Options Mastery: Theory to Trading #6

Options Trading in a Stochastic Volatility World

Junsu Park ™

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 1 / 26

Table of Contents

1 Introduction

2 Model-Free Greeks

3 P&L Dynamics

4 P&L Decomposition

5 Minimum Variance Delta

6 Skew Stickiness Ratio

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 2 / 26

Assumptions

The underlying asset’s price process is governed by


(1)
dSt = rSt dt + σt dWt ,

(2)
dσ = f (σt )dt + f2 (σt )dWt ,
 t (1) 1 (2)

dWt dWt = ρdt.
The exact specification of the model (f1 and f2 ) is unknown.
For any positive strike price, there exists a listed option contract.
An options market maker provides infinite liquidity with zero bid-ask
spread for all listed options, with knowledge of the true model.
Consequently, the options market is frictionless and efficient.
Most of the results can easily be generalized to multi-factor stochastic
volatility models [1].

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 3 / 26

Option Prices and Source of Risk

In this stochastic volatility world, the option price depends on Ft and

d
σt as FT |Ft = FT |Ft , σt .
This implies that, unlike in the Black-Scholes model, an option has
two sources of risk: dFt and dσt .
Let us denote the total implied variance of an option with expiry time
T at time t as a function of σt and log-moneyness k:
2
Ft ek (T − t)


w(t, σt , k) = σBS
Recall that in a log-type model, Ft does not provide information on
the implied volatility at a fixed log-moneyness, as long as σt is known.

Other Sources of Risk

In reality, additional factors contribute to option P&L, including interest
rates, dividends, volatility-spot correlation, and volatility of volatility.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 4 / 26

Table of Contents

1 Introduction

2 Model-Free Greeks

3 P&L Dynamics

4 P&L Decomposition

5 Minimum Variance Delta

6 Skew Stickiness Ratio

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 5 / 26

Greeks Calculation
Since the exact specifications of the model are unknown, the
functional forms of the Greeks are not available.
However, using market prices of options provided by the omniscient
market maker, some Greeks can be inferred.
Let Vt be the time-t price of an option with strike K:
Vt = e−r(T−t) [Ft Φ (d1 (t, σt , kt )) − KΦ (d2 (t, σt , kt ))] (1)
where  p
√ k 1

d 1 (t, σ t , k) = − + 2 w(t, σt , k),

 w(t,σt ,k)
p
d2 (t, σt , k) = − √ k − 21 w(t, σt , k),

 w(t,σt ,k)

k = log(K/F ).
t t

Recall that in a log-type model,

∂ Vt ∂ Vt ∂ 2 Vt 2 ∂ 2 Vt 2
Ft + K = Vt and F = K . (2)
∂F ∂K ∂ F2 t ∂ K2
Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 6 / 26
Greeks Calculation

Using equations (1) and (2), we can derive the following results. The
proof is left as an exercise to the readers.

Model-free Delta and Cash Gamma

√
∂ Vt −r(T−t) −r(T−t) ∂ w
=e Φ (d1 (t, σt , kt )) −e φ (d1 (t, σt , kt ))
∂F | {z } ∂k k=kt
=∆BS
∂ 2 Vt 2 Ft φ (d1 (t, σt , kt ))
2
Ft = e−r(T−t) p g (t, σt , kt )
∂F w (t, σt , kt )
| {z }
=$ΓBS

k ∂w 2 1 1 1 ∂ w 2 1 ∂ 2w
    
where
g (t, σt , k) = 1 − − + + [2]
2w ∂ k 4 w 4 ∂k 2 ∂ k2
√ 2  √ 2 √
√ ∂2 w

k ∂ w w ∂ w
= 1− √ − + w ≥ 0.
w ∂k 4 ∂k ∂ k2

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 7 / 26

Table of Contents

1 Introduction

2 Model-Free Greeks

3 P&L Dynamics

4 P&L Decomposition

5 Minimum Variance Delta

6 Skew Stickiness Ratio

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 8 / 26

P&L Dynamics

Consider a self-financing portfolio, Π, that holds one option. Its P&L

dynamics is given by:

dΠt =dVt + r(Πt − Vt )dt

∂ Vt ∂ Vt 1 ∂ 2 Vt 2 2
= dt + dFt + F σ dt + r (Πt − Vt ) dt
2 t t
t }
|∂{z |∂ F{z } |2 ∂ F {z |
} Funding {z }
Cost P&L
Theta P&L Delta P&L Gamma P&L
∂ Vt ∂ 2 Vt 1 ∂ 2 Vt
+ dσt + dFt dσt + (dσt )2
|∂ σ{z } | ∂ F∂ σ {z 2 ∂σ2
} | {z }
Vega P&L Vanna P&L Volga P&L

