Digital Finance

https://doi.org/10.1007/s42521-019-00002-1

ORIGINAL ARTICLE

Advanced model calibration on bitcoin options

Dilip B. Madan1 · Sofie Reyners2   · Wim Schoutens2

Received: 21 September 2018 / Accepted: 13 March 2019

© Springer Nature Switzerland AG 2019

Abstract
In this paper, we investigate the dynamics of the bitcoin (BTC) price through the
vanilla options available on the market. We calibrate a series of Markov models on
the option surface. In particular, we consider the Black–Scholes model, Laplace
model, five variance gamma-related models and the Heston model. We examine
their pricing performance and the optimal risk-neutral model parameters over a
period of 2 months. We conclude with a study of the implied liquidity of BTC call
options, based on conic finance theory.

Keywords  Cryptocurrency · Modelling · Bitcoin · Calibration

JEL Classification  C52 · C60 · G10

1 Introduction

In 2008, an anonymous (group of) author(s) posted a white paper under the name
Nakamoto (2008). The paper suggests an electronic payment system, Bitcoin, that
does not depend on any central authority. In particular, the authority is replaced by a
distributed ledger, the so-called blockchain, that records all transactions. Only veri-
fied blocks are added to the blockchain. Verification is—roughly speaking—based
on solving a complicated puzzle and hence demands a lot of CPU power. People
who put their computer at work are, therefore, rewarded with bitcoins: they are min-
ing bitcoins. This is how new bitcoins are created. For a detailed overview of the key

* Sofie Reyners
[email protected]
Dilip B. Madan
[email protected]
Wim Schoutens
[email protected]
1
Robert H. Smith School of Business, University of Maryland, College Park, MD 20742, USA
2
Department of Mathematics, University of Leuven, Celestijnenlaan 200B, 3001 Leuven,
Belgium

13
Vol.:(0123456789)
 Digital Finance

$ 20 000

$ 15 000

$ 10 000

$ 5000

$0
3 7 8 8
01 01 01 01
r2 r2 e2 t2
tobe be J un ugus
Oc em 29 31 A
1 Dec
18

Fig. 1  Historical price chart of the BTC–USD rate

concepts of Bitcoin, we refer to Becker et al. (2013), Dwyer (2015) and Segendorf
(2014).
In October 2009, New Liberty Standard determined the first bitcoin exchange
rate. Its value was based on the price of the electricity needed for mining in the
United States. They concluded that one US Dollar was worth 1309.03 bitcoins, or
a BTC–USD rate of approximately 0.0008. Later on, multiple exchange platforms
emerged. The price of bitcoin gradually increased and reached parity with the US
Dollar in 2011. At the end of 2013, the barrier of 1000 USD was reached, but this
did not hold for long. The real breakthrough came in 2017, due to a combination of
events, among which the announcement of bitcoin derivatives.
On 24 July 2017, the US Commodity Futures Trading Commission (CFTC)
granted permission to LedgerX to clear and settle derivative contracts for digital cur-
rencies. Three months later, LedgerX became the first US exchange platform trading
bitcoin options. Bitcoin derivatives were born. In December 2017, the CFTC author-
ized the Chicago Mercantile Exchange (CME) and the Cboe Futures Exchange
(CFE) to start trading BTC futures. Some argue that the peak in BTC–USD rate
mid-December (Fig. 1) was caused by the launch of their futures. Before the exist-
ence of BTC futures, investors had indeed few possibilities to bet on a decreasing
bitcoin price. Futures made this easier, which could be a reason for the sharp decline
in price late December 2017.
A lot of research has been carried out to gain insight into the dynamics of the bit-
coin price. A substantial part of this research focuses on time series modelling, see for
instance (Kristoufek 2013; Garcia et  al. 2014; Kjærland et  al. 2018). These authors
explain the bitcoin price fluctuations using several technical and socio-economical fac-
tors, e.g. the cost of mining bitcoins. Baur et al. (2018) and Dwyer (2015) on the other
hand compare historical bitcoin returns with those of other assets. However, the market
of bitcoin derivatives is not taken into account yet. Literature on bitcoin derivatives
usually focuses on their regulation instead of their valuation. Bitcoin derivatives and
their prices are barely studied from a modelling point of view. Some articles regarding

13
Digital Finance

option pricing (models) are available, see for instance (Chen et al. 2018; Cretarola and
Figà-Talamanca 2017), but an extensive (empirical) analysis is missing.
In this paper, we aim to fill the gap between the bitcoin price modelling and bitcoin
derivative products. We use some traditional and more advanced asset pricing models
to get a better understanding of the bitcoin market. In particular, we investigate how
well a series of frequently used Markov–Martingale models match the bitcoin market.
Bitcoin vanilla options lie at the heart of this research. Given the option surface(s), we
fit the Black–Scholes model, Laplace model, variance gamma model, bilateral (double)
gamma model, VG Sato model, VG-CIR model and Heston model on the bitcoin mar-
ket. We start by calibrating the models on the option surface and elaborate on the differ-
ent fits. We discuss and compare the implied volatility smiles according to the Laplace
model and the Black–Scholes model. Finally, we investigate the Black–Scholes implied
liquidity, based on the theory of conic finance.
We believe this study is useful, because it gives us for the first time a glimpse of
the risk-neutral ℚ-measure. Time series modelling lacks this feature, and also bitcoin
futures do not contain this information. The calibrated models provide us moreover
with new insights in bitcoin’s volatility, for instance through the initial volatility and the
vol-of-vol parameter in the Heston model. Due to the turbulent behavior of the bitcoin
price, the traditional arguments against the Black–Scholes model in an equity setting
do also apply here. In fact, these effects are here even more pronounced, which encour-
ages the use of advanced jump processes and stochastic volatility to model the bitcoin
market even more.

