How does scipy.stats.distributions.fit works?

Question

I was trying to approximate some hard distribution with normal one. But scipy.stats.norm.fit gives some awkward results. Approximational result is very far from ori