S to F, on a Polymarket wallet that 0xinsider tracks. It ranks the wallet’s realized profit against every other tracked wallet, so it describes money the wallet has already taken out of the market. Use min_grade to filter trades, positions, and flows by it.
Terms on this page
The
cohort parameter on Pre-game side observations is unrelated. There it names one of three measurement groups, not the graded wallets.
The bands
score runs from 0 to 100. It is where the wallet’s combined score, described in the next section, ranks among the graded wallets. The API holds score inside its grade’s band, so a wallet never sorts past a band boundary.
When a wallet has no grade,
grade and score are absent from the response rather than null, so read them with a default. A wallet with fewer than 5 traded markets never gets one, and a wallet whose Polymarket volume 0xinsider has not verified is not regraded at all.
A grade is relative, so it can change when other wallets move and this wallet places no trade.
What sets the grade
The combined score is 95% the wallet’s realized P&L percentile and 5% a blend of six skill measures. The skill blend only separates wallets whose realized P&L percentiles are almost identical. The realized P&L part is the average of two separate rankings, each worth half:- Realized P&L over the wallet’s whole history.
- Realized P&L with each day’s earnings weighted by age, so a day’s earnings count for half as much every 90 days.
Every measure is ranked against the other graded wallets, and a thin sample is corrected before that ranking:
- Calibration edge, max drawdown, and equity smoothness are pulled toward the population’s median, less so as the wallet’s resolved-market count grows.
- Sortino is dropped entirely below 50 resolved markets, then clipped to the population’s 2nd and 98th percentiles.
- A measure 0xinsider cannot compute for a wallet is left out, and the remaining weights are rescaled to fill the 5%.
Caps that lower a grade
A cap beats the percentile. Every cap only lowers a grade, never raises one.
A wallet passes the drawdown test when its worst peak-to-trough loss is at most 10% of everything it ever put to work. When 0xinsider has no cost basis to divide by, it falls back to the worst loss as a share of the profit built up before it, and passes at 60% or less.
Floors that raise a grade
A floor lifts a proven winner that the percentile ranks too low. Both floors require a positive calibration edge, positive realized P&L, a passing drawdown test,$50,000 or more volume, and a trade inside the last 30 days.
The caps that read only the wallet’s own record run last: the two idle caps (30 and 120 days), fewer than 10 resolved markets, and realized P&L that is not positive. A floor can never lift a wallet past one of them.
Filter by grade
min_grade is an allowlist of grades at or above the one you name. min_grade=A returns S and A, and min_grade=B returns S, A, and B.
On the two routes that take only
S, A, and B, a min_grade of C, D, or F answers 400 with error.param set to min_grade.
min_grade parameter. It lists S, A, and B wallets only.
What this does not tell you
- What the wallet does next. A grade ranks profit the wallet has already realized.
- Anything about open positions.
pnl.totalincludes them, but the grade readspnl.realized. - Recent form. That is
streak_tier, a separate measure described on Scores. - Why a wallet’s grade moved. A wallet that did not trade can still be regraded when other wallets move around it.