2 The data

The study is performed on vanilla options with underlying BTC price index, con-
sisting of quotes according to six leading BTC–USD exchanges: Bitfinex, Bitstamp,
GDAX, Gemini, Itbit and Kraken. Options traded are European style and cash settled
in BTC, while USD acts as the numeraire. Option datasets used in this study were col-
lected manually from various unregulated exchanges. The data are extracted each Fri-
day approximately 2 h after expiration, to avoid extremely short-dated options. In this
way, we collect ten option surfaces, starting from Friday 29 June 2018 until Friday 31
August 2018. Each surface contains options with four or five different expiries, ranging
from 1 week to at most 8 months. Longer term options are not (yet) frequently traded
and are hence not included in the analysis.
Before we start modelling, some additional remarks are given about the option sur-
faces. First of all, the options are expensive, which is directly related to bitcoin’s high
volatility. More remarkable is the range of the strikes. For short-term options (up to
1 month), the strikes roughly range from 85 to 125% of the initial value of the underly-
ing. For higher maturities, however, it is not exceptional to encounter strike ranging
from 100 to 500%. Last, we checked whether the no-arbitrage assumption is roughly
satisfied. Consider the two-price put-call parity, i.e.
bidEC(K, T) − askEP(K, T) ≤ S0 − K ≤ askEC(K, T) − bidEP(K, T), (1)

13
 Digital Finance

where S0 is the initial bitcoin price and e.g. bidEC(K, T) denotes the bid price of a
European call option with strike price K and time to maturity T. The inequality is
respected with only a few exceptions. Out of the 491 strike–maturity combinations,
ten violate this relation. Note that we assumed here zero risk-free interest rates, both
for BTC and USD. The same assumption is implicitly made in the remainder of the
article.

3 Model calibration

We first examine how well the pricing models fit the atypical bitcoin option surface.
To this purpose, we calibrate the models on the entire surface of one particular day,
Friday 29 June 2018, for which the data are provided in “Appendix B”. Note that
the choice of this day is arbitrary. Second, the stability of the risk-neutral model
parameters over time is investigated based on a weekly time series of option surfaces
ranging from 29 June until 31 August. The length and frequency of this time series
are again arbitrarily chosen.

3.1 Calibration procedure

The option surface is cleaned by considering only out of the money options for
which both a bid and an ask price are available. Since out of the money options
have no intrinsic value, their prices only consist of the so-called time value. Consid-
ering only out of the money options, therefore, leads to a more robust calibration.
The optimization procedure is based on the mid-prices of the remaining options.
We determine optimal model parameters by minimizing the root mean squared error
(rmse) of the model prices with respect to the market prices. The resulting fit is eval-
uated in terms of three other measures: the average absolute error (aae), the average
pricing error (ape), i.e. the aae as a percentage of the average option price, and the
average relative pricing error (arpe).
√
√ n
√1 ∑
rmse = √ (marketpricei − modelpricei )2 , (2)
n i=1

n
1∑
aae = |marketpricei − modelpricei |, (3)
n i=1

∑n
i=1
�marketpricei − modelpricei �
ape = ∑n , (4)
i=1
marketpricei

n
1 ∑ |marketpricei − modelpricei |
arpe = . (5)
n i=1 marketpricei

13
Digital Finance

Table 1  Summary of the pricing rmse ape aae arpe

performance on 29 June
BS 27.8368 0.1142 22.0807 0.3379
Laplace 34.2618 0.1384 26.7563 0.4990
Heston 9.2144 0.0360 6.9505 0.1207
VG 20.4969 0.0858 16.5809 0.2410
BG 16.5971 0.0691 13.3652 0.2383
BDG 9.9357 0.0322 6.2255 0.1436
VG Sato 10.5010 0.0414 8.0007 0.1011
VG-CIR 9.2275 0.0350 6.7651 0.1028

When a model does not support a closed-form pricing formula for vanilla options,
model prices are calculated with the Fast Fourier-Transform (FFT) pricing method
developed by Carr and Madan (1999). The price of a European call option with
strike K and time to maturity T is given by
exp(−𝛼 log(K)) ∞
∫0 (6)
C(K, T) = exp(−iv log(K))𝜌(v)dv,
𝜋
where
𝜙T (v − (𝛼 + 1)i)
𝜌(v) = (7)
𝛼 2 + 𝛼 − v2 + i(2𝛼 + 1)v
and 𝛼 = 1.5 . The characteristic function 𝜙t of the log-price process at time t is speci-
fied below for each model.

3.2 The Black–Scholes model

The Black–Scholes model is a natural reference point to start the analysis. Let (St )t≥0
denote the bitcoin price process. In a zero interest rate Black–Scholes world, the
price process is given by
( 2 )
−𝜎 t
St = S0 exp + 𝜎Wt , (8)
2
where (Wt )t≥0 is a standard Brownian motion, S0 the initial value of the bitcoin price
and 𝜎 denotes the volatility. Calibrating the model on the option surface of 29 June
results in an overall best fitting volatility of 70.47%. In Fig. 2a, the Black–Scholes
prices are plotted against the observed market prices to visualize the fit. As expected,
the optimal volatility is relatively high in comparison with more common assets.
The evolution of the optimal Black–Scholes volatility in the next weeks of the data-
set is shown in Fig. 2b. More details on the goodness of fit are given in Table 1 and
Fig. 11.
Alternatively, we can compute the implied volatility for each option in the sur-
face. Implied volatility smiles based on the mid-prices of out of the money vanilla

13
 Digital Finance

Black-Scholes ( = 0.70, rmse = 27.84, ape = 0.11) Black-Scholes calibration

1200 1
Market price
Black-Scholes model price
0.95
1000

0.9

Optimal volatility
800
Price (USD)

0.85
600

0.8
400
0.75

200
0.7

18

18

18

18

8
18

8
01

01

01

01

01
20

20

20

20
20
0

t2

t2

t2

t2

t2
ly

ly

ly

ly
ne

us

us

us

us

us
0.5 1 1.5 2 2.5 3 3.5 4

Ju

Ju

Ju

Ju
Ju

g
Au

Au

Au

Au

Au
6

13

20

27
Strike (USD) 104

29

10

17

24

31
(a) Calibration on 29 June 2018. (b) Evolution over time.

