How Forecasting Works

Forecasting is not predicting the future with certainty. It is the disciplined skill of attaching a probability to an uncertain outcome, then being scored on how well those probabilities match reality over many predictions.

A good forecaster does not aim to be right every time. They aim to be well calibrated: when they say 80%, the thing happens about 80% of the time. That shift, from "will it happen?" to "how likely is it, and how do I know?", is the entire craft.

Step 1: Start from the base rate

The single biggest mistake in forecasting is ignoring the base rate, the underlying frequency of an event. Before you look at the specifics, ask how often this kind of thing happens in general. An incumbent’s re-election, a startup surviving five years, a favorite winning a match, each has a historical rate that anchors a sensible starting estimate.

Step 2: Update on evidence

Once you have a base rate, adjust it as you gather specific information, but only in proportion to how much that information actually tells you. Strong, relevant evidence should move your number a lot; weak or noisy evidence should move it a little. This is Bayesian thinking in plain language: new evidence nudges a prior belief rather than replacing it.

1. Write down your starting probability from the base rate.
2. List the specific factors that push the odds up or down.
3. Move your estimate for each factor, more for strong evidence, less for weak.
4. Sanity-check the final number: would you accept a bet at those odds?

Step 3: Score yourself honestly

A forecast you never grade teaches you nothing. Serious forecasters use a scoring rule that rewards being both accurate and honest about uncertainty. The most common is the Brier score.

The Brier score is the squared difference between your probability and the outcome (1 for yes, 0 for no), averaged over all your forecasts. Lower is better, and it punishes overconfidence: calling something 99% and being wrong hurts far more than calling it 70%.

Over many forecasts, the Brier score reveals your calibration. If your 70% calls come true 70% of the time and your 90% calls come true 90% of the time, you are calibrated. Most people are not at first; they are overconfident, and scoring is what fixes it.

The habits of accurate forecasters

• Think in probabilities and degrees, not in "it will" or "it won’t".
• Break a hard question into smaller, more answerable parts.
• Seek out views that disagree with yours before locking in a number.
• Update quickly when facts change, without overreacting to noise.
• Keep a written record so you can grade yourself later.

These habits were identified by studying the best amateur forecasters in the world. Read superforecasters explained to see how a research project found them.

How prediction markets fit in

A prediction market is forecasting at the scale of a crowd. Instead of one person estimating a probability, thousands trade against a price, and that price becomes a shared forecast. The scoring is automatic: put money on the wrong side and you lose it.

If the mechanic is new to you, prediction markets explained covers how a price becomes a probability.

How to practice

Forecasting is a trained skill, and the only way to train it is to make real predictions and grade them. Pick questions with clear deadlines, write down a probability, and check back when they resolve.

Clutch turns this into a game: you forecast real news and sports with in-app credits, and the app keeps score for you. It is a low-stakes way to build calibration. See how it works.