What calibration means
When you say you are 90% sure, are you right 90% of the time? Most people are not. They are overconfident: their ranges are too narrow and their "90%" answers are wrong far more than 1 time in 10.
The 90% rule
In the next step you will answer ten trivia questions. You will not give an exact answer. Instead you give a range — a low and a high value — that you are 90% sure contains the true number.
A well-calibrated person captures the true value in about 9 of 10 ranges. Capture far fewer and you are overconfident. Capture all ten with absurdly wide ranges and you are dodging the question.
Then you will answer ten true/false statements and rate your confidence. We compare what you claimed to how often you were right.
90% Confidence Intervals
For each question, enter a low and high bound you are 90% sure contains the true answer. Wide enough to be safe, tight enough to be useful.
True or False?
Mark each statement true or false, then pick how confident you are. 50% means a coin flip; 100% means certain.
Your calibration
Answer the questions in the previous steps to see your calibration.
Ranges captured
0 / 0
Target: 90%
True/False accuracy
--
Avg confidence: --
Interval test
--
Confidence test
--
Calibration curve
Each point is one confidence level. On the dashed line, your confidence matches your accuracy. Points below the line mean overconfidence.
Show the answers
90% ranges
True / False
Learning Debrief
What You Just Learned
- A 90% confidence interval should capture the truth about 9 times in 10.
- Most untrained estimators are overconfident — ranges too narrow.
- Calibration is trainable: feedback like this is how you improve.
Applying This to Cyber Risk
When you estimate breach costs, downtime, or records exposed, you give ranges. A calibrated range is the difference between a risk model the board can trust and a number you made up.
Loss ranges feed the curve
The Loss Exceedance Curve you build later turns your low/high loss estimates into a probability curve. Overconfident inputs produce a confidently wrong curve.
Calibrate before you quantify
Hubbard's research shows a few rounds of calibration feedback measurably reduces overconfidence. Run this exercise until you reliably hit 9 of 10.