There are about 800 crypto funds today, making it a daunting task for investors to decide how to invest.
With about 800 crypto funds relying on a new asset class, which has its own properties, it is essential to assess them through an appropriate framework. We provide a basic framework of useful metrics to assess the true risk of a crypto fund as a quantitative screening tool. Shortlisted funds can then be assessed in more detail through a classic due diligence process.
Assessing the return/risk profile of a directional trading crypto fund
Assessing the expected return of a directional fund
Investors in a directional fund should first have a clear understanding of the dynamic of the fund’s overall strategy in order to realize where the performance will come from and over what period before assessing whether the risk taken to achieve such results is worth it. This is achieved through discussions with the fund manager.
Warning: If a fund manager refuses to explain any of the fund’s strategies, beware!
When asking about a fund’s strategies, a truthful and experienced manager should be able to explain it in plain English. If a fund manager doesn’t want to disclose anything stating that it’s a trade secret, you could still try to understand what the fund tries to achieve by analyzing its past track record. However, in such a case, it’s unlikely that the manager will provide daily returns of the strategy for a more granular analysis, which may thus be worthless.
A transparent fund manager inspires trust, a secretive one inspires defiance, but even if a manager is transparent about strategy, investors should verify that these pitches from fund managers are credible and not take their word for granted. The Bernie Madoff Ponzi scheme was just that. Madoff explained that he was trading S&P 100 options as the basis of his strategy. Why not? But given the size of this specific market (~$100 million daily on average), there was no way he could have been trading the size of his fund ($6 billion), but he still lured many naïve investors.
Understanding the fundamentals of the strategy
Directional funds try to achieve their goals in different ways, and investors have to understand in which market environments they are going to perform well or not; some funds may perform very well during smooth trending markets but can be crushed during times of high volatility, whereas funds performing well during hectic markets can dramatically underperform in strong trending markets.
No single strategy can perform well in every market environment, as each strategy is designed to only fully capture specific moves and avoid being crushed otherwise. Directional funds tend to embed different strategies, each designed to capture specific market moves; but since these strategies are usually blended together, the resulting blend should perform well during most market environments, but will always underperform the best single strategy in a given market environment.
Understanding the strategy timeframe
Understanding the timeframe through which a fund strategy works — i.e., intraday and/or on a severalday basis — and the broad expectations of the strategy in terms of capturing market movements — e.g., captures 80% of an upward move, 30% of a downward move on average — are necessary to make a meaningful comparison against a potential benchmark.
In the example just quoted, such a fund would underperform a passive index representative of the traded underlying asset during strong upward movements but should prove its value when the passive index reverses course by limiting the losses, leading to a better performance against the passive index but over a full up/down market cycle.
Assessing the risk profile of a directional fund
In order to assess the risk profile of a directional fund, an advanced — i.e., nonlinear — hedge fund analysis framework is useful, but metrics of a crypto fund cannot be compared with the metrics of a traditional hedge fund — e.g., volatility, Sharpe ratio, etc.
We will assume that the past behavior of a fund is expected to continue more or less in the near future if the manager’s strategy is robust and well designed.
A nonlinear analysis framework
If an instrument behaves the same during different market conditions, it is said to have a linear behavior, but if it behaves differently during different market conditions, it is said to have a nonlinear behavior.
For example, when a fund gains 1% every time the broad market gains 1% and loses 1% every time the broad market losses 1%, it is linear; but when a fund gains 1% every time the broad market gains 1% and loses 2% every time the broad market losses 1%, it is nonlinear, as its behavior during negative markets doesn’t have the same amplitude as during positive markets.
Assessing the nonlinearity of a fund
The question is: “Is a given fund linear or nonlinear?” The quick answer is that most active funds will be nonlinear, but there’s a statistical test to answer the question more precisely, the Jarque–Bera test for normality.
However, metrics from a nonlinear framework can also be used to assess linear instruments, but not the other way around.
Nonlinear risk metrics
The four main metrics of a linear framework adapted to assess nonlinear asset behaviors are volatility, correlation, beta and value at risk.
