Transition framework
\[
P(X_{t+1}=s_j \mid X_t=s_i)
\]
The next state depends on the current pattern state, not the full distant past.
\[
\hat{P}(s_j \mid \pi)=\frac{N(\pi \rightarrow s_j)}{\sum_k N(\pi \rightarrow s_k)}
\]
For a sequence \(\pi\), the model estimates the next-state probability from historical transitions.