Martingale, for practical purposes

In several areas of mathematics, such as computational complexity and cryptography, there is a meaningful and important distinction between "there exists ....." and "it is practical to find ........". This distinction is not currently made within mathematical probability, but here is a setting where it would be interesting to do so.

Consider a stochastic process \(X = (X_t, 0 \le t \le 1)\). A familiar theorem states (in contrapositive form, and ignoring technical conditions)

\(X\) is not a martingale if and only if there exists a stopping time \(T\) such that \(\mathbb{E}X_T \neq \mathbb{E}X_0\).
I would like to define a class of processes, say "martingale for practical purposes" (MPP) by
\(X\) is not a MPP if and only if it is practical to find a stopping time \(T\) such that \(\mathbb{E}X_T \neq \mathbb{E}X_0\).

Problem. Make sense of this definition.

Here is a slightly different approach to the same issue. Requiring that a discrete process \(X_0,\ldots,X_n\) be a martingale is imposing an exponential (in \(n\)) number of constraints. What can we say if we are only allowed to impose a polynomial (in \(n\)) number of constraints? Is there some natural choice of a polynomial (in \(n\)) number of constraints which gives a tractable class of processes? In particular, can we do so via a choice of stopping times \(T_1, \ldots, T_{poly(n)} \) and the class of processes satisfying \(\mathbb{E}X_T = \mathbb{E}X_0\) for those \(T\) ?

This arises from the efficient market hypothesis, which is usually taken to imply short-term prices should follow a martingale, but really implies they should follow some MPP.


Back to Open Problem list