Sample size and confidence: how to read a trader's record

You find a trader on Polymarket with a 500% return across 3 markets and start mirroring their positions. Over the next month they enter 10 more markets and lose on 7. Your account drops 30%.

3 resolved markets could have been 3 coin flips that landed right. The trader with a steady 45% return across 150 markets, the one you scrolled past, compounded gains all year.

Raw returns can draw attention to short winning streaks and open-position gains. Study the realized results, the resolved-market count and the loss history before drawing conclusions about a trader.

Flip a coin 3 times and get heads each time. Nobody would bet their savings on the fourth flip, yet that is how people judge traders on a handful of markets.

At 3 markets, luck and skill look identical. At 30, patterns start to emerge. At 300, the picture is clear. A small sample actively misleads: it produces a story ("this trader never loses") that falls apart when more data arrives. The chart at the top of this guide shows how far a measured win rate can sit from the truth at each sample size.

Baseball learned this decades ago. A batter who goes 3-for-3 in April does not lead the league all season. The same math applies here: the Wilson lower bound on a win rate sits far below the raw rate on a small sample and tightens toward it as resolved markets accumulate.

Think of hiring. One candidate claims a 95% success rate on 4 projects. Another claims 72% across 150. The longer record gives you more evidence to examine. Bayesian shrinkage is one statistical way to account for limited observations.

In a shrinkage estimate, the observed result is combined with a baseline. With little data, the baseline has more influence. As observations accumulate, the estimate can move closer to the measured result. The choice of baseline and model matters.

The illustrations in this guide explain that statistical idea. They do not specify how 0xinsider constructs a trader grade.

More observations generally reduce statistical uncertainty. A measured win rate over a handful of markets can be far from the underlying rate; a longer record usually gives a narrower interval.

The curve illustrates the effect of sample size. It assumes comparable observations, so also check whether markets share one event or whether one payout dominates the record.

A shrinkage estimate combines an observed value with a baseline. With a short record it stays closer to the baseline. With more observations the measured performance has more influence.

This works in both directions: unusually strong and unusually weak short records can both move toward the baseline. It reduces overconfidence in a few outcomes; it cannot make an unsuitable model reliable.

A trader profile gives you several ways to read the record. Realized profit describes closed and settled positions; open-position gains remain exposed to changing prices.

Calibration compares the prices paid with resolved results. Read win rate beside entry prices and loss sizes, since frequent wins can still leave a wallet unprofitable.

The resolved-market count shows how much evidence is available. Drawdown describes losses from an equity peak, and the equity curve shows whether gains came steadily or in a few large moves.

These metrics provide context for the letter. An unavailable field means the displayed record cannot support that measurement.

S is the strongest rating, followed by A and B. C is a middle rating, D is below average, and F is the lowest. The letter summarizes historical performance.

Open the profile before acting on a grade. Two wallets with the same letter can differ in profit, sample size, drawdown, current exposure and the markets they trade.

An Unranked wallet has no grade. It should not be read as an F. The guide at 0xinsider.com/learn/prediction-market-trader-grades-explained covers the scale.

Suppose Alex has a short resolved record with strong profit and a high win rate. Those are observations worth studying, but they leave uncertainty about how the results would hold across more markets.

Check the prices Alex paid, the size of the losses, and whether several positions depended on one event. A few wins can produce strong headline numbers without establishing repeatable performance.

If Alex later builds a longer record with comparable results, you have more evidence to assess. This example illustrates how confidence can change; it does not predict a grade or a promotion.

Use a grade to begin your research, then inspect the underlying record. Give a short history less confidence than a comparable long one, and distinguish banked profit from open-position gains.

A profile describes observed results. It cannot establish that a trader will keep performing well, or that you can copy their positions at the same prices.