This thesis investigates a robust approach to option pricing that departs from the classical stochastic assumptions of the Black-Scholes-Merton model. While traditional models rely on specific assumptions on the price dynamics of the underlying asset, such as the Geometric Brownian Motion, which often fail to capture discontinuous and jump-like dynamics, this work adopts a methodology based on Game Theory and online trading algorithms, following the framework developed by DeMarzo, Kremer and Mansour in [4] and [3]. The core of the analysis rests on the concept of regret minimization, where the performance of a dynamic trading strategy is compared to the best static alternative ex-post. By assuming an absence of arbitrage and imposing weak constraints on the asset’s price path (specifically bounds on the quadratic variation (Q) and the maximum single-period return (M)), robust upper and lower pricing bounds for European call options are derived. The method employs specific online trading algorithms to establish these bounds. The generic algorithm is used to determine the upper bound, while a combination of squaremomentum and trade-once strategies is developed to construct the lower bound. Furthermore, the problem of finding the optimal bound is modeled as a finite-horizon Zerosum game between an investor and an adversary who controls the price path. Using the Minimax theorem, the value of this game is shown to correspond to the optimal option price bound. Finally, the derived robust bounds are compared numerically with the standard Black- Scholes prices. The results demonstrate that while the game theoretic bounds are naturally wider due to the lack of distributional assumptions, they exhibit a pricing behavior and trend that is remarkably consistent with the classical model, validating the robustness of the proposed approach.
Questa tesi analizza un approccio robusto per il pricing delle opzioni che si discosta dalle classiche assunzioni stocastiche del modello di Black-Scholes-Merton. Mentre i modelli tradizionali si basano su specifici processi di prezzo del titolo sottostante, come il Moto Browniano Geometrico, che spesso falliscono nel catturare il caso di processi discontinui e con salti, questo lavoro adotta una metodologia basata sulla Teoria dei Giochi e su algoritmi di trading online, seguendo il framework sviluppato da DeMarzo, Kremer e Mansour in [4] e [3]. Il nucleo dell’analisi poggia sul concetto di minimizzazione del rimpianto (regret minimization), dove la performance di una strategia di trading dinamica viene confrontata ex-post con la migliore alternativa statica. Assumendo una condizione di assenza di arbitraggio e imponendo vincoli deboli sul percorso di prezzo dell’asset (in particolare limiti sulla variazione quadratica (Q) e sul rendimento massimo di un singolo periodo (M)) vengono derivati limiti di prezzo robusti, superiori e inferiori, per le opzioni call europee. Il metodo impiega specifici algoritmi di trading online per stabilire tali limiti. L’algoritmo generic viene utilizzato per determinare il limite superiore, mentre una combinazione di strategie square-momentum e trade-once è sviluppata per costruire il limite inferiore. Inoltre, il problema di trovare il limite ottimo viene modellato come un Gioco a somma zero a orizzonte finito tra un investitore e un avversario che controlla il percorso del prezzo. Applicando il teorema del Minimax, si dimostra che il valore di questo gioco corrisponde al limite ottimo del prezzo dell’opzione. Infine, i limiti robusti derivati vengono confrontati numericamente con i prezzi standard di Black and Scholes. I risultati dimostrano che, sebbene i limiti basati sulla teoria dei giochi siano naturalmente più ampi a causa della mancanza di assunzioni sulla distribuzione, essi mostrano un comportamento di prezzo e un trend notevolmente coerenti con il modello classico, validando la robustezza dell’approccio proposto.
Option pricing through game theory: how online trading algorithms and zero-sum games lead to robust pricing bounds
Boni, Andrea
2025/2026
Abstract
This thesis investigates a robust approach to option pricing that departs from the classical stochastic assumptions of the Black-Scholes-Merton model. While traditional models rely on specific assumptions on the price dynamics of the underlying asset, such as the Geometric Brownian Motion, which often fail to capture discontinuous and jump-like dynamics, this work adopts a methodology based on Game Theory and online trading algorithms, following the framework developed by DeMarzo, Kremer and Mansour in [4] and [3]. The core of the analysis rests on the concept of regret minimization, where the performance of a dynamic trading strategy is compared to the best static alternative ex-post. By assuming an absence of arbitrage and imposing weak constraints on the asset’s price path (specifically bounds on the quadratic variation (Q) and the maximum single-period return (M)), robust upper and lower pricing bounds for European call options are derived. The method employs specific online trading algorithms to establish these bounds. The generic algorithm is used to determine the upper bound, while a combination of squaremomentum and trade-once strategies is developed to construct the lower bound. Furthermore, the problem of finding the optimal bound is modeled as a finite-horizon Zerosum game between an investor and an adversary who controls the price path. Using the Minimax theorem, the value of this game is shown to correspond to the optimal option price bound. Finally, the derived robust bounds are compared numerically with the standard Black- Scholes prices. The results demonstrate that while the game theoretic bounds are naturally wider due to the lack of distributional assumptions, they exhibit a pricing behavior and trend that is remarkably consistent with the classical model, validating the robustness of the proposed approach.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_03_Boni.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: testo della tesi
Dimensione
631.21 kB
Formato
Adobe PDF
|
631.21 kB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/250738