Fig. 2  Black–Scholes calibration

Laplace ( = 0.67, rmse = 34.26, ape = 0.14) Laplace calibration

1200 0.9
Market price
Laplace model price
1000 0.85
Optimal volatility

800
0.8
Price (USD)

600
0.75

400
0.7

200
0.65
18

18

18

18

18
18

18

18

18

18
20

20

20

20

20
20

20

20

20

20
0
ly

ly

ly

ly

st
ne

st

st

st

st
0.5 1 1.5 2 2.5 3 3.5 4
Ju

Ju

Ju

Ju

gu

gu

gu

gu

gu
Ju

Au

Au

Au

Au

Au
6

13

20

27

Strike (USD) 104

29

10

17

24

31

(a) Calibration on 29 June 2018. (b) Evolution over time.

Fig. 3  Laplace calibration

options as published on Friday 29 June 2018 are displayed in Fig. 4 for each matu-
rity. The 1-week implied volatility smiles are the only graphs that actually resemble
the shape of a smile. However, recall that the strike range for the larger maturities is
highly asymmetric around the initial BTC–USD rate.

3.3 The Laplace model

Instead of modelling the log-returns with a normal distribution, the Laplace model
employs the heavier tailed Laplace distribution. The model has again one param-
eter, 𝜎 , corresponding to the volatility. A brief summary of the model specifica-
tions is given in “Appendix A”. Model calibration on the option surface of 29 June
leads to an optimal volatility 𝜎 of 67.35%. The optimal volatility is similar to its
Black–Scholes counterpart, but the overall price fits in Figs. 2a and 3a are not the
same. While the Black–Scholes model underestimates the deep out of the money

13
Digital Finance

T = 7 days, S = 5906 T = 28 days, S = 5872

1 1
Black-Scholes Black-Scholes
Laplace Laplace

0.9 0.9
Implied volatility

Implied volatility
0.8 0.8

0.7 0.7

0.6 0.6

0.5 0.5
5400 5600 5800 6000 6200 6400 6600 6800 5500 6000 6500 7000 7500 8000 8500 9000 9500 10000
Strike Strike

T = 91 days, S = 5867 T = 182 days, S = 5906

1 1
Black-Scholes Black-Scholes
Laplace Laplace

0.9 0.9
Implied volatility

Implied volatility
0.8 0.8

0.7 0.7

0.6 0.6

0.5 0.5
0.5 1 1.5 2 0.5 1 1.5 2 2.5 3 3.5 4
Strike 104 Strike 104

Fig. 4  Implied volatility smiles corresponding to out of the money options on 29 June 2018

call option prices in the right tail, the Laplace model overestimates them. However,
neither of these models captures the option surface accurately.
We can extend the comparative analysis by calculating the implied volatili-
ties according to the Laplace model. Across all maturities, the Laplacian  implied
volatility in Fig.  4 exceeds the Black–Scholes  implied volatility near the money.
For larger strikes, the Laplacian  implied volatility decreases and stabilizes, while
the Black–Scholes  implied volatility behaves adversely and increases. This can be
explained by the tail behavior of the log-returns in both models, being, respectively,
Laplacian and Gaussian distributed. The phenomenon of differing implied (Lévy)
volatilities was already discussed in Corcuera et al. (2009).

3.4 Heston’s stochastic volatility model

Besides the high volatility of the bitcoin price, the high variability of this volatility
should be analyzed. Incorporating a stochastic volatility process is hence a reason-
able next step. We consider Heston’s stochastic volatility model (Heston 1993) with
parameters 𝜅, 𝜌, 𝜃, 𝜂 and v0.1 Formally, we model the risk-neutral bitcoin price pro-
cess (St )t≥0 as follows:

1
  𝜅 = rate of mean reversion, 𝜌 = correlation stock—vol, 𝜃 = vol-of-vol, 𝜂 = long-run variance, v0 =
initial variance.

13
 Digital Finance

Heston ( = 2.08, = -0.09, = 2.36, = 0.93, v0 = 0.41,

rmse = 9.21, ape = 0.04) Heston calibration
1200 25
Market price
Heston model price
1000 20

v0

Optimal parameters
15
800
Price (USD)

10
600
5

400
0

200
-5

18

18

18

18

8
18

8
01

01

01

01

01
20

20

20

20
20

t2

t2

t2

t2

t2
ly

ly

ly

ly
ne
0

us

us

us

us

us
Ju

Ju

Ju

Ju
Ju

g
0.5 1 1.5 2 2.5 3 3.5 4

Au

Au

Au

Au

Au
6

13

20

27
29

10

17

24

31
Strike (USD) 104

(a) Calibration on 29 June 2018. (b) Evolution over time.

Fig. 5  Heston calibration based on the FFT method

dSt √
= vt dWt , (9)
St
where the squared volatility process vt is given by
√
̃t
dvt = 𝜅(𝜂 − vt )dt + 𝜃 vt dW (10)
with (Wt )t≥0 and (W
̃ t )t≥0 two correlated standard Brownian motions such that
( )
Cov dWt dW ̃ t = 𝜌dt. (11)
The characteristic function 𝜙t (u) of log(St ) is given by
𝜙t (u) = exp(iu log(S0 ))
( ( ( )))
× exp 𝜂𝜅𝜃 −2 (𝜅 − 𝜌𝜃ui − d)t − 2 log (1 − g exp(− dt))(1 − g)−1
( )
× exp v0 𝜃 −2 (𝜅 − 𝜌𝜃iu − d)(1 − exp (− dt))(1 − g exp(− dt))−1
(12)
with
d = ((𝜌𝜃ui − 𝜅)2 − 𝜃 2 (−iu − u2 ))1∕2 , (13)

g = (𝜅 − 𝜌𝜃ui − d)(𝜅 − 𝜌𝜃ui + d)−1 . (14)

Figure 5a shows a good fit. The optimal vol-of-vol parameter 𝜃 takes a high value
of 236%, indicating that the previous assumption of constant volatility was indeed
too strong. The vol-of-vol parameter gives, moreover, new information about bit-
coin’s volatility, which is often said to be high. This study indeed reflects a high
volatility—as confirmed by the optimal values of 𝜂 and v0—but it additionally
shows that bitcoin’s volatility itself is very volatile. Note that the optimal value of 𝜃
increases sharply over time, reaching a maximum of nearly 1300% at the end of July.

