Economics Letters 228 (2023) 111118

Contents lists available at ScienceDirect

Economics Letters
journal homepage: [Link]/locate/ecolet

Performance measurement of crypto funds

∗
Niclas Dombrowski a , Wolfgang Drobetz a , Paul P. Momtaz b,c,d ,
a
Hamburg University, Germany
b
UCLA Anderson School of Management, United States
c
TUM School of Management, United States
d
UCL Computer Science Centre for Blockchain Technologies, United Kingdom

article info a b s t r a c t

Article history: Crypto funds (CFs) are a growing intermediary in cryptocurrency markets. We evaluate CF performance
Received 14 March 2023 using metrics based on alphas, value at risk, lower partial moments, and maximum drawdown. The
Received in revised form 29 March 2023 performance of actively managed CFs is heterogeneous: While the average fund in our sample does
Accepted 9 April 2023
not outperform the overall cryptocurrency market, there seem to be some few funds with superior
Available online 26 April 2023
skills. Given the non-normal nature of fund returns, the choice of the performance measure affects
JEL classification: the rank orders of funds. Compared to the Sharpe ratio, the most commonly applied metric in the
G11 asset management practice, performance measures based on alphas and maximum drawdown lead to
G12 diverging fund rankings. Depending on their ranking order of preferences, CF investors should consider
G23 a bundle of metrics for fund selection and performance measurement.
G24 © 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license
Keywords: ([Link]
Crypto funds
Active investment management
Performance measurement
Fund selection
Blockchain-based digital assets

1. Introduction following questions: Do CFs add value for investors? Does the
choice of performance measure matter for the evaluation of CFs?
Crypto funds (CFs) are an arising type of actively managed The contribution of our paper is twofold. First, we add to a
funds that invest in blockchain-based assets. Their performance long debate on the value of active asset management through
has become the focus of attention as the crypto market exhibits the lens of a new and growing asset class.1 To the best of our
extreme levels of volatility. According to CoinMarketCap, the knowledge, only two studies exist that address the performance
market valuation of all crypto assets reached its all-time high of of CFs (Cumming et al., 2022; Bianchi and Babiak, 2022).2 We
almost $3,000 billion in November 2021 only to drop to about extend their work by benchmarking CF returns against Liu et al.’s
$800 billion seven months later – a loss of over 70%. Neverthe- (2022) multi-factor crypto asset pricing model, considering a
less, market growth since its early stages and the multi-faceted larger set of funds, and incorporating the crypto market turmoil
promises of blockchain technology have attracted many investors, during the 2021–22 period.
including CFs. Fig. 1 illustrates the number of fund foundings Second, we evaluate CF performance along a broad set of
over time and also reveals the impact of market swings: Growth performance measures. The question which measure to use is
slowed after the first peak in 2018, but CFs continue to gain in crucial from a CF investor’s perspective given the divergent find-
relevance. As of June 2022, more than 800 CFs existed with assets ings in the investment fund literature whether or not the choice
under management (AuM) close to $58 billion (Crypto Fund Re- of the performance measure influences the evaluation of funds.3
search, 2022). Most important, CF returns are highly non-normal, and a frequent
Despite these strong market dynamics and the increasing im-
portance of institutional investors in crypto markets, we still
1 See Agarwal et al. (2015) and Cremers et al. (2019) for literature reviews
know very little about CFs’ value contribution and performance
on hedge fund and mutual fund performance, respectively.
measurement. Our paper aims to fill this gap by addressing the 2 A related literature examines institutional investors, adding cryptocurren-
cies to their portfolios (Huang et al., 2022).
∗ Correspondence to: 110 Westwood Plaza, Los Angeles, CA 90095, United 3 In a more general framework, Platanakis and Urquhart (2019) show that the
States. method for risk estimation matters to assess the performance of cryptocurrency
E-mail address: momtaz@[Link] (P.P. Momtaz). portfolios.

[Link]
0165-1765/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license ([Link]
N. Dombrowski, W. Drobetz and P.P. Momtaz Economics Letters 228 (2023) 111118

Fig. 1. Evolution of crypto funds.

