Chemist_MZ 发表于 2012-5-11 10:45 
我姑且假设存续期是一年,如果有1000步的话u绝对不会有1.2这么大,这个你可以去看看。二叉树速度当然比si ...
vol=50%(10Y tenor 1Y expiry USD swpation vol quoted from SuperD), TTM=1Y, step size u=exp(0.5*sqrt(1/1000))=1.0159, u^1000=7.35*10^6, and you need to multiply the risk neutral probability then discount them back, lots of rounding would happen. I am not arguing with you whether the model is good and not, instead of telling you that it is not a market model, and no traders are applying the model in the same way in your mind. In the real trading market, American option is very liquid and the profit margin is low. To be able to apply a model then price back the market quotes, PDE+local vol is often used in major banks to give the price of American options. But I believe with you that binomial tree is better then the longstaff schwartz model used for Monte Carlo.