Simple time series are used in the section below to illustrate the purpose.

Volatility
Volatility measures the degree of dispersion of returns around their mean. The higher the volatility, the higher the dispersion of the returns. If an asset has a linear behavior, a high dispersion of returns around their mean indicates that returns can be far above but also far below their mean, and this is generally considered as an easily understandable measure of risk. However, if the asset has a nonlinear behavior, overall volatility can be highly misleading, either over or underestimating the risk of loss.
In order to assess the behavior of a nonlinear asset from a volatility point of view, we will split the metric into two submetrics: positive volatility and negative volatility. Positive volatility is a classic volatility measure but is only applied to the positive returns of the asset. Likewise, negative volatility is a classic volatility measure but is only applied to the negative returns of the asset. Thus, we assess the dispersion of the returns on the positive side and on the negative side. If the asset is linear, these two metrics are close to each other.
Example: Let’s consider three funds, A, B and C as having had the following returns over the same period:
Fund A: { 3%; 8%; 5%; 58%; 1%; 2; 48%; 2%; 1%; 38% }
Fund B: { 3%; 8%; 5%; 12%; 1%; 2; 6%; 2%; 1%; 4% }
Fund C: { 45%; 8%; 5%; 12%; 1%; 2; 6%; 2%; 1%; 4% }
The volatility of Fund B is 5.3%, whereas the volatility of Fund A is 23.1%. Thus, if considering the overall volatility as a risk measure, then Fund B is much less risky than Fund A, whereas Fund C lies between.
When assessing the positive and negative volatility of funds A, B and C, we have:
Looking at the positive and negative volatility of each fund leads to a very different conclusion from just looking at their overall volatility: Fund C having the highest negative volatility and the lowest positive volatility is actually the riskiest of the three funds, whereas fund A having the highest positive volatility and the lowest negative volatility is the least risky, and fund B lies in between.
In fact, by taking a closer look at the returns of the three funds, Fund A contained its losses as much as Fund B but was able to capitalize on three strong returns that Fund B couldn’t capture. On the other hand, Fund C is similar to Fund B but has only been heavily hit once, whereas Fund B hasn’t.
Therefore, would one rather invest in a fund that delivers good returns, controlling the downside, but without any upswing either (Fund B), or invest in a fund that controls the downside as well, but which can deliver a winning lottery ticket from time to time (Fund A)?
Assessing the volatility of a crypto fund with a nonlinear framework is the only way to assess its true risk from a volatility point of view — i.e., understanding what contributes to high volatility.
Debunked myth #1: A crypto fund with overall high volatility doesn’t necessarily equate a highly risky one.

Correlation
Correlation measures how an asset is moving in relation to another one. The closer an asset is to 1, the more the assets will move in sync; the closer an asset is to 1, the more the assets will move in the opposite direction one from each other.
Again, measuring the overall correlation of a nonlinear asset can lead to misleading conclusions about how one asset moves in comparison with another.
Example:
Fund A: { 9%; 13%; 1%; 15%; 9%; 1; 28%; 6%; 2%; 0% }
Fund B: { 5%; 13%; 1%; 28%; 6%; 1; 25%; 5%; 2%; 1% }
Benchmark: { 28%; 2%; 33%; 34%; 19%; 15; 21%; 10%; 6%; 5% }
The correlation of Fund A to the benchmark is 0.81, which is similar to the correlation of Fund B to the benchmark. By looking at how these two funds correlate with their common benchmark, they are identical when assessing their overall correlation.
Now assessing the positive and negative correlations of Funds A and B with their benchmark, we have: a more subtle manner to assess the correlation of a fund with a benchmark. It consists of breaking the global correlation measure described above into two subcorrelation analyses: The positive correlation is the measured correlation of the fund with a benchmark only during positive returns of the benchmark, whereas the negative correlation is the measured correlation of the fund with a benchmark only during negative returns of the benchmark. The positive and negative correlation measures range like the standard correlation measure between 1 and +1 with the same meaning.