13
Digital Finance

3.5 Gamma models

The turbulent bitcoin price encourages us to include jumps in the model. The
presence of jumps in the bitcoin market is, moreover, extensively motivated in
Scaillet et al. (2018). In gamma-based models, the market is modelled by a sub-
traction of two independent gamma processes, being pure jump processes that
model, respectively, the upward and downward motions. We first fit the well-
known variance gamma (VG) model (Madan and Milne 1991; Madan and Sen-
eta 1990; Madan et  al. 1998) on the bitcoin option surface. Second, the bilat-
eral gamma (BG) model is considered. This model extends the variance gamma
model by modelling the speed of positive and negative jumps separately. More
generalizations and extensions consist in imposing the model parameters them-
selves to be gamma distributed, as explained in Madan et al. (2018). We consider
one such model, the bilateral double gamma (BDG) model. Two alternative gen-
eralized variance gamma models, the VG Sato and the VG-CIR models, conclude
our selection of gamma models.

3.5.1 The variance gamma model

We calibrate the variance gamma model in the original (𝜎, 𝜈, 𝜃)-parametrization.

The risk-neutral process of the bitcoin price is modelled by
exp(Xt )
S t = S0 , (15)
E[exp(Xt )]
where (Xt )t≥0 is a variance gamma process, i.e. it starts at zero, has √ independent and
stationary increments and the increment Xs+t − Xs follows a VG(𝜎 t, 𝜈∕t, 𝜃t) dis-
tribution over the time interval [s, s + t] . In this parametrization, 𝜃 controls the skew-
ness, 𝜈 the kurtosis and 𝜎 the volatility of the risk-neutral distribution of the log-
returns. Equivalently, the characteristic function of log(St ) is given by
( ( ))
𝜙t (u) = exp iu log(S0 ) − log(𝜙Xt (−i)) × 𝜙Xt (u) (16)
with 𝜙Xt the characteristic function of the VG process at time t,
( )−t∕𝜈
1
𝜙Xt (u) = 1 − iu𝜃𝜈 + 𝜎 2 𝜈u2 . (17)
2
Figure  6a summarizes the calibration results of 29 June. Figure  6b again reflects
the period of high volatility in the first 2 weeks of August. The value of 𝜃 becomes
strongly negative in this period, corresponding to more negative skewness in the
risk-neutral distribution of log-returns.

13
 Digital Finance

VG ( = 0.73, = 0.11, = -0.39, rmse = 20.50, ape = 0.09) Variance Gamma calibration
1200 1.5
Market price
VG model price
1000 1

Optimal parameters
800
0.5
Price (USD)

600
0

400
-0.5

200
-1

18

18

18

18

8
18

8
01

01

01

01

01
20

20

20

20
20
0

t2

t2

t2

t2

t2
ly

ly

ly

ly
ne

s
0.5 1 1.5 2 2.5 3 3.5 4

Ju

Ju

Ju

Ju

gu

gu

gu

gu

gu
Ju

Au

Au

Au

Au

Au
6

13

20

27
104

29
Strike (USD)

10

17

24

31
(a) Calibration on 29 June 2018. (b) Evolution over time.

Fig. 6  Variance gamma calibration based on the FFT method

BG (bp = 0.32, cp = 0.87, bn = 0.05, cn = 136.72, rmse = 16.60, ape = 0.07)

1200 Bilateral Gamma calibration
800
Market price cn
BG model price 600
1000
400
Optimal parameters

800 200
Price (USD)

600 4
bp
3 cp
400 bn
2

1
200
0
18

18

18

18

8
18

18

18

18

8
01

01
20

20

20

20
20

20

20

20
0
t2

t2
ly

ly

ly

ly
ne

st

st

st
s

s
Ju

0.5 1 1.5 2 2.5 3 3.5 4

Ju

Ju

Ju

gu

gu

gu

gu

gu
Ju

Au

Au

Au

Au

Au
6

13

20

27
29

Strike (USD) 104

3

10

17

24

31

(a) Calibration on 29 June 2018. (b) Evolution over time.

Fig. 7  Bilateral gamma calibration based on the FFT method

3.5.2 The bilateral gamma model

The VG process (Xt )t≥0 is in the CGM-parametrization2 defined by

(18)Xt = G(1) (2)
t − Gt ,
where (G(1) )
t t≥0 and (G (2)
)
t t≥0 are two independent gamma processes with, respec-
tively, parameters (C, M) and (C, G). Equivalently, the characteristic function of the
VG process at time t can be rewritten as
( )Ct ( )Ct ( )Ct
1 1 GM
𝜙Xt (u) = = . (19)
1 − iu 1 + iu GM + (M − G)iu + u2
M G

(√ )−1 (√ )−1
2
  C = 1∕𝜈 , G = 𝜃2 𝜈2
4
+ 𝜎2 𝜈
2
− 𝜃𝜈
2
and M = 𝜃2 𝜈2
4
+ 𝜎2 𝜈
2
+ 𝜃𝜈
2

13
Digital Finance

BDG (bp = 0.11, p = 1.33, p = 13.56, bn = 14.34, n = 54.21, n = 0.02

Bilateral Double Gamma model calibration
rmse = 9.94, ape = 0.03) 300
1200
p
Market price
BDG model price 200 n

1000

100

Optimal parameters
800
0
Price (USD)

15
600
bp

10 p

400 bn

n
5

200
0

18

18

18

18

8
18

18

18

18

18
01
20

20

20

20
20

20

20

20

20
2
0

ly

ly

ly

ly

st
ne

st

st

st

st
Ju

Ju

Ju

Ju

gu

gu

gu

gu

gu
Ju
0.5 1 1.5 2 2.5 3 3.5 4

Au

Au

Au

Au

Au
6

13

20

27
29

10

17

24

31
Strike (USD) 104

(a) Calibration on 29 June 2018. (b) Evolution over time.

