Probability in Practice

Most decisions arrive wrapped in words, not numbers: a doctor calls a result "concerning", an advisor calls a fund "pretty safe". Those words hide huge disagreement, because one person's "likely" is 55 percent and another's is 90 percent, and the gap quietly steers choices about health, money, and time. This capstone pulls the module together into one habit: put rough numbers on your uncertainty, anchor them to how common the thing actually is, weigh payoffs against odds, and stay honest about how wide your range really is. None of it needs heavy formulas, only the discipline to be specific where instinct prefers to stay vague.

Put a number on it

Words of estimation feel precise but are not. Force yourself to attach a percentage, even a rough one: "likely" becomes "about 70 percent", "rare" becomes "maybe 1 in 100". The number may be wrong, but now it is visible and open to revision. Vague words cannot be checked against reality; numbers can.

Start from the base rate

Before weighing the specific evidence, ask how common the thing is at all. A striking symptom or an alarming headline is the new evidence; the base rate is the background. Update from the base rate toward the evidence, rather than throwing it away because the new signal feels dramatic.

Expected value, with a floor

Expected value multiplies each outcome by its probability and adds them up. It is the right lens for repeated, survivable decisions, but it says nothing about ruin. A bet with positive expected value is still foolish if one bad draw wipes you out, because you never get to play the average. Chase expected value, but cap the downside first.

Forecasts are ranges

A single-number forecast fakes a precision nobody has. "Sales will be 400" is fragile; "between 300 and 550, most likely near 400" is honest and more useful. Well-calibrated thinkers are right about how often they are right: their 70 percent claims come true roughly 70 percent of the time.

A screening test flags a condition affecting about 1 in 100 people. It catches 90 percent of real cases but also returns a false positive for 9 percent of healthy people. Your result is positive. Imagine 1,000 people: about 10 have the condition and 9 test positive; of the 990 healthy people, roughly 89 also test positive. So about 98 test positive, yet only 9 are truly ill. Your chance of illness is around 9 in 98, under 10 percent. The alarming word "positive" hides a base rate that dominates the answer, so the next step is a follow-up test, not panic.

Expected value misleads when downside is ruinous. Suppose a wager pays 60 percent of your savings on heads and costs 60 percent on tails, and you must stake everything each round. A single round looks flat to slightly positive, so it seems worth repeating. But across many rounds a run of tails devastates you, and the typical outcome drifts toward zero even though the average looks fine. Positive expected value did not protect you, because you live in one world and cannot recover from ruin.

Psychologist Philip Tetlock spent decades studying forecasters. His earlier work found that confident experts making vague predictions were barely better than chance. His later Good Judgment Project, which won a US government forecasting tournament in the early 2010s, showed the other side: ordinary volunteers who expressed forecasts as explicit probabilities, started from base rates, updated in small steps as news arrived, and tracked their own accuracy became strikingly good. These "superforecasters" out-predicted many specialists. The documented lesson is well supported: numerical, evidence-updated, calibrated thinking beats confident storytelling.

• "A positive test means I probably have it." With a rare condition, most positives are false. The base rate rules.
• "Positive expected value means take the bet." Not if one loss can ruin you. Survivability comes first.
• "A confident forecaster is a good one." Calibration, not confidence, is the mark of skill.
• "Putting numbers on guesses is fake precision." A rough number you can check beats a vague word you cannot.

Work through everyday scenarios, converting each vague claim into a rough probability, adjusting for the base rate, and deciding whether the expected value justifies the risk.

Think Like a Maester: The headline number is almost never the answer. Anchor to how common the thing is, update in small honest steps, and never bet the farm on a good average.

1. Why is it useful to convert a word like "likely" into a rough percentage?
2. In the screening example, why is the chance of illness under 10 percent despite a positive result?
3. What does the coin-wager counterexample show about relying on expected value alone?
4. What does it mean for a forecaster to be well-calibrated?
5. What common thread links base rates, expected value, and calibration?

Good judgment under uncertainty is a routine, not a talent. Turn vague words into rough numbers so your beliefs can be checked, anchor them to base rates before dramatic evidence sways you, pursue expected value only after capping any ruinous outcome, and state forecasts as ranges you can be held to. Tetlock's research shows this style of thinking reliably beats confident guessing, turning probability into a working tool for real decisions.

• Statistics in Everyday Decisions
• Statistics for Everyday Life