void	s30acc (Nag_CallPut option, Integer n, const double p[], const double k[], const double s0[], const double t[], const double r[], double sigma[], Integer mode, Integer ivalid[], NagError *fail)

The function may be called by the names: s30acc or nag_specfun_opt_imp_vol.

3 Description

The Black–Scholes formula for the price of a European option is

P_{call} = S_{0} Φ (\frac{\ln (\frac{S_{0}}{K}) + [r + \frac{σ^{2}}{2}] T}{σ \sqrt{T}}) - K e^{- r T} Φ (\frac{\ln (\frac{S_{0}}{K}) + [r - \frac{σ^{2}}{2}] T}{σ \sqrt{T}}),

for a call option, and

P_{put} = K e^{- r T} Φ (\frac{- \ln (\frac{S_{0}}{K}) - [r - \frac{σ^{2}}{2}] T}{σ \sqrt{T}}) - S_{0} Φ (\frac{- \ln (\frac{S_{0}}{K}) - [r + \frac{σ^{2}}{2}] T}{σ \sqrt{T}}),

for a put option, where

Φ

is the cumulative Normal distribution function,

T

is the time to maturity,

S_{0}

is the spot price of the underlying asset,

K

is the strike price,

r

is the interest rate and

σ

is the volatility.

Given arrays of values

P_{i}

K_{i}

S_{0 i}

T_{i}

and

r_{i}

, for

i = 1, 2, \dots n

, s30acc computes the implied volatilities

σ_{i}

s30acc offers the choice of two algorithms. The algorithm of Glau et al. (2018) uses Chebyshev interpolation to compute the implied volatilities, and performs best for long arrays of input data. The algorithm of Jäckel (2015) uses a third order Householder iteration and performs better for short arrays of input data.

4 References

Glau K, Herold P, Madan D B and Pötz C (2018) The Chebyshev method for the implied volatility Accepted for publication in the Journal of Computational Finance

Jäckel P (2015) Let's be Rational Wilmott Magazine 2015(75) 40–53

5 Arguments

1: $option$ – Nag_CallPut Input

On entry: determines whether the option is a call or a put.

$option = Nag_Call$: A call; the holder has a right to buy.
$option = Nag_Put$: A put; the holder has a right to sell.

Constraint:

option = Nag_Call

Nag_Put

2: $n$ – Integer Input

On entry:

n

, the number of implied volatilities to be computed.

Constraint:

n \geq 0

3: $p [n]$ – const double Input

On entry:

p [i - 1]

must contain

P_{i}

, the

i

th option price, for

i = 1, 2, \dots, n

Constraint:

p [i - 1] \geq 0.0

, for

i = 1, 2, \dots, n

4: $k [n]$ – const double Input

On entry:

k [i - 1]

must contain

K_{i}

, the

i

th strike price, for

i = 1, 2, \dots, n

Constraint:

k [i - 1] > 0.0

, for

i = 1, 2, \dots, n

5: $s0 [n]$ – const double Input

On entry:

s0 [i - 1]

must contain

S_{0 i}

, the

i

th spot price, for

i = 1, 2, \dots, n

Constraint:

s0 [i - 1] > 0.0

, for

i = 1, 2, \dots, n

6: $t [n]$ – const double Input

On entry:

t [i - 1]

must contain

T_{i}

, the

i

th time, in years, to maturity, for

i = 1, 2, \dots, n

Constraint:

t [i - 1] > 0.0

, for

i = 1, 2, \dots, n

7: $r [n]$ – const double Input

On entry:

r [i - 1]

must contain

r_{i}

, the

i

th interest rate, for

i = 1, 2, \dots, n

. Note that a rate of 5% should be entered as 0.05.

8: $sigma [n]$ – double Output

On exit:

sigma [i - 1]

contains

σ_{i}

, the

i

th implied volatility, for

i = 1, 2, \dots, n

9: $mode$ – Integer Input

On entry: specifies which algorithm will be used to compute the implied volatilities. See Sections 7 and 8 for further guidance on the choice of mode.

$mode = 0$: The Glau et al. (2018) algorithm will be used. The nodes used in the Chebyshev interpolation will be chosen to achieve relative accuracy to approximately seven decimal places;
$mode = 1$: The Glau et al. (2018) algorithm will be used. The nodes used in the Chebyshev interpolation will be chosen to achieve relative accuracy to approximately $15$ decimal places, but limited by the machine precision;
$mode = 2$: The Jäckel (2015) algorithm will be used, aiming for accuracy to approximately $15$ – $16$ decimal places, but limited by machine precision.

