It is well known that hedging with implied volatility involves a PnL:


In the Wilmott paper (http://web.math.ku.dk/~rolf/Wilmott_WhichFreeLunch.pdf), they imply that the collective PnL from such a strategy is the integral of above expression across time.

However, this seems to assume that the market implied volatility stays constant at $σ_i$. Otherwise, one would also encounter the mark-to-market PnL governed by the sensitivity of the option to implied volatility among other terms:

$C_{σ}* (dσ)+.....$

Why is the mark to market PnL not accounted for in the above analysis?


2 Answers 2


Consider any function $f(S(t),K,t,T,\{x_i(t)\})$ with payoff $(S(T) - K)_+$ when $t=T$, where $\{x_i(t)\}$ are other variables/parameters so that at $t=0$ you are able to choose (i.e. calibrated) these so that your function matches the market price of the option: $f(S(0),K,0,T,\{x_i(0)\}) = C^{market}(t=0)$.

As the payoff of the option does not depend on $\{x_i(T)\}$, if you decide to look only at the option value at maturity, then you are free to keep these other variables fixed and only hedge changes in $S_t$. In this case, according to your chosen `reality' (this is 'marking to model' as opposed to 'marking to market') the change in option value is $$ df = \theta dt + \Delta dS + \frac{1}{2} \Gamma (dS)^2 $$ since you have chosen all the others variables/parameters to be constant. $dS$ is whatever change in stock price is observed.

However, if you decide to / or are forced to 'look' at the option value in the market before expiration, then your delta-hedge P/L will equal: $$ P\&L = C^{market}(t=0) + \int_0^u \left( \theta_t dt + \frac{1}{2} \Gamma_t (dS_t)^2 \right) - C^{market}(t=u) $$


If you assume that the vols $\sigma_r,\sigma_i$ are deterministic functions of time their formula (1) still holds $$\tag{1} dV(t)=\frac{1}{2}(\sigma^2_r(t)-\sigma_i^2(t))\,\Gamma^i(t)\,dt. $$ Integrating gives the accumulated hedge PnL $$ V(t)=\frac{1}{2}\int_0^t(\sigma^2_r(s)-\sigma_i^2(s))\,\Gamma^i(s)\,ds. $$ One could extend the derivation to the case of stochastic vol $\sigma_r(t)$ by applying Ito's formula to the call price with two state variables $C(t,S(t),\sigma_r(t))\,.$ I am however not sure how useful such a general result will be in practice. Formula (1) holds approximately for small time intervals when $\sigma_r(t)$ can be assumed to be nearly deterministic.

  • $\begingroup$ If volatility is stochastic, how can you write the true dynamics of $C$ in 2 state variables? $\endgroup$
    – user121416
    Dec 22, 2021 at 12:17
  • $\begingroup$ You are right. I missed something. Will delete that answer for now. $\endgroup$
    – Kurt G.
    Dec 22, 2021 at 12:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.