Note: This figure displays the cumulative number of crypto fund foundings from 2009 until September 2022. In total, our sample includes about 800 CFs with a
reported founding year.

concern is that funds with non-normal return distributions can- the U.S., and the same applies for similar efforts at national and
not be adequately evaluated by using the classic Sharpe ratio. supranational levels around the globe.
While Eling and Schuhmacher (2007) and Eling (2008) document To estimate CF alphas, we run market model regressions us-
that various alternative performance measures result in identical ing a value-weighted crypto market index and the U.S. equity
rank orderings for hedge funds and mutual funds, other stud- market index. In addition, we estimate the three-factor alpha
ies (Ornelas et al., 2012; Zakamouline, 2010) find that the choice based on Liu et al.’s (2022) crypto market model. They document
of the performance measure does influence the ranking across that three factors, market, size, and momentum, are sufficient to
funds. explain the cross-section of returns of a variety of crypto trading
Our results confirm that CF returns are heavily right-skewed strategies.5 Following Ferson and Schadt (1996) and Ferson and
and fat-tailed. Compared to a value-weighted crypto market in- Qian (2004), we assess the distribution of the estimaed alphas via
dex, the average CF outperforms by as much as 1.9% per month. their t-statistics.
This outperformance vanishes when applying Liu et al.’s (2022) A growing literature shows that the use of different perfor-
multi-factor crypto asset pricing model based on market, size, and mance measures is important because higher moments of the
momentum. While CFs outperform the crypto market, they are return distribution play a significant role in performance evalua-
unable to add value relative to dynamic factors known from the tion, and their effect depends on the choice of the performance
asset pricing literature. Performance measures based on alphas measure (Ornelas et al., 2012; Zakamouline, 2010). Since CF re-
and maximum drawdown lead to diverging rankings compared to turns strongly deviate from the normality assumption, this new
the classic Sharpe ratio. Tests of the significance of rank correla- asset class is an ideal example, where correlations between the
tions reveal that, in particular, maximum drawdown performance Sharpe ratio (the most commonly used benchmark in the asset
measures are independent from other metrics. We conclude that management industry) and alternative performance measures are
CF investors, depending on their ranking order of preferences, expected to be low.
should rely on a bundle of metrics for performance measurement. To determine whether the choice of the measure matters in
The remainder of our paper is structured as follows. Section 2 the CF industry, we compare 14 different performance measures.
reviews existing literature on CFs and describes our fund perfor- In particular, these encompass classic approaches based on the
mance measures. Section 3 presents the data sample. We discuss normality assumption together with a variety of alternative met-
our empirical results in Section 4 and conclude in Section 5. rics: Sharpe ratio (Sharpe, 1966), single- and multi-factor alphas,
information (appraisal) ratio (Treynor and Black, 1973), excess
2. Crypto funds and performance measurement return on value at risk (VaR) (Dowd, 2000), conditional Sharpe
ratio (Agarwal and Naik, 2004), modified Sharpe ratio (Gregoriou
While attention on CFs has strongly increased from investors, and Gueyie, 2003), Omega ratio (Keating and Shadwick, 2002),
researchers, and regulators in recent years, the literature on this Sortino ratio (Sortino and Van Der Meer, 1991), Kappa 3 ra-
new asset class is still scarce. Most related to our study, Bianchi tio (Kaplan and Knowles, 2004), upside potential ratio (Sortino
and Babiak (2022) document that CFs achieve outperformance et al., 1999), Calmar ratio (Young, 1991), Sterling ratio (Kestner,
against the crypto market. Although this outperformance reduces 1996), and Burke ratio (Burke, 1994). Table 1 provides detailed
when applying a multi-factor pricing model, they conclude that definitions of these performance measures. To evaluate whether
CF performance is driven by superior skills of active fund man- the choice of a particular measure is critical for CFs’ perfor-
agers. Cumming et al. (2022) document similar outperformance mance evaluation, we compute all measures per CF in our sample,
relative to the market. They also show that CFs positively in- rank funds, and calculate Spearman rank correlations coefficients
fluence the financial outcome of token-backed ventures.4 From between the funds’ performance measures.
a regulatory perspective, Mokhtarian and Lindgren (2018) docu-
ment that CF regulation still remains at a very early stage even in
5 For the sake of brevity, we refer to Liu et al. (2022) for a detailed description
of factors and strategies. We note that their factor data are only available until
4 The positive impact of institutional investors on the financial success of December 2021, and thus our three-factor model results are based on a shorter
blockchain-based ventures has also been shown by Fisch and Momtaz (2020). sample period.