Constraint:

mode = 0

1

2

10: $ivalid [n]$ – Integer Output

On exit:

ivalid [i - 1]

indicates any errors with the input arguments that prevented

σ_{i}

from being computed. If

ivalid [i - 1] \neq 0

sigma [i - 1]

contains

0.0

$ivalid [i - 1] = 0$: No error.
$ivalid [i - 1] = 1$: $P_{i} < 0.0$ .
$ivalid [i - 1] = 2$: $K_{i} \leq 0.0$ .
$ivalid [i - 1] = 3$: $S_{0 i} \leq 0.0$ .
$ivalid [i - 1] = 4$: $T_{i} \leq 0.0$ .
$ivalid [i - 1] = 5$: The combination of $P_{i}$ , $K_{i}$ , $S_{0 i}$ , $T_{i}$ and $r_{i}$ is out of the domain in which $σ_{i}$ can be computed. See Section 9 for further details.

11: $fail$ – NagError * Input/Output

The NAG error argument (see Section 7 in the Introduction to the NAG Library CL Interface).

6 Error Indicators and Warnings

NE_ALLOC_FAIL: Dynamic memory allocation failed.
See Section 3.1.2 in the Introduction to the NAG Library CL Interface for further information.
NE_BAD_PARAM: On entry, argument $⟨ value ⟩$ had an illegal value.
NE_INT: On entry, $mode = ⟨ value ⟩$ .
Constraint: $mode = 0$ , $1$ or $2$ .

On entry, $n = ⟨ value ⟩$ .
Constraint: $n \geq 0$ .
NE_INTERNAL_ERROR: An internal error has occurred in this function. Check the function call and any array sizes. If the call is correct then please contact NAG for assistance.
See Section 7.5 in the Introduction to the NAG Library CL Interface for further information.
NE_NO_LICENCE: Your licence key may have expired or may not have been installed correctly.
See Section 8 in the Introduction to the NAG Library CL Interface for further information.
NW_IVALID: On entry at least one input argument was invalid.
Check ivalid for more information.

7 Accuracy

mode = 0

1

then s30acc uses Chebyshev interpolation. For

mode = 0

it aims for relative accuracy to roughly single precision (approximately seven decimal places). For

mode = 1

it aims for relative accuracy to roughly double precision (approximately sixteen decimal places).

mode = 2

, a Householder iteration is used to achieve relative accuracy to roughly double precision (approximately sixteen decimal places). In practice there is very little difference in accuracy between

mode = 1

and

2

, though for more extreme input values,

mode = 2

is likely to be more accurate.

8 Parallelism and Performance

Background information to multithreading can be found in the Multithreading documentation.

s30acc is threaded by NAG for parallel execution in multithreaded implementations of the NAG Library.

Please consult the X06 Chapter Introduction for information on how to control and interrogate the OpenMP environment used within this function. Please also consult the Users' Note for your implementation for any additional implementation-specific information.

When

mode = 0

1

, s30acc uses an algorithm optimized for performance on large vectors of input data (

n ≳ 1000

) by exploiting vector instructions when available.

When

mode = 2

, s30acc uses an algorithm optimized for performance on smaller vectors of input data.

Numerical experiments suggest that for

n ≲ 1000

, the best performance is always achieved by choosing

mode = 2

, whereas for

n ≳ 1000

mode = 0

1

should be used, according to the desired accuracy.

9 Further Comments

The domain of inputs,

P_{i}

K_{i}

S_{0 i}

T_{i}

and

r_{i}

, for which s30acc is able to accurately compute

σ_{i}

is very large and should cover all practical applications of this function. Thus, encountering arguments for which

ivalid [i - 1] = 5

is returned is highly unlikely. Note that it is not possible to give a closed-form expression for the allowed range of arguments because this range is based on a transformation to the normalized call price.

Note that some formulations of the Black–Scholes equation also include the annual yield rate,

q

. If you wish to incorporate

q

here, then you must first compute the quantities

{\tilde{S}}_{0} = S_{0} e^{- q T}

\tilde{K} = K e^{- q T}

and

\tilde{r} = r - q

. s30acc should then be called with

{\tilde{S}}_{0}

\tilde{K}

and

\tilde{r}

in place of

S_{0}

K

and

r

respectively.

s30acc can also be used with Black's model for European futures options. In this case, the forward price,

F

, and the discount factor,

D

, are used, which are related to

S_{0}

via

F = S_{0} / D

. In addition, the option price is scaled by a factor

e^{- r T}

Approximately

7 \times n

of real allocatable memory and

3 \times n

of integer allocatable memory is required by the function.

10 Example

This example reads in values of

P_{i}

K_{i}

S_{0 i}

T_{i}

and

r

from a file, evaluates the implied volatilities,

σ_{i}

, and prints the results.

s30ac: FL CL CPP AD PY MB

NAG CL Interfaces30acc (opt_​imp_​vol)

▸▿ Contents

1 Purpose

2 Specification

3 Description

4 References

5 Arguments

6 Error Indicators and Warnings

7 Accuracy

8 Parallelism and Performance

9 Further Comments

10 Example

10.1 Program Text

10.2 Program Data

10.3 Program Results

NAG CL Interface
s30acc (opt_imp_vol)