Skip to content
Spark & Whetstonethe grown-up commons

← The Forecast Table

The Forecast Table

Three short lessons for a family forecasting game: six questions that settle before you next meet, a percentage from everyone sealed in an envelope, and a score for how well your percentages matched what happened.

Spark & Whetstone · sparkwhetstone.org/learn/the-forecast-table · CC BY-NC-SA 4.0. Answers are at the end, so you can try each question first.

Lesson 1. Questions that settle

A good forecasting question will have a clear answer by a known date: "Will it rain on Grandma's birthday?" "Will the local team win on Saturday?" "Will the bean plant be taller than 10 cm by next month?"

Avoid questions nobody can check, or ones someone at the table controls. Six questions is a good evening.

Question. Which is a good forecasting question?

  1. Will next year be good?
  2. Will it rain here next Sunday?
  3. Will I eat my vegetables tonight?

Say it back. In your own words: what makes a forecasting question good?

________________________________________________

________________________________________________

Lesson 2. A percentage, not a yes or no

Instead of yes or no, each person writes how sure they are as a percentage: 80% means "I would expect this to happen eight times out of ten".

Write them on paper, put them in an envelope, seal it and sign across the flap. Nobody changes their number after they hear the others.

Question. What does writing 70% mean?

  1. It will definitely happen.
  2. You would expect it to happen about seven times in ten.
  3. It will happen 70% of the way.

Say it back. Say it back: why write a percentage instead of yes or no?

________________________________________________

________________________________________________

Lesson 3. Scoring how well you knew

When you next meet, open the envelope. For each question, score how far your percentage was from what happened: if it happened, count it as 100%; if not, 0%. The distance from your number is your miss. Smaller is better.

A forecaster who says 90% about everything will miss badly when they are wrong. One who says 60% and is right six times in ten is well calibrated. Over several evenings you learn what your percentages really mean.

Question. You wrote 80% and it happened. You wrote 80% on another and it did not. Which miss is bigger?

  1. The one that happened.
  2. The one that did not happen.
  3. They are the same size.

Say it back. In your own words: what does it mean to be well calibrated?

________________________________________________

________________________________________________

At the table: The envelope

Players: three to eight people, 9 or older. You'll need: paper, pens, an envelope, and a date to meet again.

  1. Agree six questions that will be settled before you next meet.
  2. Everyone writes a percentage for each question on their own slip.
  3. Seal all the slips in one envelope and sign across the flap.
  4. Next time, open it and score each miss. Talk about which questions surprised you.

Answers

  1. Will it rain here next Sunday? — Rain next Sunday has a clear answer by a known date, and nobody at the table controls it.
  2. You would expect it to happen about seven times in ten. — 70% means you expect it about seven times in ten similar cases.
  3. The one that did not happen. — A miss of 20 when it happened; a miss of 80 when it did not. Being sure and wrong costs more.