Thus, unlike in the Black-Scholes model, where the sources of P&L

are primarily Delta, Gamma, and Theta, this model also includes
Vega, Vanna, and Volga as additional sources of P&L.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 9 / 26

Dynamics of Implied Volatility
In reality, σt is not directly observable but the implied volatility curve
is available, providing some information about dσt .
The dynamics of the implied volatility for a fixed strike is as follows:
p p p
p ∂ w(kt ) ∂ w(kt ) 1 ∂ 2 w(kt )
d w(kt ) = dt + dσt + (dσt )2
∂t ∂σ 2 ∂σ2
√  √ √ !
∂2 w dFt 2

∂ w dFt 1 ∂ w
− + +
∂ k k=kt Ft 2 ∂ k k=kt ∂ k2 k=kt Ft
√
∂2 w dFt
− dσt .
∂ k∂ σ k=kt Ft

The dynamics of the implied volatility for a fixed log-moneyness is as

follows:
p p p
p ∂ w(k0 ) ∂ w(k0 ) 1 ∂ 2 w(k0 )
d w(k0 ) = dt + dσt + (dσt )2 .
∂t ∂σ 2 ∂σ2
Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 10 / 26
P&L Dynamics using Implied Volatility

Let us rewrite the P&L dynamics using the implied volatility of the strike
K:
dFt 2
 
1 p p
dΠt =rΠt dt + ∆BS dFt + $ΓBS + $ΓBS w (t, σt , kt )d w (t, σt , kt )
2 Ft
dFt p 1  p 2
− $ΓBS d2 (t, σt , kt ) d w (t, σt , kt ) + $ΓBS d1 d2 d w (t, σt , kt ) .
Ft 2

Inconsistent P&L Decomposition

Some market practitioners
q decompose P&L similarly to the above
t ,kt )
expression, assuming d w(t,σ
T−t ∝ dσt . However, this is not consistent
with a conventional stochastic volatility model.
p
Now, by substituting the SDE for w (t, σt , kt ) into the equation, we can
rearrange and expand the SDE accordingly:

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 11 / 26

P&L Dynamics using Implied Volatility
 √ 
∂ w
dΠt = ∆BS − e−r(T−t) φ (d1 ) dFt + rΠt dt
∂k
| {z }
= ∂∂ VF
 √ √  √  √  
√ ∂ w ∂2 w dFt 2
 
1 ∂ w ∂ w
+ $ΓBS 1 + w + + d2 2 + d1
2 ∂k ∂ k2 ∂k ∂k Ft
| {z }
2V
=∂ Ft2
∂ F2
√ √
√ ∂ w √ ∂ w
+ $ΓBS w dt + $ΓBS w dσt
| {z ∂ t } | {z ∂ σ }
= ∂∂Vt −rVt = ∂∂ σV
√ √  √ 
√ ∂2 w

∂ w ∂ w dFt
−$ΓBS w + d2 1 + d1 dσt
∂ k∂ σ ∂σ ∂k Ft
| {z }
2
= ∂∂σ ∂VF Ft
√  √ 2 !
1 √ ∂2 w ∂ w
+ $ΓBS w + d1 d2 (dσt )2
2 ∂σ2 ∂σ
| {z }
2V
=∂
∂σ2

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 12 / 26

Table of Contents

1 Introduction

2 Model-Free Greeks

3 P&L Dynamics

4 P&L Decomposition

5 Minimum Variance Delta

6 Skew Stickiness Ratio

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 13 / 26

P&L Decomposition

The evolution of the implied volatility of a fixed log-moneyness is

completely explained by dt, dσt , and (dσt )2 and this allows for an
approximate P&L decomposition.
However, without knowing the exact dynamics of σt , it is not possible
to fully separate the impacts of dt, dσt , and (dσt )2 .
Therefore, the P&L can only be approximately decomposed into
1 Funding Cost P&L = r (Πt − Vt ) dt
2 Interest Rate Theta P&L = rVt dt
3 Delta P&L = ∂∂VFt dFt
1 ∂ 2 Vt
4 Gamma P&L = (dFt )2
2 ∂ F2 
5 Volatility Theta P&L = ∂∂Vtt − rVt dt
+ Vega P&L + ∂∂Vσt dσt
2
+ Volga P&L + 12 ∂∂ σV2t (dσt )2
6 Vanna P&L = ∂∂F∂Vtσ dFt dσt

