On optimal arbitrage (Q990375)

The paper is devoted to a characterization of the highest return on investment (best arbitrage with respect to the market portfolio) on a given time horizon. The market model that they consider does not consist of the prices of a collection of \(n\) basic assets, bit of their capitalization in the market. They propose a Black-Scholes type model for these quantities. This may be stretching it too far, but it is a model anyway. If \(X_i(t)\) denotes the capitalization of the \(i\)th asset at time \(t\), the market portfolio is \(\mu_i(t)/X(t)\), where \(X(t):=\sum_i X_i(t)\) is the market portfolio. After eight preparatory sections, in which the preliminaries are laid down, they devote the rest of the paper examine under what conditions there exists arbitrage relative to the market. They define a portfolio (investment strategy) \(\rho\), with initival value \(v\) to be an arbitrage relative to \(\mu\) if \[ P(V^{v,\rho}(T)\geq P(V^{v,\mu}(T))= 1\text{ and }P(V^{v,\rho}(T)> P(V^{v,\mu}(T))> 0. \] To state the question they pose and answer we need the following definition. Set \[ u(T):= \text{inf}\{w> 0|\exists\rho\in{\mathcal H}\text{ s.t. }V^{wX(0),\rho}(T)\geq X(T)as\}, \] where \({\mathcal H}\) is the collection of admissible investment strategies. \(u(t)\) is the smallest amount that must be invested in some portfolio that beats the market portfolio. They prove that it is the probability that an appropriately defined process never leaves \([0,\infty)^n\).