What's a Good Brier Score? A Polymarket Trader's Guide.

If you have spent any time trying to find sharp Polymarket traders, you have run into win rate — and if you have read why win rate lies, you already know it mostly measures which prices a wallet likes to buy, not whether its judgment is any good. The metric that actually measures judgment is calibration, and the standard way to score calibration is the Brier score. This post explains what a Brier score is, what counts as "good" on Polymarket specifically, and why the number is worthless until you compare it to the market itself.

What is a Brier score?

A Brier score is the mean squared error of a probability forecast. You make a prediction as a probability — "this event happens with 70% likelihood" — and reality delivers a 1 (it happened) or a 0 (it didn't). The Brier score for that single forecast is the squared difference:

• You said 0.70, the event happened (outcome = 1): error is (0.70 − 1)² = 0.09.
• You said 0.70, the event didn't (outcome = 0): error is (0.70 − 0)² = 0.49.

Average that squared error across every forecast a trader has made, and you get their Brier score. Because it is a squared error on probabilities between 0 and 1, the number always lands in a fixed range:

The key intuition: lower is better, and 0.25 is the line that separates "informative" from "no better than a coin flip." Anything above 0.25 means your confident predictions were, on net, working against you. A Brier score is not a grade out of 100 and it is not a win rate — it rewards being confident and correct, and it punishes being confident and wrong far more than it punishes hedging.

Why calibration beats win rate

Win rate asks a yes/no question: did the position pay out? Calibration asks a much harder one: when this wallet's entry prices implied 70%, did those events actually happen about 70% of the time? That distinction is the whole game on a prediction market.

Here is why the Brier score can't be gamed the way win rate can. On Polymarket, the price you pay is a probability. Buy a share at 95¢ and you are stating a 95% forecast. A favorite-collecting wallet — the kind that posts a spectacular win rate by only ever buying 90¢+ favorites — will win most of its trades, but its Brier score exposes the trick. The market already said 95%; being right 95% of the time proves nothing, because that is exactly what a fair price predicted. To earn a good Brier score, a wallet has to be right more often than its entry prices implied — and that is precisely the thing luck can't fake over a real sample.

How Polyrank benchmarks Brier against the market's own prices

This is the part most calibration write-ups skip, and it is the part that makes the number usable. A Brier score in a vacuum is unanchored — you can't tell whether 0.18 is impressive or just what the market handed you for free.

So we don't score a wallet's calibration in a vacuum. For every resolved position, we compute two Brier scores:

1. The wallet's Brier — using the price the wallet actually paid as its forecast.
2. The market baseline Brier — using the market's own mid-price at the same moment as the forecast.

The difference between them is the signal. A wallet whose Brier score is meaningfully lower than the market baseline was systematically more accurate than the prices it traded against — it saw mispricings the crowd didn't. A wallet whose Brier score merely matches the baseline is riding the market's own information; it has no calibration advantage, however good its raw number looks. And a wallet above the baseline is actively worse-calibrated than the price it paid, which is the on-chain fingerprint of paying up for stories.

Every number in this comparison traces back to a public Polygon transaction: the fill price the wallet paid, the market mid at that block, and the on-chain resolution that settled it. Nothing is self-reported. This is calibration inside Polyrank's 13 calibration-family metrics (one of 9 metric families, 84 skill metrics total across 3.3M wallets and 3.0B on-chain fill events since 2022).

What does a "good" Brier score look like on Polymarket?

The honest answer is that there is no universal cutoff, and anyone who hands you one ("below 0.15 = a genius") is selling you a number without its context. Calibration depends heavily on what a wallet trades. Consider two traders with identical skill:

• A wallet that lives in efficient, heavily-traded markets (major elections, headline sports) is forecasting events the crowd already prices tightly. Even a genuinely sharp trader posts a Brier score close to the market baseline there, because there's little mispricing left to capture.
• A wallet that hunts thin, contested, longer-dated markets — where prices are noisier and the crowd is less sure — has far more room to beat the baseline, and a strong one will.

