it appears as though r does not depend on t if Walter Lewin at MIT differentiates it correctly.
The chain rule is required because \frac{d((f(t))^{2})}{dt} = 2 f \frac{d(f(t))}{dt}
Thus, your derivative of 1/2mr^2 is wrong because you forgot the factor of dr/dt but that isn't really relavent...