I guess it depends on the terms and conditions of the products you're trading. If you cannot trade at the NAV but your algo is bound to reflect it, then you take on the transaction costs, but usually your fees reflect them. Indeed for MF/HF, usually the bid/ask spread from NAV is fixed, so I would not call it a transaction cost per say, as it is likely to be a fixed amount no matter what amount you trade. Whether it be for the equity or the rate part, spreads from fair are usually set up in the termsheet anyway. I.e. you tell the client you the ZCB at libor + 20bps to cover for transaction costs. That also is a way to hedge liquidity/trading costs without needing a proper transaction cost model.As per the distribution and evalutation of the option price, I think it really depends on the liquidity. And usually you're right. Though a 5 y-o quartely traded fund might be thought to be a good investment, but 20 observation points is not enough not matter what your assumptions are, especially for a fund. How realiable is your vol estimate then...Well the proba of being closed to the boundary is small when properly managed, but 10% drop in the underlying can very well happen in the real world, and if it happens when your exposure to the fund is say 150%, the CPPI could still go below the floor, even if you were way away from it.

I am new to CPPI options and I have a problem pricing one. If youcould give me a hint about my problem I would be most grateful.Product 1) (not the real one, just to see if I am getting the thingsright). Suppose the CPPI is not guaranteed. An amount N is taken by the issuer, andinvested in an underlying (A) + risk free invesment (B). The leveragefactor is monthly reviewed and possibly modified depending on theevolution of A and according to some predefined and well-knownrules. Part of the benefit is given in coupons, what remains is givenat the end. Price spreads are taken into account.If we forget fees, it seems to me that the price isexactly N, even if A is tradable only once every month. The CPPIissuer is bearing no risk, just handling our money and giving back the(good or bad) results. Although it is pointless, if we would insist onpricing this through MC, A should be modeled using risk-neutralmeasure in order to get the price N, not real-rate.Product 2) (The REAL one). Like the previous, but now N isguaranteed at the end. So, we could writeProduct 2 = Product 1 + Option,where Option pays (BF - V)^+ in case the value v of the portfoliois less than the bond floor needed to honor the guaranteed amount.I understand that now liquidity is an issue, because continous deltahedging is not possible anymore. The document CPPI.DOC by GiacomoGalli linked in the Wikipedia refers to this kind of situations, andpoints out that in this reason:"The discrete nature also means that the relevant growth rate isnot the risk-free rate, but rather the real rate of return..."Now some doubts arise:* Beyond which frequency of trade for A (dayly, weekly, monthly) is itreasonable to use the real rate instead of the risk-free?. Or somehowa weighted average would be better? The price obtained, is still theamount needed to manage a hedging-portfolio? Or is it just the price that,probabilistically, would be fair if we could "play the game" manytimes, like in a statistical hedging?* If I would decide to price using MC, one way to do it would be toprice the whole Product 2 using the real rate. But another, to pricejust the Option using real rate, and Product 1 normally (because itis not sensible to liquidity), with risk-free rate. The results willbe very different. I prefer the second choice, but I don't find iteasy to justify.* Last and probably most disturbing: in general the value of an optionmay result increased using real probabilities instead ofneutral-risk probabilities, but it can also be decreased (if theoption is more valueble in case of poor performance of theunderlying). Surely I can't say that, "now that I can't hedgecontinuosly, the option is cheaper".Will you please shed some light on my poor understanding, or providereferences where I can learn these specific topics?Thank you, and I apologize for my broken English.

CPPI options are very simple to price.At any time the cushion can be written as : C(t)=d*S^mSo an auropean call on cushion is a power option with payoff at maturity : max(d*S^m-K;0)If S0'=d*S0^(m)*exp(0.5*sigma*sigma*m*(m-1)*T) then the call on cushion is simply [exp(r*(m-1)*T)] *CBS(S0',K,m*sigma,m*r,T)with CBS the standard Black-Scholes formula.