I've made up a code that solves PDEs for a vanilla call/put. But I cant get i right for exotics, where one needs some more state variables, so therefore I'd appreciate comments (have I missunderstood it all?):1) Suppose I have a european asian with arithmetic discrete average to start with, then I need to keep track of the average (Ai), the value will be V(Si,ti,Ai) in the three dimensional grid.2) If I understands things right, the statevariable (average) will be constant except when we come to a point when a new sample takes place and affects average to date. Since statevariable is constant between sampling points we simply have to solve N regular B & S between these dates. When we start backwards, we know the initial conditions max (AT-K,0). When going backwards we can use the same bounding conditions as usual when S goes to zero or "the limit", but we dont have to set bounding cond for the new statevariable? 3) When we get to a timepoint when sampling occurs, we instead have to use a jump condition when updating the optionvalue (arbitrage condition). 4) And so, if I have completed the grid calc. and whants to pick an optionvalue at the current shareprice and time I can get it from state: V(S0,t0,0) ?

Rather: V(S0,t0,A0) where A0 is the current average.

Thanks,But if the first sampling point for the average is, lets say 6 months ahead from current time. Shoud I then realy define the average to equal the current share price, as you say?

I am a postgraduate student. I study "MSc RISK MANAGEMENT & FINANCIAL SERVICES", I have some strong educational background in the trading and use of financial derivatives. I need some good idea for a dissertation that will combine financial risk management techniques, with the uses in financial institutions, banks, intermediaries, hedge funds etc.I am particullary interested in the uptodate key sector aspects.