How to do it
- Write the call and a percentage. "70% the new job is the better one." A number can be checked later; a feeling can't.
- Telling someone else, give the number or a range, not a shrug. Numbers and ranges escaped the credibility penalty; plain doubt didn't.
- Change the number when something new comes in. The more accurate forecasters updated more often.
- Look back once you know how it went. Do your 70s come true about seven times in ten?
Why a number helps
- Not knowingtravels with worry
Across many studies, people who found uncertainty hard to bear reported more worry and anxiety. A snapshot, so it doesn't show which comes first.
- A percentagemakes it precise
In a two-year forecasting tournament, people randomly given probability training made better calls. Forecasts on world events, not your choices.
- Checking backsharpens the next one
The more accurate forecasters updated their numbers more often. Looking back at your own calls is my addition.
- Youwrite it, check later
A number on paper can be checked when you know how it went. That it helps your own calls, or eases the worry, is a guess.
Why it works
Stuck between two options, most of us sit with a vague "I don't know". Intolerance of uncertainty, finding not-knowing hard to bear, goes with more worry and anxiety across many studies. Evidence: decent Those were snapshots, so they don't show which comes first. Evidence: decent
A percentage turns the fog into something you can check. In a two-year forecasting tournament, people given probability training made more accurate forecasts, and the more accurate updated more often. Evidence: decent They were forecasting world events. Whether the habit helps with your own calls, or quiets the worry, nobody has tested. Evidence: worth trying
It's the same move as guess before you look it up, for decisions instead of facts: commit first, check after.
The long readEvery step of the chain, each study with its numbers, and what's still unproven. 3 min read.
What happens in your brain
This chain is behaviour, not brain areas. Each of the first three stops is a different paper, and the last is my extension.
Not knowing travels with worry. A meta-analysis, which pools many studies, found that people who found uncertainty hard to bear reported more symptoms of generalised anxiety, the worry-about-everything kind, and of depression and obsessive-compulsive disorder. The links were of similar size. Evidence: decent It was measured at one point in time: it went with worry, it isn't shown to cause it. Evidence: decent
A percentage makes it precise. In a two-year forecasting tournament, people randomly assigned to probability training made more accurate forecasts than those without it. Evidence: decent A 2025 reanalysis of the same data found that effect shrank or vanished once other factors were controlled. Evidence: decent
Checking back sharpens the next one. Across 743 forecasters, the more accurate ones updated their numbers more often. Evidence: decent That's a link, not proof that updating made them better. Evidence: decent
You write it, check later. Writing a number down and looking back at it is the everyday version. That it helps outside a tournament is my guess. Evidence: worth trying
The proof
- Mellers et al., 2014: a two-year forecasting tournament on world events. People randomly assigned to probability training made more accurate forecasts. Evidence: decent Hauenstein et al., 2025, reanalysed the data and found the training effect shrank or vanished once other factors were controlled. Evidence: decent
- Mellers et al., 2015: 743 forecasters, over 150,000 forecasts. The more accurate ones updated their numbers more often. Correlational. Evidence: decent
- Gaertig and Simmons, 2018: 11 studies, 4,806 people. Advisors who sounded fully confident were rated a bit better than ones who said they weren't sure. Evidence: decent Advisors who gave a range, a percentage or "more likely" weren't penalised against flat certainty. Evidence: decent
- Gentes and Ruscio, 2011: a meta-analysis linking intolerance of uncertainty with generalised anxiety, depression and obsessive-compulsive symptoms, at similar strength for all three. Measured at one point in time. Evidence: decent
How to do it
- Write the call and a percentage: "70% the new job is the better one." A number can be checked later. A feeling can't.
- Telling someone else, give the number or a range, not "no idea". Numbers and ranges weren't marked down; plain "I'm not sure" came out a bit lower. Evidence: decent
- When something new comes in, change the number. The more accurate forecasters updated more often. Evidence: decent
- Look back once you know how it went. That step is my addition. Evidence: worth trying Over time, the check is simple: do your 70s come true about seven times in ten?
Watch out
The forecasting results come from people answering scored questions about world events, not you choosing between two jobs. Evidence: decent The training effect is disputed: a 2025 reanalysis found it shrank or vanished once other factors were controlled. Evidence: decent That a number helps your own calls, or eases the worry, is untested. Evidence: worth trying
The link between not tolerating uncertainty and worry comes from snapshot studies, so it doesn't show which drives which. Evidence: decent And a number isn't a shrug: in the advice studies, advisors who sounded fully sure still came out a bit ahead of ones who said they weren't. Evidence: decent