Fig. 8  Bilateral double gamma calibration based on the FFT method

The bilateral gamma model distinguishes between the rate of the positive and nega-
tive motions by introducing the parameters cp and cn . When defining bp = 1∕M and
bn = 1∕G , i.e. the scale of respectively positive and negative jumps, the resulting
characteristic function is given by
( )cp t ( )cn t
1 1
𝜙(BG)
Xt
(u) = . (20)
1 − iubp 1 + iubn

Calibrating this model on the bitcoin option surface of 29 June leads to the fit dis-
played in Fig.  7a. The optimal BG model parameters in the weeks thereafter are
given in Fig. 7b. Note that the optimal values of cp and cn are overall significantly
different. The speed at which negative jumps occur is much higher than that of posi-
tive jumps, while the scale is larger for positive jumps.

3.5.3 The bilateral double gamma model

The bilateral double gamma model (Madan et  al. 2018) goes one step further in the
generalization by allowing the speed parameters cp and cn to vary randomly. They are
assumed to be gamma distributed, with characteristic functions:
( ) 𝜂p ( )𝜂n
1 1
𝜙cp (u) = and 𝜙cn (u) = . (21)
1 − iu𝛽p 1 − iu𝛽n

The resulting bilateral double gamma (BDG) process depends on six parameters:
bp , 𝛽p , 𝜂p , bn , 𝛽n and 𝜂n , and is defined by
( )𝜂p ( ) 𝜂n
1 1
𝜙(BDG)
X
(u) = . (22)
t 1 + 𝛽p t log(1 − iubp ) 1 + 𝛽n t log(1 + iubn )

13
 Digital Finance

VG Sato ( = 0.78, = 0.51, = -0.22, = 0.55, rmse = 10.50, ape = 0.04) VG Sato model calibration
1200 1.2
Market price
VG Sato model price
1000
0.8

Optimal parameters
800
Price (USD)

0.4
600

400 0

200
-0.4

18

18

18

18

8
18

8
01

01

01

01

01
20

20

20

20
20
0

t2

t2

t2

t2

t2
ly

ly

ly

ly
ne

s
0.5 1 1.5 2 2.5 3 3.5 4

Ju

Ju

Ju

Ju

gu

gu

gu

gu

gu
Ju

Au

Au

Au

Au

Au
6

13

20

27
104

29
Strike (USD)

10

17

24

31
(a) Calibration on 29 June 2018. (b) Evolution over time.

Fig. 9  VG Sato calibration based on the FFT method

The results of calibrating this model on the bitcoin option data are displayed in
Fig. 8a, b.

3.6 The VG Sato model

Alternatively, the variance gamma model can be extended to a Sato model (Sato 1999).
The VG Sato process (Yt )t≥0 is defined by scaling the VG process as follows:
Yt ∼ VG(𝜎t𝛾 , 𝜈, t𝛾 𝜃), (23)
or alternatively,
( )−1∕𝜈
1
𝜙Yt (u) = 1 − iut𝛾 𝜃𝜈 + 𝜎 2 𝜈u2 t2𝛾 . (24)
2
Figure 9a shows that the VG Sato model fits the observed bitcoin option prices sig-
nificantly better than the ordinary VG model. An overview of the evolution of the
risk-neutral parameters is given in Fig. 9b.

3.7 The VG‑CIR model

None of the previously considered gamma-based models took into account any notion
of stochastic volatility. The calibration results of the Heston model in Sect. 3.4, how-
ever, indicate that it may be interesting to also include a similar effect here. The VG-
CIR model (Carr et al. 2003) brings this in by making time stochastic. Assuming for
the rate of time change yt the classical Cox–Ingersoll–Ross (CIR) process,

(25)
1∕2
dyt = 𝜅(𝜂 − yt )dt + 𝜆yt dWt ,
where 𝜂 is the long-run rate of time change, 𝜅 the rate of mean-reversion and 𝜆 gov-
erns the volatility of the time change. The economic time elapsed in t units of calen-
dar time is then given by the integrated CIR process (Yt )t≥0 , where

13
Digital Finance

VG-CIR (y0 = 1.57, G = 26.43, M = 27.78, C = 95.01, = 0.78, = 5.87, = 3.55

rmse = 9.23, ape = 0.03) 104 VG-CIR model calibration
1200 3
Market price C
VG-CIR model price G
2 M
1000

Optimal parameters
800
0
Price (USD)

20
y0
600
15

10
400
5

200 0

18

18

18

18

8
18

18

8
01

01

01

01
20

20

20

20
20

20
t2

t2

t2

t2
ly

ly

ly

ly
ne

st
s

s
Ju

Ju

Ju

Ju

gu

gu

gu

gu

gu
Ju
0

Au

Au

Au

Au

Au
6

13

20

27
29
0.5 1 1.5 2 2.5 3 3.5 4

10

17

24

31
Strike (USD) 104

(a) Calibration on 29 June 2018. (b) Evolution over time.

Fig. 10  VG-CIR calibration based on the FFT method

RMSE APE
120 0.3
Black-Scholes Black-Scholes
Laplace Laplace
Heston Heston
100 0.25
Variance Gamma Variance Gamma
Bilateral Gamma Bilateral Gamma
Bilateral Double Gamma Bilateral Double Gamma
80 VG Sato
0.2
VG Sato
VG-CIR
rmse

ape

60 0.15

40 0.1

20 0.05

0 0
18

18
18

18

18

18

18

18

18

18
18

18
18

18

18

18

18

18

18

18
20

20
20

20

20

20

20

20

20

20
20

20
20

20

20

20

20

20

20

20
ly

ly
ly

ly

ly

st

ly

ly

ly

st
ne

ne
st

st

st

st

st

st

st

st
Ju

Ju
Ju

Ju

Ju

Ju

Ju

Ju
gu

gu
gu

gu

gu

gu

gu

gu

gu

gu
Ju

Ju
Au

Au
Au

Au

Au

Au

Au

Au

Au

Au
6

6
13

20

27

13

20

27
29

29
3

3
10

17

24

31

10

17

24

31

Fig. 11  Evolution of the pricing error over time

∫0
Yt = ys ds. (26)

The risk-neutral bitcoin price process is under the VG-CIR model hence given by

exp(XYt )
S t = S0 , (27)
E[exp(XYt ) | y0 ]

where (Xt )t≥0 is a variance gamma process. The corresponding characteristic func-
tion of log(St ) is, therefore, given by

( ( )) 𝜙Yt (−i𝜙Xt (u))

𝜙t (u) = exp iu log(S0 ) , (28)
𝜙Yt (−i𝜙Xt (−i))iu

where 𝜙Yt is the characteristic function of Yt . Figure  10a shows that the VG-CIR
model fits the bitcoin option surface fairly well. The optimal model parameters
according to the other weeks in the dataset are displayed in Fig. 10b.