2
N. Dombrowski, W. Drobetz and P.P. Momtaz Economics Letters 228 (2023) 111118

Table 1
Definitions of performance measures.
Performance measure Formula Variable definition
Panel A : Traditional performance measures
Sharpe Ratio (ri − rf )/σi ri is the average monthly return of CF i, rf the monthly
risk-free rate, and σi the standard deviation of monthly excess
returns
αi,cmkt , αi,emkt ri − (rf + βi,m × (rm − βi,m is the sensitivity of CF i’s returns to the returns of the
rf )), m = index m and rm the monthly index return; betai,m can be
{cmkt , emkt } Cov (r ,r )
expressed as Var(ri m) ; the index m can be either the
m
value-weighted crypto market index cmkt or the U.S. equity
market index emkt
Information Ratio αi,cmkt /σ (ϵi,cmkt ) αi,cmkt is CFi ’s monthly crypto market alpha and σ (ϵi,cmkt ) the
tracking error, which is the standard deviation of the
difference between the returns of CFi and the crypto market
index
αi,cmft ri − (rf + βi,cmkt × rcmkt is the monthly crypto market return; SMB is the size
(rcmkt − rf ) + βi,SMB × factor of crypto returns; MOM is the momentum factor of
SMB + βi,MOM × MOM) crypto returns; the three factors follow the work of Liu et al.
(2022)
Panel B : Performance measures based on v alue at risk
Excess Return on VaR (ri − rf )/VaRi VaRi represents CFi′ s value at risk computed by
VaRi = −(ri − zα × σi ), where zα is the standard normal
distribution quantile at the significance level α
Conditional Sharpe (ri − rf )/CVaRi CVaRi is the conditional value at risk, defined as
Ratio CVaRi = E [−ri,t |ri,t ≤ −VaRi ]
Modified Sharpe Ratio (ri − rf )/MVaRi MVaRi is the modified value at risk based on the
Cornish–Fisher expansion (Favre and Galeano, 2002), computed
as MVaRi = −(ri + σi × (zα + 1/6 × (zα2 − 1) × Si + 1/24 × (zα3 −
3 × zα ) × Ei − 1/36 × (2 × zα3 − 5 × zα ) × Si2 )), where Si and Ei
denote the skewness and excess kurtosis of CFi , respectively
Panel C : Performance measures based on low er partial moments
Omega Ratio (ri − τ )/LPM1,i (τ ) + 1 Lower partial moments (LPMs) measure negative return
deviations relative to a minimal acceptable return τ as
∑T
LPMn,i (τ ) = T1 t =1 max[τ − ri,t , 0] ; LPMs are of order n = 1
n

for the Omega Ratio

(ri − τ )/ 2 LPM2,i (τ )
√
Sortino Ratio LPMs are of order n = 2
(ri − τ )/ 3 LPM3,i (τ )
√
Kappa 3 Ratio LPMs are of order n = 3
HPM1,i (τ )/ 2 LPM2,i (τ )
√
Upside Potential Ratio Higher partial moments (HPMs) measure positive return
deviations relative to a minimal acceptable return; HPMs are
of order n = 1, LPMs of order n = 2
Panel D : Performance measures based on maximum draw dow ns
Calmar Ratio (ri − rf )/(−MDDi,l ) Max. drawdowns (MDDs) measure risk as the largest return
losses of CFi during the sample period; MDDi,l is CFi ’s largest
decline
Sterling Ratio (ri − Evaluates risk as the average of the N largest drawdowns
∑N
rf )/( N1 j=1 −MDDi,j )
√∑
N
Burke Ratio (ri − rf )/ 2
j=1 MDD2i,j Measures risk as the square root of the sum of the N largest
squared drawdowns

Note: This table defines the measures used to evaluate the performance of crypto funds. The measures are grouped into different categories based on their approach
to assess a fund’s risk-return profile. Variables are defined at the first occurrence and suppressed thereafter for brevity. All measures, except αi,cmft , are based on
data for the sample period from January 2017 to June 2022. For αi,cmft , we rely on the crypto market factors as provided by Liu et al. (2022), which are available
only until December 2021. The size factor (SMB) is constructed using market capitalization, and the momentum factor (MOM) is based on past three-week return
windows. Liu et al. (2022) provide detailed descriptions. In addition, the following parameter values are chosen for measures in Panel B, C, and D: The VaR-based
ratios are computed at a significance level of α = 5%. For the measures based on lower and higher partial moments, we apply a minimal acceptable return of 0%.
For the Sterling and Burke ratios, we use the five largest drawdowns (N = 5).

