Abstract
Consider an American option that pays $G(X^*_t)$ when exercised at time $t$, where $G$ is a positive increasing function, $X^*_t := \sup_{s\le t}X_s$, and $X_s$ is the price of the underlying security at time $s$. Assuming zero interest rates, we show that the seller of this option can hedge his position by trading in the underlying security if he begins with initial capital $X_0\int_{X_0}^{\infty}G(x)x^{-2}dx$ (and this is the smallest initial capital that allows him to hedge his position). This leads to strategies for trading that are always competitive both with a given strategy's current performance and, to a somewhat lesser degree, with its best performance so far. It also leads to methods of statistical testing that avoid sacrificing too much of the maximum statistical significance that they achieve in the course of accumulating data.
Introduction
In traditional quantitative finance, such as the Black-Scholes model, pricing a derivative requires assuming a specific probabilistic model for asset price movements, typically Geometric Brownian Motion. However, if the market deviates from these assumed distributions, hedging strategies can fail, potentially leading to significant losses. This is particularly true for path-dependent derivatives like lookback options, whose value depends on the running maximum ($X^*_t$) of the asset price.
This paper introduces a "probability-free" framework for pricing and hedging American lookback options. By utilizing game-theoretic probability, we derive the minimum capital required to guarantee a payout regardless of market behavior. In this paradigm, the "risk-neutral" measures of standard finance emerge naturally as worst-case scenarios rather than a priori assumptions.
Methods: Game-Theoretic Probability
The core methodology views market dynamics as a game between two players: the "Market," who determines prices, and the "Investor," who chooses trading positions. The interaction is formalized in a sequential protocol.
Technical Components
- Lookback Adjusters (LA): These are functions $F(X^*_t, X_t)$ representing the capital an investor can guarantee relative to the running maximum and current price.
- The "Spine": A central discovery is that an Admissible Lookback Adjuster (ALA)—one that cannot be improved upon—is characterized by its "spine," which is its value when the current price equals the running maximum ($X_t = X^*_t$).
- The $x^{-2}$ Density: The authors show that the "upper price" is calculated by integrating the payoff function $G$ against a $1/x^2$ density.
- Hedging Strategy: The paper provides an explicit trading rule: at any time $t$, the investor should hold a number of units of the asset equal to the right-derivative of the "spine" function at the current maximum.
Key Insights
- Emergent Risk-Neutrality: The $1/x^2$ distribution, often a result of Brownian motion assumptions in standard finance, emerges here as a game-theoretic necessity for safety against any price path. It turns a probabilistic assumption into a game-theoretic necessity.
- Geometric Constraints: For a hedging strategy to be optimal (admissible), the capital function must be linear in the current price $X_t$ and concave in the running maximum $X^*_t$.
- Statistical Robustness: These strategies can be applied to "insure against loss of evidence" in scientific testing. They allow researchers to "lock in" a fraction of the maximum significance (minimum p-value) achieved during data accumulation, preventing the decay of significance as more data arrives.
Results
The primary result is the pricing formula for an American option paying $G(X^*_t)$. The upper price at time zero is given by: $$X_0 \int_{X_0}^{\infty} G(x)x^{-2} dx$$
Furthermore, Theorem 4.1 provides the necessary and sufficient conditions for a capital process to be an ALA, mapping the space of perfect hedging strategies. The paper also proves that any increasing function $F$ can be guaranteed as a capital process if and only if it satisfies the integral inequality $\int_1^\infty F(y)y^{-2} dy \le 1$.
Limitations and Future Work
The current model assumes zero interest rates and frictionless trading. Real-world applications would need to account for the time value of money and transaction costs associated with frequent rebalancing. Additionally, while the methods are optimized for options depending on the running maximum, applying them to options depending on the running minimum (e.g., lookback puts) requires further refinement of the "stop at zero" logic. Future research may explore deeper connections between these adjusters and the Azéma–Yor solution to the Skorokhod embedding problem.
Summary updated