Is every local martingale right càdlàg (i.e. right continuous with left limits)?
At the university, in the definition of martingale we assume martingales to be right càdlàg processes.
We call an $X$ process local martingale if there exists a $\left(\tau_{n}\right)_{n},\tau_{n}\nearrow\infty$ sequence of stopping times where $X^{\tau_{n}}$ stopped process at $\tau_{n}$ is a martingale for all $n$.
Does this mean every local martingale is right càdlàg as well? Implicitly, the question is if we can only localize a right-càdlàg $X$ processes with $\tau_{n}$ stopping times in order to get a right càdlàg process (, which is a martingale in this case)?
Assume that your objects are defined on some probability space $(\Omega,\mathcal{F},\mathbb{P})$, and that for every $n$, $X^{\tau_n}$ is almost surely càdlàg (don't need to say "right càdlàg" the rigth continuity is included in the "càd").
This means that you can find a measurable set $E$ such that $\mathbb{P}(E)=1$ and for all $\omega\in E$,
So, take $\omega\in E$, $t\in \mathbb{R}_+$, and $n$ such that $\tau_n(\omega)> t$ (this is possible because of what precedes).
Thus, on a neighborhood of $t$, $X^{\tau_n}(\omega)=X(\omega)$. Càdlàgness is a local property, $X^{\tau_n(\omega)}(\omega)$ is càdlàg at $t$, hence so do $X(\omega)$. This, show that $X$ is almost surely càdlàg .
But, in full generality, a martingale does not need to be right continuous (but even so, it can be regularized).