3. Data monthly fund returns during our sample period, which strongly
deviates from normality.
Our CF sample comes from Crypto Fund Research (CFR), a
4. Empirical results
U.S.-based data aggregator that provides the most comprehen-
sive database for returns and characteristics of funds focused on Table 2 describes the performance characteristics of our sam-
blockchain investments. While CFR provides characteristics for ple of actively managed CFs. Summary statistics in Panel A show
over 800 CFs, monthly net-of-fees performance data are available that the average fund generates a monthly mean return of 7.8%,
for 352 CFs from January 2017 to June 2022. To avoid survivorship with a standard deviation of 5.9%. For the average fund, monthly
bias, our sample includes live as well as defunct funds. The differ- returns are highly volatile, right-skewed, and fat-tailed. The aver-
ence between the average monthly returns of surviving funds and age fund’s Sharpe ratio is 0.248. The alphas benchmarked against
all CFs in our sample is 0.20%. Fig. 2 reports the distribution of all the value-weighted crypto market index (αcmkt ) indicate that
3
N. Dombrowski, W. Drobetz and P.P. Momtaz Economics Letters 228 (2023) 111118

Fig. 2. Return distribution of crypto funds.

Note: This figure shows the distribution of monthly (discrete) crypto fund returns in % during our sample period from January 2017 to June 2022. The sample contains
returns for 352 crypto funds with a total of 16,808 fund-month observations.

Table 2
Performance characteristics of crypto funds.
Panel A: Summary statistics of fund returns
Crypto Fund Mean SD Q1 Median Q3
Return mean, in % 7.752*** 5.861 3.563 8.331*** 11.779
Return SD, in % 32.213 14.524 20.130 35.431 43.843
Return skewness 1.166 0.747 0.823 1.259 1.606
Return excess kurtosis 2.109 2.846 0.349 1.557 3.154
Sharpe Ratio 0.248 0.226 0.171 0.260 0.330
αcmkt , in % 1.868*** 3.726 −0.089 1.813*** 3.331
αemkt , in % 6.308*** 5.286 2.636 6.738*** 9.832
αcmft , in % −0.486** 4.576 −2.859 −0.215** 2.058
# of monthly returns 47.750 16.076 37.750 52.000 61.000

Panel B: Distribution of t-statistics of estimated fund alphas

Crypto market Equity market Crypto three-
Null model model factor model
(1) (2) (3) (4)
Minimum t-statistic −2.910 –2.551 −3.193
Bonferroni p-value (−) 1.000 1.000 1.000
t ≤ –2.326 1.760 4 3 7
–2.326 < t ≤ –1.960 7.040 6 1 5
–1.960 < t ≤ –1.645 8.800 6 2 11
–1.645 < t ≤ 0 158.400 80 30 159
0 < t ≤ 1.645 158.400 205 160 128
1.645 < t ≤ 1.960 8.800 17 50 14
1.960 < t ≤ 2.326 7.040 10 47 9
t > 2.326 1.760 24 59 18
Maximum t-statistic 6.655 7.106 5.662
Bonferroni p-value (+) 0.001 0.0003 0.003

Note: Panel A reports summary statistics for the 16,808 monthly returns of the 352 sample crypto funds. The return statistics are first calculated on the fund-level
and then aggregated across CFs. Alphas are based on CAPM regressions relative to the crypto market, the U.S. equity market, and the crypto three-factor model
based on Liu et al. (2022). Equity market data are retrieved from the website of Kenneth R. French. ∗ , ∗∗ , and ∗∗∗ indicate statistical difference from zero at the 0.10,
0.05, and 0.01 level, respectively, and are based on t-tests for means and Wilcoxon-tests for medians. Panel B shows the distribution of robust t-statistics for the
estimated alphas of all CFs. The rows contain the number of CFs for which the t-statistic falls within distinctive ranges of a standard normal distribution. As a point
of reference, column (1) shows the number of CFs per range for the hypothetical scenario that fund alphas followed a normal distribution. Columns (2)-(4) display
the actual distribution of alpha t-statistics for the crypto market model, the U.S. equity market model, and the crypto three-factor model. The Bonferroni p-value is
the one-tailed p-value of the minimum and maximum t-statistics multiplied by the number of CFs. It tests the null hypothesis that all alphas are jointly equal to
zero against the alternative that at least one fund alpha is negative (Bonferroni p-value (–)) or positive (Bonferroni p-value (+)). The sample period is from January
2017 to June 2022. Results for the crypto three-factor alphas are based on data only until December 2021.