13
 Digital Finance

3.8 Comparison of fits

We summarize the pricing performance for the previously calibrated models.

Table  1 displays the rmse, aae , ape and arpe for the calibration on 29 June 2018,
while Fig. 11 shows for each model the evolution of the rmse and ape over time.
It is clear that both one-parameter models perform poorly. The variance gamma
model already does a better job than the former models, but extending it to the bilat-
eral gamma model and the bilateral double gamma model results in slightly better
fits. The VG Sato model, which generalizes the VG model in an alternative way,
leads in general to lower error scores than the former models. However, the VG-CIR
and the Heston model, which both incorporate stochastically changing volatility,
have overall the best performance.

4 Conic finance: implied liquidity

The bitcoin derivative market is a relatively young, but growing market. In this
section, we investigate the liquidity of the available derivatives. Instead of using
the ordinary bid–ask spread as a measure for liquidity, we use the more advanced
implied liquidity. Implied liquidity, 𝜆 , is a unitless measure for liquidity that is intro-
duced in Corcuera et al. (2012), arising from the theory of conic finance by Madan
and Schoutens (2016).
Conic finance provides an alternative to the law of one price. Instead of focusing
on one risk-neutral price, it models both the bid and ask prices of an asset. In prac-
tice, a derivative’s bid price is given by the average of the discounted distorted pay-
off, while the ask price is modelled by the negative average of the distorted distribu-
tion of the negative payoff. The distortion function applied in the analysis below is
the Wang transform (Wang 2000), defined for 0 < u < 1 as
𝛹𝜆WANG (u) = N N −1 (u) + 𝜆 , 𝜆 ≥ 0,
( )
(29)
where N(⋅) denotes the cdf of a standard normal distribution. In the Black–Scholes
model, this distortion function leads to closed-form bid and ask prices for vanilla
options, e.g. for a call option we have:
∞ � �
∫0
bidEC = exp(−rT) xd𝛹𝜆WANG F(ST −K)+ (x)
√ (30)
= EC(K, T, S0 , r, q + 𝜆𝜎∕ T, 𝜎),

0 � �
∫−∞
askEC = − exp(−rT) xd𝛹𝜆WANG F−(ST −K)+ (x)
� √ � (31)
= EC K, T, S0 , r, q − 𝜆𝜎∕ T, 𝜎 ,

where EC stands for the classical risk-neutral Black–Scholes price of a European

call option. Conic prices are hence obtained by shifting the dividend yield in the

13
Digital Finance

Black-Scholes implied liquidity

0.5
T = 7 days
T = 28 days
T = 91 days
0.4 T = 182 days
Implied liquidity

0.3

0.2

0.1

0
0.5 1 1.5 2 2.5 3 3.5 4
Strike 104

Fig. 12  Black–Scholes-implied liquidity ( 𝜆bid and 𝜆ask ) for call options on 29 June 2018

original one-price formula. Note that the resulting option prices depend on the
parameter 𝜆 . In the special case, where 𝜆 is zero, bid price equals ask price and we
are again in the one-price framework. In other words, a value of 𝜆 equal to zero cor-
responds to a bid–ask spread of zero.
Implied liquidity is defined as the particular value of 𝜆 so that conic option
prices perfectly match the observed bid and ask prices in the market. In the par-
ticular case of BTC call options, we use the following procedure. First calculate
mid-prices corresponding to the bid and ask prices available in the market. For
each mid-price, the (Black–Scholes) implied volatility 𝜎impl is computed. Implied
liquidities 𝜆bid and 𝜆ask are then calculated by matching the market bid and ask
prices with the conic bid and ask prices, respectively, where 𝜎impl is plugged into
formulas (30) and (31).
The procedure above leads to a surface of implied liquidities. The higher 𝜆 is,
the less liquid the product is. However, since the implied liquidity is unitless, it
is only appropriate to compare the liquidity of products. Figure 12 displays the
implied liquidity smiles corresponding to the call options provided on 29 June
2018.
Implied liquidity follows an upward trend with respect to the strike price. Out
of the money BTC options are hence less liquid than their near the money ana-
logs. Figure  12 further indicates that the liquidity increases ( 𝜆 decreases) with
the maturity of the options.

13
 Digital Finance

5 Conclusion

With the advent of bitcoin options, a new opportunity arose to analyze the charac-
teristics of the bitcoin market. While time series models mainly focus on historical
dynamics of the bitcoin price, the option surface contains information about the risk-
neutral distribution governing the bitcoin log-returns. In this paper, we performed an
empirical study based on the available market prices of bitcoin vanilla options. We cali-
brated a series of elementary and more advanced market models on the option surface.
While the classical Black–Scholes model did not capture the surface very well, more
advanced models managed to produce a good fit. We observed that models including
some notion of stochastic volatility, like the Heston model and the VG-CIR model, gen-
erally do a better job. In the last part, the liquidity of the bitcoin option market was
examined based on the implied liquidity measure originating from conic finance theory.
In the money, long-term options were found to be the most liquid options in the bitcoin
market.

Appendix A: The Laplace market model

A brief summary of the Laplace market model is given below. More details can be
found in Madan (2016), and Madan and Wang (2017).