Therefore, an investor should look for a fund that has a high positive (i.e., the closest to +1) positivecorrelation, meaning the fund moves up when the benchmark moves up, and a low negative (i.e., the closest to 1) negativecorrelation, meaning that the fund moves up when the benchmark moves down.
Fund A exhibits a moderate positive positivecorrelation with its benchmark (0.23) and a moderate positive negativecorrelation with its benchmark (0.30), whereas Fund B shows a very high positive positivecorrelation with the benchmark (0.97) and a medium negative negativecorrelation with its benchmark (0.45).
This means that Fund A moved more or less in sync with its benchmark either on the upside or the downside, whereas Fund B moved upward when the benchmark was up most of the time but moved also upward from time to time when the benchmark was moving down. This is exactly the characteristic of a fund investors should look for, but this is only visible in a nonlinear framework.
Debunked myth #2: A high global correlation of a crypto fund to a benchmark doesn’t necessarily mean that the fund will move in sync with the benchmark most of the time.

Beta
The beta measures the amplitude of how an asset is moving compared to another. Its value is a rough estimate of how much an asset will move vs. another one considered. A value above 1 means that an asset moves more than 1x than another one in the same direction; a value between 0 and 1 means that an asset moves less than 1x than another one in the same direction. Negative values can be interpreted as positive values in terms of multiplying effect, but with moves on the opposite directions.
Note: The beta of an asset vs. another should only be calculated if there’s a statistically significant correlation between the two assets.
Example: Let’s consider the two funds used previously with the correlation analysis, which were both highly correlated with the benchmark (0.81).
Fund A: {9%; 13%; 1%; 15%; 9%; 1; 28%; 6%; 2%; 0%}
Fund B: {5%; 13%; 1%; 28%; 6%; 1; 25%; 5%; 2%; 1%}
Benchmark: {28%; 2%; 33%; 34%; 19%; 15; 21%; 10%; 6%; 5%}
The beta of Fund A to the benchmark is 0.46, and the beta of Fund B to the benchmark 0.43 — i.e., both funds have a similar beta to their benchmark. But are they really equal?
Assessing the positive and negative beta of Funds A and B with their benchmark, we have:
Unsurprisingly, when looking at the beta of these two funds through a nonlinear prism, we have a different story. Fund A tends to capture on average about 11% of an up or down move of its benchmark, whereas Fund B tends to capture on average 48% of an up move of its benchmark while capturing 15% of a negative move of its benchmark — i.e., capturing 15% of the amplitude of the down move of its benchmark, but delivering it in positive terms instead.
Just like with the correlation, investors should seek to invest with funds showing an ashighaspossible positive positivebeta and an ashighaspossible negative negativebeta vs. the funds’ benchmarks.
Debunked myth #3: The overall beta of a crypto fund has no value unless it is assessed in a nonlinear manner.

Value at Risk
The value at risk, or VaR, is an estimate of how much an investment might lose, with a given probability, given normal market conditions, and in a set time period.
Example: VaR (Fund, 95%) = 7.5% means that over the considered period, the fund can lose more than 7.5% with 5% (= 100%–95%) probability. In other words, there’s a 95% chance that the fund will lose less than 7.5% over the considered period.
There are many ways to compute the VaR of an asset that go beyond the scope of this paper, but again, if the nonlinear behavior of the asset is not taken into account in estimating the VaR, the results lead to false conclusions.
However, given the oftenhectic behavior of digital assets, it is difficult to assess their VaR, no matter the model used, and the obtained results may not be of great help to calibrate risk. This is why VaR is not really used to assess crypto funds.
Comparing the risk metrics of traditional hedge funds and crypto funds
Now that the main diehard myths about fund metric analysis have been debunked, another misleading analysis aspect of crypto funds is to compare the metrics side by side with the wellknown metrics of traditional assets.
Essentially, digital assets are way more volatile than their traditional cousins, and some of their metrics can be of several orders of magnitude different: from annualized return and volatility to the Sharpe and Sortino ratios.
