None of these rankings would exist, if I wasn't convinced the models are as accurate as we can reasonably expect.

I have previously spent a lot of time explaining 2 things that seem contradictory:

• All rankings are “bad”-- strictly speaking, they will never be accurate or trustworthy.

• My models are less bad than the next guy.  I know this because I measure accuracy.

Most followers get confidence from knowing my methodology. Knowing how the underlying data is treated . 

But my accuracy reports have given validation. Measuring the results doesn’t only give me feedback to improve— It’s also the only way I know it's not all a huge waste of time! And I hate wasting time.

For more information about accuracy, click on the "Accuracy" tag in Articles. 

/Subvertadown