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 14 / 26

Volatility Theta, Vega, and Volga P&L

Approximation of Volatility Theta + Vega + Volga P&L

1 ∂ 2 Vt
 
∂ Vt ∂ Vt
− rVt dt + dσt + (dσt )2
∂t ∂σ 2 ∂σ2
 
p p d1 (kt )d2 (kt ) p
≈ $ΓBS w(kt )d w(k0 ) + (d w(k0 ))2
2

R T  ∂ Vt  2

0 − rVt dt + ∂∂Vσt dσt + 12 ∂∂ σV2t (dσt )2
∂t
lim R p 
T→0 T $Γ w(kt )d w(k0 ) + d12d2 (d w(k0 ))2
p p
0 BS
 √ √  √  
2
RT p ∂ w(kt ) ∂ w(kt ) 1 ∂ w(kt ) d1 d2 2
0 $ΓBS w(kt ) dt + ∂ σ dσt + 2 + √ (dσt )
∂t ∂σ2 w(kt )
= lim  √ √  √  
T→0 R T ∂ w(k0 ) ∂ w(k0 ) 2 w(k0 )
1 ∂ d1 d2
p
0 $ΓBS w(kt ) dt + ∂ σ dσt + 2 + √ (dσt ) 2
∂t ∂σ2 w(kt )

=1.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 15 / 26

Vanna P&L

Approximation of Vanna P&L

∂ Vt
dFt dσt
∂ F∂ σ
√ √ ! !
p ∂ w ∂ w p dFt
≈ −$ΓBS w(kt )d + d2 1 + d1 d w(t, σt , k0 )
∂k k=k0 ∂k k=kt Ft

RT ∂ Vt
0 ∂ F∂ σ dFt dσt
lim  √  √  
T→0 R T
w(kt )d ∂∂ kw + d2 1 + d1 ∂∂ kw d w(t, σt , k0 ) dF
p p
− 0 $ΓBS Ft
t
k=k0 k=kt
 √ √  √ 
RT p ∂2 w ∂ w(kt ) ∂ w dFt
0 $Γ BS w(k )
t ∂ k∂ σ
k=kt
+ d2 ∂σ
1 + d1 ∂k
k=kt Ft dσt
= lim  √ √  √ 
T→0 R T 2 ∂ w(k )
w(kt ) ∂∂ k∂ σw + d2 ∂ σ 0 1 + d1 ∂∂ kw dFt
p
0 $ΓBS
k=k0 Ft dσt
k=kt

=1.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 16 / 26

Table of Contents

1 Introduction

2 Model-Free Greeks

3 P&L Dynamics

4 P&L Decomposition

5 Minimum Variance Delta

6 Skew Stickiness Ratio

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 17 / 26

Vega Hedging and Minimum Variance Delta

To perfectly hedge the dσt risk, you would need to hold an additional
derivative that has exposure to dσt [3].
However, dynamically hedging the volatility risk using another
derivative can be costly in practice.
In practice, it is often preferred to hedge the Vega risk by adjusting the
position in futures contracts to minimize the variance of the P&L [4].
The concept of minimum variance delta arises naturally because
volatility is correlated with the returns of the underlying asset [5].

Minimum Variance Delta in a Stochastic Volatility World

t+h R
∂ Vt ∂ Vt ρf2 (σt ) ∂ Vt ∂ Vt dσu dFu
∆min var = + = + lim Rt t+h
∂F ∂ σ σt Ft ∂F ∂ σ h→0 t (dFu )2
| {z }
Instantaneous OLS β

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 18 / 26

Minimum Variance Delta

The minimum variance Delta can be expressed in terms of implied

volatility.
√ R t+h p
∂ Vt ρf2 (σt ) ∂ Vt ∂ w ρf2 (σt ) 1 ∂ Vt d w(u, σu , k)d log(Fu )
= √ = √ lim t R t+h
∂ σ σt Ft ∂ w ∂ σ σt Ft Ft ∂ w h→0 t (d log(Fu ))2 k=kt

Minimum Variance Delta using Implied Volatility

R t+h p
∂ Vt d w(u, σu , k)d log(Fu )
∆min var = + e−r(T−t) φ (d1 ) lim t
R t+h
∂F h→0
t (d log(Fu ))2 k=kt
 
 R t+h p √ !
−r(T−t) 
 t d w(u, σu , k)d log(Fu ) ∂ w 
=e Φ (d1 ) + φ (d1 ) h→0
lim R t+h − 
t (d log(Fu ))2 ∂ k k=kt

 k=kt 
| {z }
Difference between dynamic and static volatility slopes.

