This page will change often. If an idea developes enough steam. it will earn its own page.
These are geeky names for a simple idea that may have applications to price models. Consider the simplest case of a fixed function f of 2n+1 varialbes and a bi-infinite sequence X_t of iid variables. The stationary process defined by letting Y_t equal f(X_{-n+t}, ..., X_t, ...X_{n+t}) is your basic stripped down B-process with a window size of n. These are familiar in the fancier parts of ergodic theory, and they are nicely discussed in Shield's book.
The new suggestion here is to use these process to mode something that interests us, such as equity returns. These models don't violate the look-ahead principle since we only observe the Y values. The fact the observed Y values do inform us to some extent about the future is what makes this amusing.
One question, can we use estimation of the "f" to say something about the predictability of the Y?
Have these processes ever been used as models?