So "good" is not an absolute Brier value. Good is a Brier score that beats the market baseline, in the kinds of markets that wallet actually trades, over a real sample. A raw 0.20 in a chaotic thin market can represent far more skill than a raw 0.12 in a market that was efficient to begin with. This is exactly why we report calibration relative to baseline rather than as a naked number, and why the leaderboard lets you filter by category so you're comparing like with like.

As a rough orientation only — not a threshold to trade on — a Brier score well under 0.25 means a wallet's confident calls are net-informative rather than coin-flips, and a Brier score that sits clearly below its own market baseline over 100+ resolved markets is the pattern to study. The direction and the comparison matter far more than any single decimal.

What the numbers actually look like (measured 6 November 2026)

Here is why "below 0.25" is close to meaningless on Polymarket. We read every wallet's lifetime Brier score from our own tables on 6 November 2026 (3,268,111 wallets in total, 338,740 of them with at least 100 resolved positions):

Read the two highlighted rows together. Almost every experienced wallet clears the coin-flip line, because the market's own prices do most of the work: buy a share at 80¢ and the Brier score you inherit from the price alone is already good. Benchmark the same wallets against the forecaster that just uses the base rate, and barely one in three come out ahead. The coin-flip line separates nobody; the market baseline separates two-thirds from one-third.

The sample-size effect shows up in the same read. Among the 1,378,318 wallets with only 10 to 99 resolved positions, the median raw Brier is a better-looking 0.113 and 59.9% post a positive skill score — against 32.4% of the 60,745 wallets with 1,000 or more. Small samples are not more skilled; they are where a few lucky longshots have not yet been averaged out. That is the whole case for the gate in the next section, and the reason a glossy Brier score on 30 resolutions should make you more sceptical, not less.

The sample-size gate: 100+ resolved markets

Calibration is a statistical claim, and statistical claims need a sample. This is where most "this wallet is a savant" pitches quietly fall apart. A Brier score over twenty resolved markets is an anecdote — a couple of lucky longshots can drag the number anywhere. The same wallet across a thousand fills but only two underlying markets is one opinion repeated, not a track record.

That is why Polyrank's smart-money screen requires at least 100 resolved markets before a wallet's calibration is treated as a signal rather than noise. Below that gate, a wallet can still be browsed, but its Brier score carries a sample-size caveat and is kept out of the skill-ranked presets. It is the first filter, applied before any of the fancier metrics, precisely because a beautiful Brier score on a tiny sample is the single most common way a lucky trader gets mistaken for a skilled one.

#	Trader	Score	P&L	Win%	Volume
1	Du…te	99.8	+$172.1K	98%	$1.85M
2	mo…gl	99.0	+$30.2K	66%	$1.35M
3	la…an	98.7	+$23.0K	91%	$196.7K
4	il…is	98.6	+$79.9K	96%	$2.05M
5	An…07	98.4	+$1.8K	73%	$11.9K
6	Da…od	98.2	+$55.1K	62%	$2.07M
7	Bi…op	98.1	+$57.7K	94%	$705.1K
8	tr…13	98.0	+$13.8K	69%	$392.1K
9	su…at	98.0	+$713.38	82%	$17.3K
10	Ad…me	98.0	+$20.4K	94%	$80.8K

The Sharp Forecasters preset above pre-applies exactly this logic: wallets whose Brier score beats their market baseline, past the resolved-markets gate — calibration and execution quality, not headline win rate or raw P&L. It's the fastest way to see what a genuinely well-calibrated Polymarket track record looks like on-chain.

How to read a wallet's Brier score in 60 seconds

Every Polyrank profile shows the calibration chart — entry price against how often the side bought actually won — to anyone, no account needed; the Brier score, its baseline and the rest of the calibration family sit in the full skill breakdown that opens with a Pro plan. When you land on a wallet, run this:

1. Find the Brier score, then find the baseline next to it. Is the wallet's number below the market baseline, or just matching it? Below = genuinely more accurate than the market. Matching = riding the crowd.
2. Check the direction against 0.25. Comfortably under 0.25 means confident-and-correct on net. Near or above it means the confident calls are working against them.
3. Check the sample — resolved markets, not trade count. Under ~100 resolved markets, treat any Brier score as provisional.
4. Check what they trade. A great Brier score in one efficient category is less impressive than a solid one earned across noisy, contested markets.
5. Put it next to alpha-vs-mid and drawdown. Calibration is one leg of skill; a wallet that's well-calibrated and gets filled better than the mid and isn't risking ruin is the fingerprint worth studying.

