Show how the expected test error of a given prediction rule decomposes into the sum of the reducible error and the irreducible error.

Suppose the state-of-the-art prediction algorithm has bias 3 and variance 7. As result of extensive research you have been able to reduce the bias to 2, but unfortunately the variance has gone up to 10. Have you managed to reduce the expected test error? please take the irreducible error into account.