User: mugurel

Re:Distributions and correlations.

On: The Perils of Simplifying Risk To a Single Number

That makes sense. Other than that, I think they are just facing the bias/variance trade-off http://en.wikipedia.org/wiki/Overfitting_(machine_learning) . If you want your model to be better at predicting particular unlikely events, it will come at the cost of an overall loss of prediction accuracy.

before making fun of hard disks

On: Why Not To Shout At Your Disk Array

go and measure your own performance degradation while your hard disk does something mean to you