CFs are able to outperform their passive benchmark and gen- dynamic trading strategies captured by the size and momentum
erate large and economically relevant excess returns. The mean factors (Liu et al., 2022), the outperformance of CFs becomes
(median) monthly αcmkt is 1.9% (1.8%), with p-values below 1%. substantially weaker and even seems to disappear.
Alphas measured against the U.S. equity market (αemkt ) are even Panel B illustrates the distribution of t-statistics for fund al-
higher. In contrast, the average fund’s three-factor model al- phas. Column (1) reports the number of CFs that fall within
pha (αcmft ) turns negative. The mean (median) αcmft is −0.49% critical ranges of a standard normal distribution if the estimated
(−0.22%), statistically significant at the 5% level. Correcting for alphas followed a normal distribution. The t-statistics of these
4
N. Dombrowski, W. Drobetz and P.P. Momtaz Economics Letters 228 (2023) 111118

Table 3
Rank correlations of performance measures.
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
Traditional performance measures
1. Sharpe Ratio
2. αcmkt 0.645
3. Information Ratio 0.826 0.863
4. αcmft 0.601 0.503 0.638
Performance measures based on v alue at risk
5. Excess Return on VaR 1.000 0.645 0.826 0.584
6. Conditional Sharpe Ratio 0.990 0.668 0.823 0.561 0.990
7. Modified Sharpe Ratio 0.995 0.615 0.810 0.596 0.995 0.978
Performance measures based on partial moments
8. Omega Ratio 0.982 0.668 0.832 0.594 0.982 0.976 0.965
9. Sortino Ratio 0.972 0.695 0.830 0.570 0.972 0.978 0.949 0.988
10. Kappa 3 Ratio 0.960 0.704 0.822 0.545 0.960 0.974 0.934 0.975 0.997
11. Upside Potential Ratio 0.948 0.702 0.814 0.539 0.947 0.964 0.919 0.961 0.991 0.997
Performance measures based on maximum draw dow n
12. Calmar Ratio 0.136 0.043 0.200 0.066 0.137 0.113 0.153 0.142 0.115 0.106 0.102
13. Sterling Ratio 0.155 0.067 0.224 0.110 0.156 0.129 0.174 0.158 0.131 0.120 0.116 0.996
14. Burke Ratio 0.151 0.062 0.218 0.101 0.152 0.126 0.170 0.155 0.128 0.117 0.113 0.997 1.000
Mean 0.720 0.529 0.671 0.462 0.719 0.713 0.712 0.721 0.717 0.709 0.701 0.254 0.272 0.268

Note: This table presents the Spearman rank correlations between all 14 performance measures used to compare the performance across crypto funds. The performance
measures are grouped into categories based on their approach to assess a crypto funds’ risk-return profile. The sample period is from January 2017 to June 2022.
Since the crypto market factors are only available until December 2021, the rank correlations involving αcmft are based on the period January 2017 to December
2021. The row ‘‘Mean’’ indicates the mean rank correlation of one performance measure to all other ratios.