Steps 2 and 4 you can do on the free chart; steps 1, 3 and 5 read the skill breakdown. Nothing on either is self-reported. The Sharp Forecasters preset does steps 1–3 for you across the whole leaderboard.

FAQ

What is a good Brier score on Polymarket?+

There is no universal cutoff. A Brier score below 0.25 means a wallet's confident predictions are net-informative rather than coin-flips, but on 6 November 2026 99.4% of the 338,740 Polymarket wallets with at least 100 resolved positions cleared that line, so it separates almost nobody. The number that matters is whether the wallet's Brier score is lower than the market's own baseline — whether it was more accurate than the prices it traded against — measured over at least 100 resolved markets; only 37.2% of those wallets pass that test. A raw 0.20 in a noisy, contested market can reflect more skill than a raw 0.12 in a market that was already efficient.

What does a Brier score of 0.25 mean?+

0.25 is the score you get by forecasting 50/50 on everything, since 0.5 squared equals 0.25. It is the uninformed-coin-flip baseline. A Brier score below 0.25 means your confident calls are, on net, adding information; a score above 0.25 means they are working against you. On Polymarket the relevant baseline is usually stricter than 0.25, because the market's own prices are already a strong forecast.

Is a lower or higher Brier score better?+

Lower is better. Brier is a squared-error measure, so 0 is a perfect forecaster and 1 is someone who was 100% confident and 100% wrong every time. Unlike win rate, a Brier score rewards being confident and correct and heavily penalizes being confident and wrong, which is why it cannot be manufactured by buying favorites.

Why is Brier score better than win rate for finding skilled traders?+

Because on a prediction market the price you pay is a probability, so win rate mostly measures which prices a wallet buys — a wallet that only buys 95-cent favorites posts a huge win rate while proving nothing. A Brier score, benchmarked against the market's own prices, only improves when a wallet is right more often than its entry prices implied. That is the part luck cannot fake over a real sample.

How many resolved markets do you need before a Brier score is trustworthy?+

Polyrank's smart-money screen requires at least 100 resolved markets before treating a wallet's calibration as a reliable read. Below that, a couple of lucky longshots can swing the number anywhere, so any Brier score on a small sample should be read as provisional — an anecdote, not a track record.

Methodology & disclosures

Brier score here is the mean squared error between a forecast probability and the realized 0/1 outcome, averaged over a wallet's resolved positions. The constants are mathematical: 0 is a perfect forecaster, 0.25 is the uninformed 50/50 baseline (0.5²), and 1 is the worst case. For each wallet we compute both the wallet's own Brier score (using the fill price it paid as the forecast) and a market-baseline Brier score (using the market mid at the same moment), and report calibration relative to that baseline. Every input — fill price, market mid, and on-chain resolution — traces back to a public Polygon transaction; nothing is self-reported. Calibration is one of Polyrank's 13 calibration-family metrics (9 families, 84 skill metrics total), computed across 3.3M wallets and 3.0B on-chain fill events of Polymarket history since 2022. The smart-money screen applies a 100+ resolved-market sample gate before a Brier score is treated as a skill signal.

Polyrank is an independent analytics product and is not affiliated with Polymarket. Everything here is read-only, non-custodial analysis of public blockchain data. None of it is investment advice, and no metric — including calibration — promises anything about future results. We measure skill; we don't sell certainty.

---

Check any wallet's calibration before you trust it. Paste an address into the free wallet lookup — no login, no wallet connection — and see how often the side it bought actually won against the price it paid.

Run a free wallet lookup →