Martingale Theory and Processes
A martingale is a process modelling a fair game, in which the best prediction of the next value given the entire past is the current value, with no systematic upward or downward drift.
Definition
A martingale is a sequence or family of integrable random variables adapted to a filtration such that the conditional expectation of each future value given the present information equals the current value, formalising a fair game and generalising sums of independent zero-mean increments.
Scope
This area covers filtrations, adapted processes, and conditional expectation, the definitions of martingales, submartingales, and supermartingales, stopping times and the optional stopping theorem, Doob's maximal and upcrossing inequalities and the martingale convergence theorems, the Doob decomposition, and the role of martingales in stochastic integration and limit theorems.
Sub-topics
Core questions
- What does the martingale property say about predicting the future from the past?
- How do stopping times interact with martingales through optional stopping?
- Under what integrability conditions does a martingale converge?
- How do martingales underpin stochastic integration and limit theorems?
Key theories
- Martingale convergence theorem
- A martingale bounded in a suitable sense converges almost surely, and a uniformly integrable martingale converges both almost surely and in mean to a limiting random variable that closes it, giving a powerful tool for almost-sure limits.
- Optional stopping theorem
- Under appropriate conditions a stopped martingale has the same expectation as its starting value, so stopping a fair game at a random time chosen without foresight cannot change its expected outcome, a result with wide applications to gambling, random walks, and finance.
Clinical relevance
Martingale theory provides the conceptual foundation of no-arbitrage pricing in mathematical finance, of sequential analysis and concentration inequalities in statistics, and of convergence arguments throughout probability, and it is the natural setting for defining stochastic integrals against Brownian motion and semimartingales.
History
The term martingale entered probability through Ville's 1939 work on collectives, and Doob developed the systematic theory of martingales, stopping times, and convergence in the 1940s and 1950s, culminating in his 1953 treatise that made martingales a central tool of modern probability.
Key figures
- Joseph Doob
- Paul Levy
- Jean Ville
Related topics
Seminal works
- doob1953
- williams1991
Frequently asked questions
- What is a martingale in plain terms?
- It is a model of a fair game: given everything that has happened so far, your expected next position equals your current position, so on average you neither gain nor lose.
- Why are martingales so important in probability?
- Their convergence and stopping theorems give clean tools for almost-sure limits and expectations, and they are the foundation of stochastic calculus and of arbitrage-free pricing in finance.