wilmott.com

I am looking for a way to price this:The security is a 2-year note that pays annual fixed rate coupons and par at maturity.On the same date every month, the fixing of a currency pair (e.g., USD/CHF) is observed, say X1. This is used to set the "band", e.g. X1 +/- 0.2000 for the next fixing X2, which then becomes the centre of the band for X3, and so on.The band width is constant throughout the 12 months of each year, but the band width for Year 2 is slightly wider (e.g., +/- 0.2000 for Year 1 and +/- 0.2400 for Year 2)For each of the year, if any of the fixings X1, X2, ..., X12 breaches its band ( e.g., X5 > X4+0.2000 ), the coupon for that year becomes zero (the note pays nothing for that year)It looks a little like a double no-touch option, but hitting one of the barriers in any one of the twelve months is enough to wipe out the coupon for the entire year. The setting of exact "barrier levels" based on the most recent fixing also complicates matters.Any idea on how to proceed will be appreciated.

For something sophisticated like that I would say that Monte Carlo is the best to start. The rules are not very hard to implement.1) Generates a lot of random path under the risk neutral measure2) Apply your rules to calculate the payoff of each path and then the expectation3) make the discounting (not easy if you assume stochastic interest rate but otherwise it is simple)and then you are done.I will think a bit about the PDE approach but it would be harder I guess. The very hard point is the boundary conditions.Your option is of maturity [0,2*T], (with T=1) I think it's easier to divide it into two periods [0,T] and [T,2T] since the in case of hit the coupons are not paid only for that year period.Each interval has to be divided into 12 subintervals. It's important here that your PDE should be forward since the value of the previous subinterval fix the no touch band of the next one.If Xn is observed at the end of the nth subinterval then in (n+1) you will have a double no touch with [Xn-0.2,Xn+0.2] with the payoff coupon at rate r if Xn+1 in [Xn-0.2,Xn+0.2] or 0 for the whole interval (so [0,T] or [T,2T] otherwise).That could make it.