Superforecasting

by Philip Tetlock and Dan Gardner · 8 ideas

  1. Foxes and Hedgehogs

    A hedgehog sees the world through one big idea, while a fox draws on many small ones and doubts each. Foxes predict far better, though hedgehogs sound surer and get more attention. Trust advisors who often say "however", and borrow from several views yourself.

  2. Make It Checkable

    "There is a real chance it will rain" can never be proven wrong, because it can mean anything from 20 to 80 percent. A forecast you can't check is one you can't learn from. Put a number and a date on every prediction, like "70 percent chance of rain Saturday."

  3. Check Your Calibration

    Being well calibrated means that of all the things you call 70 percent likely, about 70 percent happen. Most people are too sure, so their 90 percent calls come true far less often. Write forecasts down, then count how often each level came true and adjust.

  4. Break It Down

    A question that seems impossible often splits into smaller parts you can estimate. Guessing each part and combining them beats one wild guess at the whole. When a question stumps you, ask what you would need to know to answer it, and estimate those pieces.

  5. Outside View First

    Before looking at the details of a case, ask how things like it usually turn out. To guess if a couple will stay married, start with how many marriages last, then adjust for what you know. Details are vivid and pull you too far, so start from the pattern.

  6. Update in Small Steps

    Good forecasters change their minds often, but usually by a little at a time. Ignoring new facts leaves you stuck, while jumping at every headline makes you swing wildly. When news arrives, ask how much it really changes the odds, and move by that much.

  7. Average Many Guesses

    The average of many separate guesses is usually more accurate than nearly any single one, because individual errors cancel out. This only works if people guess before hearing each other. Ask several people for an estimate on their own, then combine them.

  8. Forecasts Have a Horizon

    Forecasting skill fades the further ahead you look, and beyond a few years even experts do little better than chance. Small causes pile up until the future is too tangled to see. Trust short-range forecasts more than long ones, and revisit long plans often.

More on thinking