alphas benchmarked against the aggregate crypto market (αcmkt ) it substantially affects a fund’s maximum drawdown. Given ex-
in column (2) describe a distribution that is centered to the right treme crypto market swings, these three measures lead to notably
of zero, right-skewed, and fat-tailed. Most important, there is different rankings than all other performance measures.
substantial heterogeneity in individual CF performance. While Following Eling and Schuhmacher (2007), we check the sta-
205 of 352 CFs show positive but statistically insignificant alphas, tistical significance of rank correlations using two different tests
51 funds exhibit positive alphas that are statistically significant (not reported). First, we apply a standardized version of the
at least at the 10%-level (with t-values larger than 1.645). This Hotelling–Pabst statistic, testing the null hypothesis that two
compares to only 16 funds in our sample that exhibit significantly rankings are independent, i.e., the corresponding rank correlation
negative alphas. is zero (Hotelling and Pabst, 1936). For rank correlations between
performance measures 1–11 (including traditional performance
A Bonferroni multiple comparison test rejects the null hypoth-
measures and those based on VaR and partial moments) in Ta-
esis that all estimated alphas are jointly equal to zero against the
ble 3, there is no case in which the hypothesis of independence
alternative that at least one fund alpha is positive. Column (3)
between two related rankings can be confirmed (based on the
reiterates these findings using the U.S. equity market as bench-
5% significance level). In contrast, when comparing the mea-
mark. As expected, in column (4), the multi-factor alpha (αcmft ) sures building on maximum drawdowns to all other performance
t-statistics are shifted and centered to the left of zero. The Bon- measures, the null hypothesis cannot be rejected. Therefore, the
ferroni test still rejects the null hypothesis that no fund alpha Calmar, Sterling, and Burke ratios generate fund rankings that are
is positive. In contrast, the minimum t-statistics are never sta- independent from other performance measures that ignore risk
tistically significant. Overall, while the mean (median) CF cannot related to a fund’s maximum drawdown.
beat the multi-factor benchmark, there is large heterogeneity in The second test is based on the Fisher z-transformation. In-
fund performance and at least some evidence of skill even when stead of testing the independence of fund rankings, we now check
applying this conservative benchmark (Bianchi and Babiak, 2022). the hypothesis that a rank correlation is smaller than a certain
Table 3 shows Spearman rank correlation coefficients between given rank correlation x. Assuming a 5% level of significance,
performance measures. The performance measures based on the and considering our performance measures 1–11 in Table 3, the
concept of VaR and partial moments display high rank correla- hypothesis that the rank correlation is smaller than x can only be
tions with the classic Sharpe ratio. For these groups, although rejected for all x values smaller than 0.435. In contrast, for the
the measures are built on downside rather than symmetric risk performance measures based on maximum drawdown, this test
concepts, the rank correlations with the Sharpe ratio do not fall confirms that the rank correlations with the other performance
below 0.948. Comparing the Sharpe ratio with alphas as well measures are zero (and thus independent). In conclusion, it does
as the information ratio, rank correlations drop notably to a matter which measure is used to evaluate CF performance. There
can be significant changes in the evaluation of CFs as compared
range between 0.601 to 0.826. The relatively low rank correlation
to that found using the Sharpe ratio, most importantly, when ap-
between αcmkt and αcmft of 0.503 indicates that rank orders of
plying performance measures that capture maximum drawdown
the two alphas are far from perfectly aligned and highlights
risk.
the importance to correct CF performance for tradeable factor
strategies. 5. Conclusion
Considering measures based on maximum drawdown, correla-
tions with other performance measures drop substantially to low This paper assesses the performance of actively managed CFs.
values in a range between 0.043 and 0.224. The Calmar, Sterling, We find that CFs outperform a value-weighted crypto market
and Burke ratios define risk based on the largest losses investors index, but outperformance disappears when considering a three-
could potentially suffer. While the chronological sequence of factor pricing model. However, fund performance is very hetero-
returns does not change the standard deviation of fund returns, geneous, and some few funds may still achieve superior returns,
5
N. Dombrowski, W. Drobetz and P.P. Momtaz Economics Letters 228 (2023) 111118

