Loading
Frozen edition · computed 2026-06-09
This page is the frozen 2026-06-09 edition of the Luck Share, computed under the pre-2026-07-15 method from Polymarket's public API recent-trade window, which omits redeemed positions. The chain edition below reads the same cohort from the complete on-chain Polygon settlement record. The figures in this section are the 2026-06-09 reading as published.
The Luck Share · latest reading 2026-06-09
How these wallets were selected
This board is a closed set of wallet addresses selected once on 2026-04-19, not a screen over Polymarket. No wallet has entered since. The original screen was: Polymarket wallets with >=5 resolved positions and >=$5k realized PnL (2026-04-15 snapshot). Ranked by Edge Score V3b. Membership was conditioned partly on having made money, which puts profit on the causal path into the pool, so no count taken over these rows describes Polymarket wallets generally.
Share of the testable cohort whose realized record cannot be separated from chance at our frozen, FDR-corrected bar. 0 of 35 testable wallets clear the bar; 15 of 50 had too few resolved positions to test. Of the 35, 19 read as not separable from chance, 15 are too concentrated to read as a clean edge, and 1 had a single positive interval that does not survive the correction.
This is a count over a closed set of 50 addresses selected once on 2026-04-19 from a 2026-04-15 snapshot and never added to, not an estimate of how often Polymarket wallets clear. The count over wallets is a census. The read inside each wallet is not: every wallet carries a 95 percent interval over its own resolved positions. The bar itself is membership in a frozen registry built by Benjamini-Hochberg control over 3,871 wallets at q = 0.10; 29 of these 50 addresses were inside that family and 21 were never tested in it, so for those 21 not cleared means not screened rather than screened and failed.
The zero does not rest on the concentration screen or on an independence assumption: dropping the concentration screen and correcting across all 35 tests still clears none, and neither does a correction that stays valid when wallets trade the same events. Membership required realized profit, so this is not a set assembled to make wallets look bad. Every address in it had made money by the snapshot that selected it, and the record tested includes the period that money was made in.
Read basis: the 2026-06-09 scan was computed from Polymarket's recent-trade window, which omits redeemed positions and is therefore not a complete resolved record. The chain edition below reads the same cohort from the complete on-chain Polygon settlement record; the two editions are different instruments, shown side by side.
Diagnostics on public on-chain records, not investment advice; a past read is not a forecast.
Method note (2026-07-16): this analysis was computed under the pre-2026-07-15 machine, where high single-event concentration demoted a read to flagged; concentration is now a descriptive style label and the sufficiency function moved to an independent-event floor.
Quote it verbatim. The number and how the set was chosen travel together; the method is frozen and the per-wallet table is public.
On the 50 Polymarket addresses Convexly screened once in April 2026 for having at least 5 resolved positions and at least $5,000 realized PnL (2026-06-09 reading), 0 of the 35 testable wallets clear a frozen, FDR-corrected skill bar; 15 of 50 had too few resolved positions to test. This is a count over a closed set of 50 addresses selected once on 2026-04-19 from a 2026-04-15 snapshot and never added to, not an estimate of how often Polymarket wallets clear.
Convexly (2026). The Luck Share: share of a closed, profit-screened set of 50 Polymarket addresses not separable from chance at a frozen, FDR-corrected bar. Reading as of 2026-06-09. https://www.convexly.app/research/luck-share
On the 50 Polymarket addresses Convexly screened once in April 2026 for having made money, chain edition reading of 2026-08-01, 0 of 32 testable wallets clear a frozen, FDR-corrected skill bar on their complete on-chain records; 18 of 50 were refused by the completeness gate. This is a count over a closed set of 50 addresses screened once in April 2026 for having made money and never added to, not an estimate of how often Polymarket wallets clear.
Convexly (2026). The Luck Share, chain edition: share of a closed, profit-screened set of 50 Polymarket addresses not separable from chance at a frozen, FDR-corrected bar, read from the on-chain settlement record. Reading as of 2026-08-01. https://www.convexly.app/research/luck-share
Canonical URL: https://www.convexly.app/research/luck-share
Chain edition · reading as of 2026-08-01
This edition reads each wallet's record in full from the Polygon settlement record instead of the API's recent-trade window. A wallet is testable here only when its record has been read in full, its entry prices come from the current fill decoder, and at least 90 percent of its resolved positions carry an order-book entry price. 32 of the 50 board wallets pass that gate; the other 18 are refused, not scored: 16 because their chain read is not yet complete, 2 because too few of their resolved positions have order-book entry prices. This edition publishes under a rule set in advance: it does not publish until at least 25 of the 50 pass the gate.
