SERVING THE QUANTITATIVE FINANCE COMMUNITY

• 1
• 2

complyorexplain
Topic Author
Posts: 141
Joined: November 9th, 2015, 8:59 am

Odd result

Let phi be the normal density function (i.e. the Gaussian), and let F be the forward price, K the strike, V the volatility and T time to expiry. Define d1 and d2 as follows

d1 = (LN(F / K) + V ^ 2 * T / 2) / (V * SQRT(T))
d2 = (LN(F / K) - V ^ 2 * T / 2) / (V * SQRT(T))

The odd result is that

F phi(d1)  =  K phi(d2)

It's easily proved by substituting the standard exponential formula for the Gaussian. Then there is a load of cancellation and you find the ratio of phi(d1) to phi(d2) = exp(ln(K/F))  =  K/F

Is there any textbook reference for this, or has it been noted by others?

Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Odd result

Yes, it gets noticed by everyone who works with the BS or Black formula. Once you start computing Greeks, you naturally want to get rid of one of the d's in terms of the other and the result follows.

DavidJN
Posts: 1767
Joined: July 14th, 2002, 3:00 am

Re: Odd result

Lots of unusual results have been derived. For example, if you put a negative vol into the BS call price formula you get the negative of the put price.

Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Odd result

I think that was the basis of several old Star Trek episodes -- Mirror World

bearish
Posts: 5995
Joined: February 3rd, 2011, 2:19 pm

Re: Odd result

Let phi be the normal density function (i.e. the Gaussian), and let F be the forward price, K the strike, V the volatility and T time to expiry. Define d1 and d2 as follows

d1 = (LN(F / K) + V ^ 2 * T / 2) / (V * SQRT(T))
d2 = (LN(F / K) - V ^ 2 * T / 2) / (V * SQRT(T))

The odd result is that

F phi(d1)  =  K phi(d2)

It's easily proved by substituting the standard exponential formula for the Gaussian. Then there is a load of cancellation and you find the ratio of phi(d1) to phi(d2) = exp(ln(K/F))  =  K/F

Is there any textbook reference for this, or has it been noted by others?

Peter Carr is usually a good source for insights of this sort. E.g., take a look at his paper Deriving Derivatives of Derivative Securities.

complyorexplain
Topic Author
Posts: 141
Joined: November 9th, 2015, 8:59 am

Re: Odd result

I think that was the basis of several old Star Trek episodes -- Mirror World
Not with you I'm afraid. Where does Black Scholes feature in the old Star Trek?

Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Odd result

I think that was the basis of several old Star Trek episodes -- Mirror World
Not with you I'm afraid. Where does Black Scholes feature in the old Star Trek?
It doesn't -- small joke -- very small. Neg. vol => mirror world, get it?

Cuchulainn
Posts: 63379
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Odd result

I think that was the basis of several old Star Trek episodes -- Mirror World
Not with you I'm afraid. Where does Black Scholes feature in the old Star Trek?
It doesn't -- small joke -- very small. Neg. vol => mirror world, get it?
Even I can't remember that one .. 4th episode of season 2. It was always on TV on Friday afternoons around 4 pm.

Another mirror image from the same era.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl

complyorexplain
Topic Author
Posts: 141
Joined: November 9th, 2015, 8:59 am

Re: Odd result

That's enough Star Trek. The question is whether we would need to prove that result in a paper, or just assume it as a standard result, perhaps with a reference.

Alan
Posts: 10468
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Odd result

@comply,
If it was my paper, I would just say,
"familiar computations show F phi(d1)  =  K phi(d2)".

@Daniel,
Good Prisoner. Amazon prime has early season Star Trek's, so happened to watch it recently.

complyorexplain
Topic Author
Posts: 141
Joined: November 9th, 2015, 8:59 am

Re: Odd result

Thanks Alan. We worked out how to fit the proof into a footnote.

Cuchulainn
Posts: 63379
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Odd result

@comply,
If it was my paper, I would just say,
"familiar computations show F phi(d1)  =  K phi(d2)".
So long as it's not
1) its is obvious
2) a 6-year old child can do it.
3) using simple but mind-numbing algebra.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl

Cuchulainn
Posts: 63379
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Odd result

Lots of unusual results have been derived. For example, if you put a negative vol into the BS call price formula you get the negative of the put price.
Take a call spread option PDE, for example. Taking $\sigma1 = -0.4, \sigma2 = -0.2$ gives the same result as  $\sigma1 = 0.4, \sigma2 = 0.2$. Not surprising I suppose because PDE only 'sees' $\sigma^2$.

Ditto conclusion for put spread. In fact, my FDM gives the same results as Kirk 1995. Even for $\sigma1 = +0.4, \sigma2 = -0.2$ .

Is formula BS 'wrong'?

Getting a negative put price in a PDE would be heresy in the current context. Pardon me if I misread your post.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl

DavidJN
Posts: 1767
Joined: July 14th, 2002, 3:00 am

Re: Odd result

I could see a 'correct' model producing garbage output when supplied with garbage input. Good argument validation would take care of that particular one. Anyway, the result with the negative of the put price emerging from a BS call formula supplied with a negative vol is in a paper by Rolf Poulsen from the University of Copenhagen titled Four Things You Might Not Know About the Black Scholes Model. I think there is a formal proof of the result if memory serves me correct. And at least one typo (in the pseudo code).

Cuchulainn
Posts: 63379
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

Re: Odd result

Something else (thinking out loud) is that the SDEs

dS/S  = rdt + sig dW
dS/S = rdt - sig dW

give the same option value (antithetic variates) ? It would be consistent with the PDE approach?

//
Poulson paper,
https://papers.ssrn.com/sol3/papers.cfm ... _id=991344

$c(-\sigma) = - p(\sigma)$
Last edited by Cuchulainn on August 24th, 2020, 11:14 am, edited 1 time in total.
My C++ Boost code gives
262537412640768743.999999999999250072597198185688879353856337336990862707537410378210647910118607313

http://www.datasimfinancial.com
http://www.datasim.nl