasonable in terms of framework.
Ruiqi:
My feelings may be different, which also depends on our understanding of this framework. What I have observed is that there are more people doing time-series trading on mainstream coins, such as trend trading on Bitcoin and Ethereum. But if we say that there are trend trading on 100 tokens, there are very few such teams. There are many people doing time-series trading, but few doing cross-sectional trading. This is the phenomenon I have observed.
If I have to attribute it, I think there are mainly the following reasons:
First, there is the issue of data length. Most assets may have only experienced one cycle, and there is no longer data to verify and backtest.
Secondly, even assets that have gone through multiple cycles, such as EOS, became inactive after 2017 and 2018 and are difficult to be selected into the target pool. There are many similar targets in the Crypto market, but there are few assets that can go through several circles and maintain activity and liquidity, basically only Bitcoin and Ethereum. Others, such as Solana, have also been silent for a long time and have only recently become active.
Third, relatively speaking, the effectiveness of time series factors may be more significant than the effectiveness of cross-sectional factors in practice. The underlying logic is that reactions to emotional momentum are long-lived and can be planned very well using a traditional trend trading framework. The relative strength factor of the cross-section is unstable because many targets themselves are unstable. Unlike traditional commodities or stocks, which have experienced multiple bull and bear market cycles, their relative strengths and weaknesses are relatively stable. In the Crypto market, the target of this wave may disappear in the next wave, and it is impossible to verify whether its relative strength exists.
What do you think is the standard for measuring the value of Crypto assets? What is the value of Crypto assets?
Ruiqi:
From the current situation, the value of the Crypto market is equivalent to attention. In other words, it is now an attention-driven market. No matter what the underlying logic of the project is, as long as it can gain attention, it can gain value.
This may have some similarities to the market momentum mentioned by Wiz, but I don’t think it is exactly the same. In short, this is more like a product of the attention economy.
In the long run, we expect and many practitioners and VCs are also working hard to promote a direction in which future value is reflected in practical applications and the competitiveness of the ecosystem as much as possible. But at least at present, the state of the market is not entirely like this.
Easter egg: What do you think of the market now? How do you think Bitcoin will develop in the future? (Subjective and irresponsible)
Wiz:
If I were to guess, it would not have much room to go up if it continues to fluctuate at this position. Even if it breaks through a new high, the increase would probably be only about 30%, and then it might have to pull back. At the current level, I think the world's major risk assets may not have much room to go up.
This is really a slap in the face, it's very revealing.
Ruiqi:
I am more optimistic because I think interest rates have not started to fall yet. Although I did not believe in Bitcoin before, I am basically a half-believer in Bitcoin now. Therefore, I think it is still possible to reach 150,000 within two years in this bull market cycle.
