Thursday, March 24, 2011

This Bears Repeating

Tyler Cowen over at Marginal Revolution cites a tidbit that is worth repeating. Over and over. Folks should IMHO be pounded over the head with this until the lightbulb goes on:
…what separated those with modest but significant predictive ability from the utterly hopeless was their style of thinking. Experts who had one big idea they were certain would reveal what was to come were handily beaten by those who used diverse information and analytical models, were comfortable with complexity and uncertainty and kept their confidence in check.
Cowen cites that from Tetlock and Gardner's Why Most Predictions Are So Bad, at Forbes. This shouldn't be a revelation to many folks, but apparently it is. It's a multivariate world, whether you we like it or not. Base your predictions on one variable, and you'll usually get it wrong.


  1. Note that "those who used diverse information and analytical models, were comfortable with complexity and uncertainty and kept their confidence in check" are also highly unlikely to make long-term predictions (if any) without hedging the hell out of them, while those "one big idea" people tend to make firm predictions of such things with conviction.

    And once you're on record firmly predicting it, it impacts your batting average. :-)

  2. Someone with a really good model or a really good understanding is capable of seeing that one variable can under certain circumstances have the power to swamp the others, driving the model. And that can be almost the same thing as "one big idea" except for not being an ideologue about it.

    Course, it's in my nature to strongly prefer a well-reasoned prediction that has been heavily hedged to a bold prediction based on little more than strong faith in a tiny handful of things.

    BTW: awhile back you mentioned that the cheapo espresso machines can make a decent brew, but the gasket wears out and you have to fix that. When you've done this, did you fashion the replacement gasket yourself? I have a hard time imagining that I could even buy a replacement gasket for my machine, which was only like $30, and has now begun leaking... .

  3. It's pretty rare for any one factor to have such an overwhelming effect in any but the short term. In the longer run SDIC and the unforseen tend to derail predictions. Smart forecasters watch events and adjust to conditions rather than stand around hoping things will re-align to the original prediction. Remember my example of a few year back of the long-run predictive power of iterative models at 90% and 95% accuracy levels? Uh huh. You maintain model utility by resetting the model when it wanders into error territory, and try to find ways to reduce the error factors.

    What brand espresso machine? Check their website for parts. The only tool I required to replace the cap gasket on my ancient cheap Krups was a phillips head, but on some newer models you may not even need that. Just something to pry the old seal out.

    I'm assuming it's the cap seal/gasket that's going bad on you. For a Krups that can run anywhere from $2 to $8 plus shipping for the gasket.

  4. Or you can just haunt thrift stores for partial machines of the same maker, and get the whole cap assmebly that way. Most people who get those machines really don't use them much. Me and the missus use the hell out of ours (to me breakfast is two mugs of half-strength espresso) and they tend to last several years anyway.