Juggling snowballs

Previous work on the valuation of cancellable snowball swaps in the Libor market model suggested the use of nested Monte Carlo simulations was needed to obtain accurate prices. Here, Christopher Beveridge and Mark Joshi introduce new techniques that allow tight bounds to be obtained quickly without sub-simulations. A key part of their work is a new identification of points of provably sub-optimal exercise

Many advances have been made in recent years in the pricing of early exercisable derivatives using Monte Carlo simulation, but it is still a problem. The fundamental problem is that while existing techniques work, they generally require a degree of specialist handcrafting or sub-Monte Carlo simulations. The former requires an undesirable amount of research for new payouts and the latter results in models that are slower than is satisfactory.

We study the pricing of contracts that give the holder