Sharpe ratio
For example, a Sharpe ratio above 1 is more of an exception rather than the norm for funds dealing with traditional assets, as their annualized return is usually in the 5%–15% range and an annualized volatility of 10%–15% that doesn’t imply insignificant returns from their means.
However, with Bitcoin (BTC), for example, its annualized return from 2016 to date has been slightly above 100%, while its annualized volatility is close to 85%, leading to a ratio above 1 despite its frequent booms and busts.
Thus, the Sharpe ratio of a good crypto fund — one that is able to provide to capture most of the upside of its underlying asset while protecting on the downside — can be in a high single to a low doubledigit range, which can appear highly suspicious if compared to the Sharpe ratio of a typical hedge fund.
Sortino ratio
The same is even more true for the Sortino ratio. For example, Bitcoin has a 30% annualized downside volatility, which is roughly three times that of the S&P 500, meaning negative returns reaching three times further than the ones of the S&P 500, which leads to a three times lower value of the denominator of the Sortino ratio of Bitcoin. However, if Bitcoin has an annualized return 10 times bigger than that of the S&P 500, the numerator of the Sortino ratio of Bitcoin will be 10 times higher than the numerator of the Sortino ratio of the S&P 500. Thus, when calculating the Sortino ratio of Bitcoin, dividing a numerator that is 10 times bigger (than the one of the S&P 500) by a denominator that is 3 times bigger (than the one of the S&P 500), we obtain roughly a ratio for Bitcoin that is about 3.3 (=10/3) times higher than that of the S&P 500. More precisely, the Sortino ratio of Bitcoin is above three, whereas the Sortino ratio of the S&P 500 is about 0.8.
Therefore, for a good crypto fund, posting a high annualized return over limited downside volatility can easily lead to a high doubledigit Sortino ratio.
Drawdowns
Drawdowns are bounded metrics between 0% and 100%, contrary to the unbounded metrics that are the Sharpe and Sortino ratios described above. Thus, an investor can compare side by side the drawdowns of a crypto fund to the ones of a traditional fund without having to take into account the scaling of the metrics.
However, investors have to understand that the magnitude of drawdowns of crypto funds can be more substantial than the ones of a fund trading only traditional assets, as the digital assets can swing more wildly. For example, a 40% drawdown for a crypto fund can be “equivalent” to a 15% drawdown for a traditional fund, but the crypto fund lost is nevertheless more than the traditional fund. The idea is just to put things into perspective here.
A loss due to a drawdown is never pleasant to experience, especially when it is a big loss; therefore, investors have to pay more attention to the shapes of the fund drawdowns. The shape of a drawdown refers to the shape described by the drawdown curve of a fund. These shapes are triangles more or less tilted, which tell how the fund manager dealt with losses and are highly instructive, as we will detail below.
Let’s consider these three funds:
Fund A: { 1%; 3%; 1%; 5%; 2%; 23.5; 2%; 6%; 2%; 3%; 1%; 5%; 2%; 3%; 6%; 3% }
Fund B: { 1%; 2%; 1%; 0.5%; 2%; 1.5%; 2%; 0.5%; 2%; 3%; 1%; 2%; 1%; 23%; 1%; 2% }
Fund C: { 2%; 1%; 3%; 1%; 0.5%; 1%; 0.5%; 19%; 21%; 3%; 2%; 1%; 0.5%; 2%; 0%; 1% }
They all have the same performance (around +5%) and maximum drawdown (around 20%) over the same period, but the shapes of their drawdowns depict a very different story for each fund.
Generally, there are three cases:

A sudden loss followed by a steady recovery over several weeks. This is the shape of the drawdowns one could expect. At some point, the fund manager’s strategy is caught wrongfooted and a sudden, steep loss occurs. As discussed earlier, as the old Wall Street adage says “markets take the elevator down, but the stairs up” — i.e., a sudden panic move downward happens quickly, but it takes time for the markets to calm down and realize that what caused the panic move in the first place is over, which explains the slow recovery. These drawdowns are normal and inherent to the strategy. Investors have to simply make sure that all of the past major drawdowns were about the same magnitude, showing the robustness of the underlying strategy; bad trades occur, but they are always controlled and will eventually recover.

