A plain-English guide to the method behind the predictions — and how to use the app.
Most people predict things by gut feeling. They read the news, form an opinion, and say "I think X will happen." The problem is that gut feelings are often wrong — and we rarely keep score.
ForecastIQ uses a method developed by Philip Tetlock, a professor who spent 20 years studying thousands of predictions made by experts and ordinary people. His finding? Most experts are barely better than random chance — but a small group of "superforecasters" could consistently beat the odds.
What made superforecasters different wasn't insider knowledge. It was their process. This app follows that process.
Start by forgetting everything specific about the situation. Instead ask: "How often does this type of thing happen in general?"
Now bring in the specifics. What's actually happening in this case right now? The app searches current news and scores 6 key factors.
The outside view always counts as one full term. Each inside factor is weighted by how important it is.
For political forecasts, the app automatically searches the news and scores these six factors. Each has a weight showing how much it matters. You can customise these weights in the Weights page.
What do polls, surveys, and approval ratings say? High poll numbers = positive signal.
Is the government stable? Is there opposition pressure, protests, or a strong incumbent? Stability is a powerful predictor.
What are foreign governments, international bodies, or external actors doing? Sanctions, diplomacy, or proxy interference all shift probabilities.
GDP growth, inflation, unemployment, and trade pressure. Economic pain often precedes political change.
What are analysts, think tanks, and prediction markets saying? Aggregated expert opinion is a strong signal.
How is the media framing the story? Is the narrative turning positive or negative? Propaganda and disinformation are also factored in.
The app assigns a credibility score to every news source it finds. A Reuters article counts more than a random blog. This prevents low-quality sources from skewing the result.
The point of forecasting is to get better over time. ForecastIQ tracks your accuracy using the Brier Score — the standard measure used by forecasting researchers.
No password needed. Just pick a username on the homepage to start tracking your forecasts and building your accuracy score over time.
Good forecast questions have a definite YES or NO answer by a specific date. Example: "Will X happen before December 2026?" Bad: "Will things get better?"
Political mode automatically searches current news and scores 6 factors. Sports mode lets you manually score 6 factors with sliders.
Choose how many days back to search for news — from 7 days (very current) to 90 days (longer trend). Default is 14 days.
The centre panel streams every step in real-time — searching news, scoring factors, applying weights, calculating the blend. You can see exactly how the number was reached.
The right panel shows your final probability, the outside view base rate, and an evidence trail. Click "View Dataset" to see the academic sources behind the base rate.
Go to Weights to adjust how much each factor matters. If you think polling matters more than media sentiment for a particular type of question, adjust accordingly.
When the event happens (or doesn't), go to History or Profile and mark the outcome YES or NO. Your Brier score updates automatically. Over time, you'll see how well-calibrated you are.
Many people jump straight to "but this situation is different." Start with the base rate first, then adjust.
Say 63%, not "probably." Vague language hides bad thinking. Exact numbers force you to commit.
Run the forecast again with a fresh news window as events evolve. Small updates beat big flip-flops.
Your Brier score only improves if you mark outcomes. Don't forecast and forget — close the loop.
If your question doesn't match a standard category, type your own reference class. The app fires a 9-query search battery across RAND, Brookings, UCDP, SIPRI, Metaculus, and academic databases — no date limit — to estimate a base rate from real data.
Check the Community page to see how others have forecasted the same type of question. Learn from superforecasters.
If you're an expert in economics, weight the economic indicator factor higher. Your domain knowledge should be reflected in your weights.
It takes less than a minute. Pick a question, run the model, see the result.
Start Forecasting →