This representation is valid if the model is log-type and Markovian.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 19 / 26
Minimum Variance Delta-hedged P&L Dynamics
Again, consider a self-financing portfolio, Π, that holds one lot of an option
V and is short ∆min var lots of F. Its P&L dynamics is given by:
dΠt =dVt − ∆min var dFt + r(Πt − Vt )dt
∂ Vt 1 ∂ 2 Vt 2 2 ∂ 2 Vt 1 ∂ 2 Vt
 
2
= + F σ + dFt dσt + (dσt ) dt
∂t 2 ∂ F 2 t t ∂ F∂ σ 2 ∂σ2
 
∂ Vt ρf2 (σt )
+ dσt − dFt +r(Πt − Vt )dt.
∂σ σt Ft
| {z }
=Vega P&L −(∆min var − ∂∂ VF )dFt

The proportion of the variance of Vega P&L that can be hedged by holding
the minimum variance Delta is (1 − ρ 2 ):
R T  ∂ Vt 2  ρf2 (σt )
2
0 ∂σ dσ t − σ F
t t
dF t
2 = 1 − ρ 2.
R T  ∂ Vt
0 ∂ σ dσt

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 20 / 26

Table of Contents

1 Introduction

2 Model-Free Greeks

3 P&L Dynamics

4 P&L Decomposition

5 Minimum Variance Delta

6 Skew Stickiness Ratio

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 21 / 26

Skew Stickiness Ratio

Skew-Stickiness Ratio (SSR) [6], [7]

R t+h p , √
t d w(u, σu , 0)d log(Fu ) ∂ w
Skew Stickiness Ratio = lim R t+h
h→0
t (d log(Fu ))2 ∂k k=0

Accurate estimation and prediction of SSR is crucial when hedging the

minimum variance Delta.
In the sticky strike regime, SSR is 1, while in the sticky moneyness
regime, SSR is 0.
It is empirically observed that SSR is usually between 0.9 and 1.7 for
S&P 500 options [8].
This implies that the beta of ATM volatility with respect to log return
is generally steeper than the slope of the static volatility curve.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 22 / 26

Skew Stickiness Ratio

SSR is sensitive to the choice of a volatility model.

Let R(T) denote SSR as a function of time to expiry.
In a local volatility model, lim R(T) = 2.
T→0
In a conventional stochastic volatility model, such as the Heston
model, lim R(T) = 2 and lim R(T) = 1 in general.
T→0 T→∞
3
In a rough volatility model, lim R(T) = H + where H is the Hurst
T→0 2
exponent [8].
See [7] to learn more about this topic.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 23 / 26

Reference I

[1] Peter Christoffersen, Steven Heston, and Kris Jacobs. “The shape and
term structure of the index option smirk: Why multifactor stochastic
volatility models work so well”. In: Management Science 55.12 (2009),
pp. 1914–1932.
[2] Timothy Klassen. “State of the Smile: The Ever-Surprising Evolution
of the Equity/Listed Options Market”. In: CEO/Co-Founder, Vola
Dynamics LLC. Available at Vola Dynamics LLC booth or via email:
[email protected]. QuantMinds Conference. London, Nov.
2023.
[3] Fabrice D Rouah. The Heston model and its extensions in Matlab and
C. John Wiley & Sons, 2013.
[4] Gurdip Bakshi, Charles Cao, and Zhiwu Chen. “Empirical performance
of alternative option pricing models”. In: The Journal of finance 52.5
(1997), pp. 2003–2049.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 24 / 26

Reference II

[5] John Hull and Alan White. “The pricing of options on assets with
stochastic volatilities”. In: The journal of finance 42.2 (1987),
pp. 281–300.
[6] Lorenzo Bergomi. “Smile dynamics IV”. In: Available at SSRN
1520443 (2009).
[7] Lorenzo Bergomi. Stochastic volatility modeling. CRC press, 2015.
[8] Jim Gatheral and Bloomberg Quant Seminar. “Computing
skew-stickiness”. In: (2023).

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 25 / 26

Disclaimer

The views expressed in this presentation/article are solely those of the author and do not
necessarily reflect the views of Optiver or its affiliates. This content is intended for
educational and informational purposes only and should not be construed as investment
advice or a recommendation to engage in any specific trading strategies or transactions.
Options trading involves substantial risk and is not suitable for all investors. Individuals
should carefully consider their own financial situation and risk tolerance before engaging
in options trading or implementing any trading strategies discussed herein. The author
makes no representations or warranties as to the accuracy or completeness of any
information provided in this presentation/article and shall not be liable for any errors or
omissions in the content or for any actions taken based on the information provided
herein. Readers are encouraged to consult with a qualified financial advisor or
investment professional before making any investment decisions. By accessing this
content, you agree to release the author and Optiver from any liability arising from your
use of or reliance on the information provided herein.

Junsu Park ™ Options Trading in Stochastic Vol August 3, 2024 26 / 26