Laplace distribution

The density function of a Laplace distributed random variable L with mean 𝜇 and vari-
ance 𝜎 2 is given by
� √ �
1 �x − 𝜇�
fL (x) = √ exp − 2
𝜎
, (32)
2𝜎

and we denote L ∼ (𝜇, 𝜎 2 ) . We refer to the Laplace distribution with zero mean
and unit variance as the standard Laplace distribution L∗ . The characteristic function
of L is given by
( )−1
𝜎 2 u2
𝜙(u) = exp(iu𝜇) 1 + . (33)
2

We refer to Kotz et al. (2001) for a broad introduction to Laplace distributions and
its extensions.

Market model

Assume that the log-returns of an asset S are modelled via the Laplace distribution:
log(St+s ) − log(St ) ∼  𝜇s, 𝜎 2 s ,
( )
(34)

13
Digital Finance

and hence
� √ �
St ∼ S0 exp 𝜇t + 𝜎 tL∗ . (35)

We apply a mean-correcting measure change, assuming zero interest rates. The dis-
tribution is shifted to
� � � √ �
𝜎2t
St ∼ S0 exp log 1 − + 𝜎 tL∗ . (36)
2
Note that this equation is only valid for 𝜎 2 t < 2 . The characteristic function of the
log-price process log(St ) at time t equals:
( ( ( )))
𝜎2 t
exp iu log S0 + log 1 − 2
𝜙t (u) = E[exp(iu log(St ))] = . (37)
𝜎2 t 2
1+ 2
u

Remark This model is not identical to the variance gamma model with 𝜃 = 0

and 𝜈 = 1 . However, the VG(𝜎 , 1, 0) distribution equals the Laplace distribution
(0, 𝜎 2 ) . On the other hand, in the VG(𝜎 , 1, 0) model the log-returns are distributed
as
� √ �
log(St+s ) − log(St ) ∼ VG 𝜎 s, 1∕s, 0 . (38)

The two models hence only give the same results on a maturity of 1 year.

Vanilla option pricing

The price of a call option is given by

�
√ �
𝜎 T
C(K, T;𝜎) = S0 SL d, √ − KSL(d, 0) (39)
2

with

for x ≤ 0
{
1
(1+ s) exp((1 − s)x)
SL(x, s) = 2 , (40)
1 − 12 (1 − s) exp(−(1 + s)x) for x > 0

� �
𝜎2 T
log(S0 ∕K) + log 1 − 2
d= √ . (41)
𝜎 T
√
2

13
 Digital Finance

Appendix B: Option data of 29 June 2018

Underlying Strike Maturity (days) Is call Bid price (USD) Ask price (USD)

5906 5500 6.92 1 434.06 472.44

5906 5750 6.92 1 256.88 289.36
5906 6000 6.92 1 132.87 147.63
5906 6250 6.92 1 64.96 91.53
5906 6500 6.92 1 29.53 50.2
5906 6750 6.92 1 8.86 26.57
5906 5500 6.92 0 70.86 91.53
5906 5750 6.92 0 132.87 168.3
5906 6000 6.92 0 256.89 301.18
5906 6250 6.92 0 431.09 484.24
5906 6500 6.92 0 634.84 699.8
5906 6750 6.92 0 862.21 930.12
5906 7000 6.92 0 1098.43 1169.29
5872 5500 27.92 1 607.7 648.8
5872 6000 27.92 1 352.29 393.39
5872 6500 27.92 1 193.76 211.37
5872 7000 27.92 1 102.74 126.23
5872 7500 27.92 1 58.71 76.32
5872 8000 27.92 1 35.23 49.9
5872 8500 27.92 1 17.61 29.36
5872 9000 27.92 1 14.68 23.49
5872 10,000 27.92 1 2.94 14.68
5872 5500 27.92 0 240.73 275.96
5872 6000 27.92 0 478.53 528.43
5872 6500 27.92 0 810.27 874.85
5872 7000 27.92 0 1203.66 1288.79
5872 7500 27.92 0 1652.83 1743.84
5872 8000 27.92 0 2122.55 2222.36
5872 8500 27.92 0 2606.95 2712.63
5872 9000 27.92 0 3097.22 3202.9
5872 10,000 27.92 0 4089.5 4195.19
5872 11,000 27.92 0 5087.65 5190.41
5872 12,000 27.92 0 6085.29 6190.97
5867 5500 90.92 1 920.96 973.76
5867 6000 90.92 1 692.19 739.12
5867 6500 90.92 1 513.23 530.83
5867 7000 90.92 1 387.12 428.18
5867 7500 90.92 1 287.41 328.47
5867 8000 90.92 1 217.02 255.15
5867 8500 90.92 1 164.23 199.43

13
Digital Finance

Underlying Strike Maturity (days) Is call Bid price (USD) Ask price (USD)
5867 9000 90.92 1 126.11 158.37
5867 10,000 90.92 1 79.18 105.58
5867 11,000 90.92 1 43.99 70.39
5867 12,000 90.92 1 29.33 52.79
5867 13,000 90.92 1 17.6 38.13
5867 14,000 90.92 1 8.8 29.33
5867 15,000 90.92 1 2.93 23.46
5867 20,000 90.92 1 2.93 8.8
5867 5500 90.92 0 557.27 610.06
5867 6000 90.92 0 824.17 882.83
5867 6500 90.92 0 1132.14 1196.66
5867 7000 90.92 0 1492.9 1572.09
5867 7500 90.92 0 1886.08 1965.28
5867 8000 90.92 0 2317.27 2428.73
5867 8500 90.92 0 2748.46 2877.52
5867 9000 90.92 0 3206.04 3340.97
5867 10,000 90.92 0 4147.62 4253.21
5867 11,000 90.92 0 5112.65 5259.32
5867 12,000 90.92 0 6132.9 6238.49
5867 13,000 90.92 0 7073.79 7226.3
5867 14,000 90.92 0 8065.06 8217.57
5867 15,000 90.92 0 9056.33 9211.77
5867 20,000 90.92 0 14,027.34 14,203.31
5867 25,000 90.92 0 19,010.09 19,203.65
5867 30,000 90.92 0 23,989.9 24,206.92
5867 35,000 90.92 0 28,972.64 29,207.26
5867 40,000 90.92 0 33,955.38 34,210.53
5906 6250 181.88 1 1012.77 1086.58
5906 7500 181.88 1 658.45 729.31
5906 8750 181.88 1 439.95 504.9
5906 10,000 181.88 1 310.03 363.18
5906 12,500 181.88 1 159.44 203.73
5906 15,000 181.88 1 100.39 121.06
5906 20,000 181.88 1 41.34 59.05
5906 25,000 181.88 1 20.67 29.53
5906 30,000 181.88 1 11.81 20.67
5906 35,000 181.88 1 5.91 14.76
5906 40,000 181.88 1 2.95 11.81
5906 6250 181.88 0 1328.7 1417.28
5906 7500 181.88 0 2199.88 2350.47
5906 8750 181.88 0 3215.69 3398.77
5906 10,000 181.88 0 4328.8 4523.68
5906 12,500 181.88 0 6658.24 6900.35

