Skip to content
●Polyrank
Re-ranked daily on real-time data

How it works

How we score a trader

Profit rewards luck and size. We score whether the price a wallet paid was right, in three steps.

Chapter 1

The luck machine

Give 1,000 people a fair coin and 100 bets at 50¢. Nobody has an edge. A leaderboard sorted by profit will still crown ten geniuses.

Fig. 2 · simulated

Final profit of 1,000 coin-flippers

−$40−$20$0+$20+$40the “top 10”
+$22 or better ↓
471 made money, 455 lost it, 74 broke even — on pure chance. The best of them is +$28 with a 64 % win rate.

Round 1

The profit leaderboard

Top 10 coin-flippers by profit after 100 bets
#TraderProfitWin rate
1Flipper #0244+$2864 %
=2Flipper #0460+$2663 %
=2Flipper #0758+$2663 %
=4Flipper #0282+$2462 %
=4Flipper #0379+$2462 %
=4Flipper #0548+$2462 %
=4Flipper #0549+$2462 %
=4Flipper #0689+$2462 %
=9Flipper #0394+$2261 %
=9Flipper #0521+$2261 %

Rerun every coin and see who stays on top.

A leaderboard sorted by profit is mostly a leaderboard of coin flips. Polyrank asks a different question: would they win again?

Measured on real Polymarket traders

Skill persists. Of the top 1% of traders by luck-adjusted edge in Q2 2026 who kept trading, 81% were still in the top 10% the next quarter. Chance would be 10%.

n = 741 of 74,005 traders; windows Apr–Jun → Jul–29 Sep 2026; replication Q1 → Q2 2026: 72.1 % (n 833). Counting everyone who stopped trading as a miss gives 38.9 % (about 14× the ~3 % chance under that counting); ranked per position it is 44 %. Method and caveats ↗

Chapter 2

Two ways to be right

A trader can read the price right and the world wrong. Polyrank scores the two separately. Here is one real market, and one example trade in it.

Will Trump nominate Kevin Hassett as the next Fed chair? Yes price, ¢

0¢25¢50¢75¢100¢OctNovDecJanFebMarBought Yes · 42.5¢Sold · 55.5¢ (+13¢)Held to the end: −42.5¢ a sharePeak 84.5¢Resolved No · Yes paid 0¢

As a trader

Did the trades make money?

✓ Good trade
Closed positions in profit
1 of 1
Profit on 100 shares
+$13.00 on $42.50
Return
+30.6 %

As a predictor

Did they back what happened?

✗ Wrong call
Resolved calls right
0 of 1
Paid vs what happened
42.5¢ for a side that paid 0¢
Edge vs odds
−42.5 pp
Source: Polymarket order book, daily Yes price, 1 Oct 2025 – 4 Mar 2026; resolved No. The buyer is an example on real prices, not a real wallet.
  1. 1 · The market

    Through the autumn, traders priced a Hassett nomination between 23¢ and 47.5¢.

  2. 2 · The trade

    An example buyer takes Yes at 42.5¢ on 20 Nov and sells nine days later at 55.5¢: +13¢ a share.

  3. 3 · The resolution

    The price peaked at 84.5¢. Then the market resolved No: the Yes they backed paid 0¢.

  4. 4 · Two report cards

    Good trade, wrong call. A P&L leaderboard sees only the first card. Polyrank scores both, and tells you which one a wallet is good at.

Chapter 3

Price paid vs what happened

If you keep paying 40¢, you should win about 4 times in 10. Win more often than the prices you paid, and you are seeing something the market isn’t. That gap is what we score.

Fig. 4 · live

A top-ranked wallet (Du…nt), 4,834 resolved positions

Each jar is a price band. Dashed line: how often that price says you should win. Fill: how often the side this wallet bought actually won. Faded jars have fewer than 30 resolved positions. Overall it was right on 99.2 % of resolved positions after paying 93.3¢ on average: an edge of +5.9 points.

Move the price

Paid about
54¢
The price says win
54 in 100
It actually won
97 in 100

+43 points better than the price, over 34 positions.

Or test an idea in the Strategy Lab