In this paper we optimize spatially coupled protographs for window decoding (WD) and arbitrary rate. Previous studies found that the belief propagation (BP) threshold of spatially coupled code ensembles achieves the maximum a posteriori (MAP) threshold of their underlying block code ensemble. This property requires a large coupling length L and thus the window decoder is considered to reduce latency and complexity of the decoding. To approach the BP threshold fast in the size of the window W, it is well known that the code requires a special structure to avoid degree-1 variable nodes inside the window. We further require additional structure to construct a systematic code with low encoding complexity, which unfortunately forces degree-1 variable nodes inside the window. Thus, we formulate an optimization problem to maximize the WD threshold and solve it by applying a differential evolution (DE) based algorithm. Compared to the regular protographs obtained by edge spreading, our optimized irregular protographs show a significant improvement in terms of WD threshold and finite length performance for small window sizes, hence leads to small decoding latency and complexity. Furthermore, for high rates our codes can compete with the highly optimized LDPC block codes from the WiMAX standard.