13
 Digital Finance

Underlying Strike Maturity (days) Is call Bid price (USD) Ask price (USD)
5906 15,000 181.88 0 9064.7 9371.77
5906 20,000 181.88 0 13,966.13 14,391.31
5906 25,000 181.88 0 18,908.9 19,384.28
5906 30,000 181.88 0 23,863.48 24,442.2
5906 35,000 181.88 0 28,823.96 29,479.46
5906 40,000 181.88 0 33,787.4 34,519.66

References
Baur, D. G., Hong, K., & Lee, A. D. (2018). Bitcoin: Medium of exchange or speculative assets? Journal
of International Financial Markets, Institutions and Money, 54, 177–189.
Becker, J., Breuker, D., Heide, T., Holler, J., Rauer, H. P., & Böhme, R. (2013). Can we afford integrity
by proof-of-work? Scenarios inspired by the bitcoin currency. In: Böhme, R (Ed.), The economics of
information security and privacy (pp. 135–156). Berlin: Springer (Chap. 7).
Carr, P. P., Geman, H., Madan, D. B., & Yor, M. (2003). Stochastic volatility for Lévy processes. Math-
ematical Finance, 13(3), 345–382.
Carr, P. P., & Madan, D. B. (1999). Option valuation using the fast Fourier transform. Journal of Compu-
tational Finance, 2(4), 61–73.
Chen, C. Y., Härdle, W. K., Hou, A. J., & Wang, W. (2018). Pricing cryptocurrency options: The case of
CRIX and bitcoin. SSRN Electronic Journal,. https​://doi.org/10.2139/ssrn.31591​30.
Corcuera, J. M., Guillaume, F., Leoni, P., & Schoutens, W. (2009). Implied Lévy volatility. Quantitative
Finance, 9(4), 383–393.
Corcuera, J. M., Guillaume, F., Madan, D. B., & Schoutens, W. (2012). Implied liquidity: Towards sto-
chastic liquidity modeling and liquidity trading. International Journal of Portfolio Analysis and
Management, 1(1), 80–91.
Cretarola, A., & Figà-Talamanca, G. (2017). A confidence-based model for asset and derivative prices in
the bitcoin market. Accessed 10 Jan 2019.
Dwyer, G. P. (2015). The economics of bitcoin and similar private digital currencies. Journal of Finan-
cial Stability, 17, 81–91. Special Issue: Instead of the Fed: Past and Present Alternatives to the Fed-
eral Reserve System.
Garcia, D., Tessone, C. J., Mavrodiev, P., & Perony, N. (2014). The digital traces of bubbles: Feedback
cycles between socio-economics signals in the bitcoin economy. Journal of the Royal Society Inter-
face, 11, pp 1–18
Heston, S. L. (1993). A closed form solution for options with stochastic volatility with applications to
bonds and currency options. Review of Financial Studies, 6(2), 327–343.
Kjærland, F., Khazal, A., Krogstad, E. A., Nordstrøm, F. B., & Oust, A. (2018). An analysis of bitcoin’s
price dynamics. Journal of Risk and Financial Management, 11(4), 63.
Kotz, S., Kozubowski, T. J., & Podgórski, K. (2001). The Laplace distribution and generalizations. A
revisit with new applications. Berlin: Springer.
Kristoufek, L. (2013). Bitcoin meets Google trends and Wikipedia: Quantifying the relationship between
phenomena of the internet era. Scientific Reports, 3, 3415.
Madan, D. B. (2016). Adapted hedging. Annals of Finance, 12(3), 305–334.
Madan, D. B., Carr, P. P., & Chang, E. C. (1998). The variance gamma process and option pricing. Euro-
pean Finance Review, 2, 79–105.
Madan, D. B., & Milne, F. (1991). Option pricing with V.G. martingale components. Mathematical
Finance, 1(4), 39–55.
Madan, D. B., & Schoutens, W. (2016). Applied conic finance. Cambridge: Cambridge University Press.
Madan, D. B., Schoutens, W., & Wang, K. (2018). Bilateral multiple gamma returns: Their risks and
rewards. SSRN Electronic Journal,. https​://doi.org/10.2139/ssrn.32301​96.
Madan, D. B., & Seneta, E. (1990). The V.G. model for share market returns. Journal of Business, 63,
511–524.

13
Digital Finance

Madan, D. B., & Wang, K. (2017). Laplacian risk management. Finance Research Letters, 22, 202–210.
Nakamoto, S. (2008). Bitcoin: A peer-to-peer electronic cash system. Unpublished. http://bitco​in.org/
bitco​in.pdf.
Sato, K. (1999). Lévy processes and infinitely divisible distributions, Cambridge studies in advanced
mathematics, (Vol. 68). Cambridge: Cambridge University Press.
Scaillet, O., Treccani, A., & Trevisan, C. (2018). High-frequency jump analysis of the bitcoin market.
Journal of Financial Econometrics.
Segendorf, B. (2014). What is bitcoin? Sveriges Riksbank Economic Review, 2, 71–87.
Wang, S. S. (2000). A class of distortion operators for pricing financial and insurance risks. Journal of
Risk and Insurance, 67(1), 15–36.

Publisher’s Note  Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

13