Martingale (probability theory)

In probability theory, a martingale is a special kind of stochastic process. It describes a situation where the best guess for the next value, based on everything that has happened so far, is simply the current value. This means that if you know all the past results, the expected future result is just today’s result—no more, no less.

Martingales are often used to model fair games. For example, in a fair casino game where you either win or lose equal amounts, your average expected winnings after each round stay the same as now, no matter what has happened before. This helps mathematicians and scientists understand how random events balance out over time.

The idea of a martingale helps us study many real-world situations, from financial markets to games of chance. It gives a clear way to think about what “fairness” means when things are unpredictable. By studying martingales, we can make better predictions and understand how chance influences outcomes over long periods.

For those interested in gambling strategies, note that the martingale betting strategy is a different concept, used in betting systems and not related to the mathematical martingale in probability theory. See martingale (betting system) for more information.

History

Originally, the word "martingale" described a popular betting strategy in 18th-century France. In simple games, like guessing if a coin will land on heads or tails, some people would double their bet after every loss. They believed this would help them win back all their losses and more, if they kept playing long enough. However, this strategy often led to big losses because bets grow very quickly.

Later, mathematicians like Paul Lévy began studying martingales in probability theory. The term was officially used by Ville in 1939. This work helped show that no betting strategy can truly guarantee success in games of chance.

Definitions

A martingale is a special kind of process in probability theory. It describes a sequence of events where the best guess for the next event, based on all past events, is simply the current event. This means that if you know everything that has happened up to now, the most likely future value is the same as the present value.

Martingales are often used to model fair games, like a coin flip game where your expected winnings after each flip are the same as what you have now, no matter what has happened before. This idea helps mathematicians understand random processes and make predictions about them.

Main article: Martingale

Examples of martingales

An unbiased random walk is a simple example of a martingale. Imagine flipping a fair coin repeatedly: if you win $1 for heads and lose $1 for tails, your total fortune after each flip follows a martingale pattern. This means that, on average, your expected fortune after the next flip is exactly what it is now, no matter what has happened before.

Other examples include certain gambling strategies and natural processes. For instance, in Pólya's urn, where marbles are drawn and replaced with more of the same color, the proportion of any color remains a martingale. These examples show how martingales appear in both games and nature, illustrating the idea that future averages stay consistent with current values.

Submartingales, supermartingales, and relationship to harmonic functions

Submartingales and supermartingales are two ways to extend the idea of a martingale. In a submartingale, the expected next value is greater than or equal to the current value. This means the process tends to increase over time. In a supermartingale, the expected next value is less than or equal to the current value, so the process tends to decrease over time.

These ideas connect to something called harmonic functions, which are important in studying processes that stay balanced over time. For example, if you have a fair game where the expected future value equals the current value, you have a martingale. If the game favors the player, you get a submartingale, and if it favors the house, you get a supermartingale.

Martingales and stopping times

Main article: Stopping time

A stopping time is like deciding when to stop something based only on what has already happened, not what will happen next. For example, imagine a gambler who decides to leave the table only when they have lost all their money—they can’t decide to leave based on future rolls of the dice!

When we combine stopping times with martingales, something interesting happens. If you “stop” a martingale at a certain point in time, the new process you create still follows the same fair-pattern rules. This idea helps prove important results, like the optional stopping theorem, which tells us that, under certain rules, the expected value of the martingale when you stop remains the same as its starting value.

Martingale problem

The martingale problem is a way to study certain kinds of mathematical processes, called stochastic differential equations, by looking at their expected values. In simple terms, it helps us understand how these processes behave over time by using the idea of a "fair game."

When we talk about a martingale, we mean a process where the best guess for the next value, based on all the information we have so far, is simply the current value. This idea is used in many areas, including finance and physics, to model situations where there is no advantage to waiting or acting — the future looks just like the present from a statistical point of view.