# Why is low variance important in model selection?

Lets say I have two models with the same bias, but one seem to have a much higher variance. I know it is suggested to pick the one with the lower variance, but I'm not quite sure why.

Models with lower variance are clearly better because they are more stable and we often like stability.

But I was wondering if we also prefer low variance because it may indicate that our low bias on the test set is more 'real' and we expect, on average, to get lower bias on an unseen test-set with the low-variance model than the high-variance model.