I love everything about Wrong Tomorrow, except for the nasty philosophical flaw that makes it worse than useless.

The first time I visited the site, I didn’t notice this, but the founder wrote an admirably honest explanation of what he wanted out of this site:

At the start of March, an economist named Robert Barro gazed deep into historical data and announced in a Wall Street Journal article that we had a 20% chance of falling into economic depression.

I found myself mesmerized by this prediction, a shining diamond of unfalsifiability. In August 2008, Barro had made a similar prediction, citing a 5-10% chance of a severe financial crisis—and this low-probability event had come true! It was an impressive feat of prognostication, especially considering that the opposite outcome would have made Barro equally right.

It is a serious issue to think that probabilities are worthless. If you’re chasing certainty, you have two huge problems:

• You are bound to be wrong, at some point. Unless everything is purely deterministic, and you’re omniscient, you need to round 100% down to 99.99%, to at least concede that things could go differently.

• It’s much easier to adjust a probability than to move from certainty to uncertainty. People who are certain spend time defending their beliefs rather than adjusting them — you’ll fight way harder over “Oil is definitely going to $100 in the next year,” than over a 1% adjustment to your estimated chance that it’ll hit $200.

• The most useful predictions we can make are about the most unusual events, so those will necessarily be couched in low-probability language. Which is a more important prediction? “The US GDP will change by some number between -5% and +5% over the next 12 months,” or “There is a ~33% chance that we will have routine 300% annual growth in GDP within the next century.”? Run the numbers, and choice two — which, remember, is something that is most likely not to happen — is far more significant.

The problem with Wrong Tomorrow is that it encourages artificial precision, promotes dogma over data, and encourages people to focus on the trivial but measurable. For a prediction site meant to solve these problems, that’s troubling.