Continuous and increasing losses over several months recovered in just a few weeks. Such drawdowns are more problematic, as they may show that the manager’s strategy hasn’t worked for a long time, but facing investors’ redemptions, the fund manager went “all in” in order to stop the bleeding: It’s make or break. However, such drawdown shapes can sometimes also be explained by the way the strategy works and may not be a sign of a gambling fund manager. This is why it is always important to understand what the fund strategy tends to capture in order to assess its behavior.

A sudden loss, followed by a quick recovery. These drawdowns can take place from time to time and are usually linked to a market dislocation, leading to a fast and deep loss followed by an equally strong recovery.
Finally, when looking at fund drawdowns, having datasampling as precise as possible is key: Looking at drawdowns on a daily basis or on a monthly basis can lead to very different conclusions.
If managers just report their performance on a monthly basis, as is generally the case, only the change of the fund’s net asset value, or NAV, between the last day of the current month and the last day of the previous month are disclosed. There’s no information about what occurred during the month. For performancereporting purposes, that’s fine, but for risk assessment, this can be highly misleading.
Indeed, if the fund witnessed a 30% drawdown during the month that fully recovered by the end of the month, then looking only at monthly NAVs won’t show it, and investors will have a false sense of confidence by assuming that the fund never had any 30% drawdown in this example. Reporting performance on a daily basis shows what happened from day to day, which is far more informative than just from month to month.
For passive index, drawdowns measured on a daily or monthly basis are very close because there’s no active management involved. However, with actively traded strategies, short but steep drawdowns can occur from time to time, and if investors are not aware of that possibility, they may be in for a rude awakening, possibly panicking and selling their holdings.
Conclusion
Crypto funds come in different shapes and sizes, as we have briefly described in this article.
No matter their nature, since they are all dealing with highly volatile underlying assets, they tend to exhibit nonlinear behavior, which requires a proper framework to analyze them. Through a nonlinear analysis of such funds, we have highlighted that:
 A crypto fund with overall high volatility doesn’t necessarily equate to a highly risky one.
 A high global correlation of a crypto fund to a benchmark doesn’t necessarily mean that the fund will move in sync with the benchmark most of the time.
 The global beta of a crypto fund has no value unless it is assessed in a nonlinear manner.
Another point we touched upon is that comparing metrics of traditional funds vs. crypto funds is like comparing apples to oranges, given the very different nature of the underlying instruments traded.
We concluded on the drawdowns of crypto funds, which, to us, are a very powerful risk metric when properly analyzed. If an investor had to look at just one risk metric to assess the risk taken vs. the delivered performance, it would be the fund drawdowns, not just their depth, but also their shapes.
We gave some directions on which metrics to look at and analyze, but metrics without their context are meaningless. This is why such an analysis should always be conducted under the supervision of the professional fund manager’s explanations about his strategy.
This is part two of a twopart series on how to sort crypto funds — read part one with an overview of the main types of crypto funds here.
This article does not contain investment advice or recommendations. Every investment and trading move involves risk, you should conduct your own research when making a decision.
The views, thoughts and opinions expressed here are the author’s alone and do not necessarily reflect or represent the views and opinions of Cointelegraph.
David Lifchitz is the chief investment officer and managing partner at ExoAlpha — an expert in quantitative trading, portfolio construction and risk management. With over 20 years of experience in these fields and 8+ years in information technology with financial firms, he has notably been the former head of risk management at the U.S. subsidiary of Ashmore Group, which had $74 billion in assets under management in 2018. ExoAlpha has developed proprietary, institutionalgrade trading strategies and infrastructure to operate seamlessly in the digital asset markets applying strong risk management principles.
Source: cointelegraph