Start with how often it usually happens

Asked whether a particular thing will happen, almost everyone starts with the particulars. Who is involved, what they have said, what the latest reporting suggests, how it feels. That is natural, and it is backwards.

The better first question is duller: of all the situations that look like this one, how often did they turn out this way? That number is the base rate. It is where a forecast should start, and the specifics of the case are what you use to move away from it.

Inside and outside

The distinction has names. Daniel Kahneman and Dan Lovallo described it in 1993, in a paper in Management Science called “Timid choices and bold forecasts”. The inside view builds a forecast from the details of the case in front of you. The outside view ignores those details at first and asks how a whole class of similar cases tended to go.

Their observation was that people default heavily to the inside view, and that the inside view is reliably overconfident. A story built from specifics feels complete. It does not include the long list of things that usually go wrong, because those have not happened yet and so do not feature in the story.

Philip Tetlock and Dan Gardner put the same idea at the centre of practical forecasting in Superforecasting (2015), drawing on Tetlock’s forecasting tournaments. Their recommended order is outside view first, inside view second. Start from the rate, then adjust for what makes this case different.

Why it matters on a priced contract

On an event contract the market has already given you a number. A contract at 35 cents is the market’s estimate of a roughly 35 percent chance. That makes the order of operations more important, not less, because the price is the most persuasive anchor available and it is sitting on the screen.

A hit rate on its own tells you nothing argued for writing your probability down before you look at the price. The base rate is how you produce that number honestly. Without it, “my probability” is usually the inside-view story with a percentage attached, and the percentage drifts toward whatever the tile says.

Choosing the reference class is the hard part

“How often do things like this happen” depends entirely on what you count as “like this”, and there is rarely one right answer.

Take a question of the form will this bill be signed into law by a stated date. Possible reference classes include all bills introduced, bills that have passed one chamber, bills of this type in this kind of session, and bills that reached this stage with this many days left. Each gives a different rate, and the gap between them can be large.

A few rules of thumb help:

• Match the stage, not the topic. A question about something already halfway through a process belongs with others at the same point. How far a thing has got is often more predictive than what it is about.
• Match the clock. If the contract has a deadline, the class is not “does this ever happen” but “does this happen within this much time”. Deadlines pull the rate down, often a lot.
• Try more than one. If two sensible classes give similar rates, you have a firm anchor. If they disagree, that disagreement is your honest uncertainty, and a confident number would be pretending otherwise.
• Prefer a rough rate to none. A base rate counted by hand from a short list is still better than no anchor at all.

Then adjust, and keep it modest

Once you have an anchor, the specifics earn their place. They move you up or down from the rate. The discipline is in the size of the move.

Here is an illustration. The numbers are invented to show the method, not drawn from any real market. Suppose a reasonable reference class says events of this kind happen about 30 percent of the time. You know something specific that makes this one more likely. The inside view says “almost certainly”. The outside-view habit says: start at 30, ask how much this kind of detail has historically moved outcomes, and move perhaps to 40 or 45. If the market is at 35, you may have a modest view. If your adjusted number only differs from the price by a point or two, you do not have a trade, you have agreement.

That last case is the common one, and it is worth taking seriously. Most of the time a careful base-rate exercise lands near the market price, because other participants can count too.

Why this suits an evaluation

The structure of a Your Call evaluation rewards exactly the kind of forecasting the outside view produces. Maximum profit per event is capped at 0.5% of account size, and the Equity Growth Target is 10%, so even the theoretical best case needs twenty capped winning events in thirty days. At a 70% hit rate it is roughly fifty. At 60%, roughly a hundred. At 55%, roughly two hundred. The rulebook sets out the full table.

That is a volume problem. It is won by many modest, well-calibrated views rather than a few dramatic ones, and dramatic views are usually inside-view views. A process that starts from the base rate on every question is slower per question and far more consistent across a hundred of them.

It also tells you when to pass. If you cannot name a reasonable reference class for a question, you are missing your anchor, and the useful response is usually to leave that market alone and look for one where you have it.

Evaluations are conducted on simulated accounts. Not financial advice.