eventually improving the efficiency of the cryptocurrency mar- Eling, M., Schuhmacher, F., 2007. Does the choice of performance measure
ket (Urquhart, 2016). Our results further indicate that the choice influence the evaluation of hedge funds? J. Bank. Financ. 31 (9), 2632–2647.
of the performance measure does matter. Fund rankings diverge Favre, L., Galeano, J.-A., 2002. Mean-modified value-at-risk optimization with
hedge funds. J. Altern. Invest. 5 (2), 21–25.
greatly when using alphas and performance measures based on
Ferson, W.E., Qian, M., 2004. Conditional Performance Evaluation, Revisited.
maximum drawdown compared to the classic Sharpe ratio and Research Foundation of CFA Institute.
other performance metrics. Depending on their ranking order of Ferson, W.E., Schadt, R.W., 1996. Measuring fund strategy and performance in
preferences, CF investors should use a bundle of metrics for fund changing economic conditions. J. Finance 51 (2), 425–461.
performance measurement. Fisch, C., Momtaz, P.P., 2020. Institutional investors and post-ICO performance:
An empirical analysis of investor returns in initial coin offerings (ICOs). J.
Corp. Finance 64, 101679.
Data availability
Gregoriou, G.N., Gueyie, J.-P., 2003. Risk-adjusted performance of funds of hedge
funds using a modified sharpe ratio. J. Wealth Manag. 6 (3), 77–83.
Data will be made available on request. Hotelling, H., Pabst, M.R., 1936. Rank correlation and tests of significance
involving no assumption of normality. Ann. Math. Stat. 7 (1), 29–43.
Acknowledgment Huang, X., Lin, J., Wang, P., 2022. Are institutional investors marching into the
crypto market? Econom. Lett. 220, 110856.
We thank Yukun Liu and Aleh Tsyvinsk for sharing their crypto Kaplan, P.D., Knowles, J.A., 2004. Kappa: A generalized downside risk-adjusted
performance measure. J. Perform. Meas. 8, 42–54.
risk factor data.
Keating, C., Shadwick, W.F., 2002. An introduction to Omega. AIMA Newslett..
Kestner, L.N., 1996. Getting a handle on true performance. Futures 25 (1), 44–47.
References Liu, Y., Tsyvinski, A., Wu, X., 2022. Common risk factors in cryptocurrency. J.
Finance 77 (2), 1133–1177.
Agarwal, V., Mullally, K.A., Naik, N.Y., et al., 2015. The economics and finance of Mokhtarian, E., Lindgren, A., 2018. Rise of the crypto hedge fund: Operational
®
hedge funds: A review of the academic literature. Found. Trends Finance issues and best practices for an emergent investment industry. Stan. JL Bus.
10 (1), 1–111. Fin. 23, 112.
Agarwal, V., Naik, N.Y., 2004. Risks and portfolio decisions involving hedge funds. Ornelas, J.R.H., Silva Júnior, A.F., Fernandes, J.L.B., 2012. Yes, the choice of
Rev. Financ. Stud. 17 (1), 63–98. performance measure does matter for ranking of US mutual funds. Int. J.
Bianchi, D., Babiak, M., 2022. On the performance of cryptocurrency funds. J. Finance Econ. 17 (1), 61–72.
Bank. Financ. 138. Platanakis, E., Urquhart, A., 2019. Portfolio management with cryptocurrencies:
Burke, G., 1994. A sharper sharpe ratio. Futures 23 (3), 56.
The role of estimation risk. Econom. Lett. 177, 76–80.
Cremers, K.M., Fulkerson, J.A., Riley, T.B., 2019. Challenging the conventional
Sharpe, W.F., 1966. Mutual fund performance. J. Bus. 39 (1), 119–138.
wisdom on active management: A review of the past 20 years of academic
Sortino, F.A., Van Der Meer, R., 1991. Downside risk. J. Portf. Manag. 17 (4), 27.
literature on actively managed mutual funds. Financ. Anal. J. 75 (4), 8–35.
Sortino, F.A., Van Der Meer, R., Plantinga, A., 1999. The dutch triangle. J. Portf.
Crypto Fund Research, 2022. Crypto fund quarterly report – 2022 Q2.
Cumming, D.J., Dombrowski, N., Drobetz, W., Momtaz, P.P., 2022. Decentralized Manag. 26 (1), 50–57.
Finance, Crypto Funds, and Value Creation in Tokenized Firms, Vol. 4102295. Treynor, J.L., Black, F., 1973. How to use security analysis to improve portfolio
SSRN Working Paper. selection. J. Bus. 46 (1), 66–86.
Dowd, K., 2000. Estimating value at risk: A subjective approach. J. Risk Finance Urquhart, A., 2016. The inefficiency of bitcoin. Econom. Lett. 148, 80–82.
1 (4), 43–46. Young, T.W., 1991. Calmar ratio: A smoother tool. Futures 20 (1), 40.
Eling, M., 2008. Does the measure matter in the mutual fund industry? Financ. Zakamouline, V., 2010. The choice of performance measure does influence the
Anal. J. 64 (3), 54–66. evaluation of hedge funds, Vol. 1403246. SSRN Working Paper, Citeseer.