Of the 32 testable: 0 clear the corrected bar, 10 carry no concentration flag (5 of those include zero and 5 sit entirely below it), and 22 are too concentrated to read as a clean edge. This is a count over a closed set of 50 addresses selected once on 2026-04-19 from a 2026-04-15 snapshot and never added to, not an estimate of how often Polymarket wallets clear. The count over wallets is a census. The read inside each wallet is not: every wallet carries a 95 percent interval over its own resolved positions, and the false-discovery correction controls errors across these 32 tests only.
What this measures: the share of testable wallets whose complete on-chain record does not clear the frozen corrected bar at these sample sizes.
What it does not prove: that any wallet has no skill, that a record is current (no chain probe runs in this lane; each record is scored as stored at its last full read), or anything about the 18 wallets the gate refused.
Each testable wallet's row shows its chain-read state, resolved-position count, realized edge, and 95 percent interval, next to its 2026-06-09 state. Read the two columns as two different instruments, not as before-and-after: the June reading scored the API window, this one scores the complete record, so a changed state reflects a change in what was read, not evidence the wallet changed how it trades. One example of why this matters: the one wallet the June window read with an uncorrected positive interval has a 95 percent interval entirely below zero on its complete record. A below-zero interval is also what paying the spread on entry produces under calibrated prices, so it says nothing about forecasting.
| Wallet | Chain state | n | Realized edge, points [95 percent BCa] | Frozen 2026-06-09 state (API window) |
|---|---|---|---|---|
| yuo45 | too concentrated to read | 238 | -1.1 [-4.7, +2.5] | too concentrated to read |
| ShouShouKKos | interval entirely below zero | 126 | -12.8 [-20.3, -6.1] | interval clears zero on the positive side (uncorrected) |
| Mongabc123 | not separable from chance | 61 | -3.1 [-10.8, +6.0] | not separable from chance |
| PinkPunks- | interval entirely below zero | 124 | -16.2 [-20.8, -10.2] | not separable from chance |
| bioo | too concentrated to read | 828 priced of 829 resolved |
Sign in to see the per-wallet cohort behind this reading: each wallet, whether it was testable, and its register-safe read. Sign in. The full per-wallet table is also public at the top-50 skill scan.
Most prediction-market trackers rank wallets by raw profit. Profit at these sample sizes does not separate skill from variance. The Luck Share is the one number that does the opposite: it takes the 50 Polymarket wallets we screened once in April 2026 for having made money, and have never added to, and reports how much of that set survives a real skill test. The screen and the test do not measure the same thing: membership was won on dollar profit, while the test asks whether entries beat the prices paid, position by position. Read the zero as a fact about these records, not as a result about Polymarket wallets in general. For each wallet with at least 30 scored resolved positions we compute the realized entry edge, whether entries resolved true more often than the price paid implied, with a bootstrap 95 percent interval and a concentration screen, then check membership in a frozen registry of wallets that cleared Benjamini-Hochberg control at q = 0.10 across 3,871 wallets. Twenty-nine of these 50 were inside that screen; the other 21 were never in it.
What "not separable from chance" means, precisely: the record does not clear the corrected bar at this sample size. It does NOT mean the wallet has no skill. At these resolved-position counts most records cannot distinguish skill from chance in either direction, and a profit screenshot tells you even less. That honest limit is the finding, and it is why we publish it as a recurring number rather than a leaderboard.
| As of | Luck Share | Cleared / testable | Cohort | Note |
|---|---|---|---|---|
| 2026-06-09 | 100% | 0 / 35 | 50 | Of the 35 testable, 19 read as not separable from chance, 15 are too concentrated to read as a clean edge, and 1 had a single positive interval that does not survive the multiple-comparisons correction and is net-negative. 35 one-sided tests at a 2.5 percent threshold work out to 0.875 positive intervals if no wallet had any edge and the 35 had been drawn at random; these 35 were not drawn at random, so read 0.875 as a floor, not a forecast. |
The method is frozen and version-controlled, and the per-wallet table behind this reading is public. Diagnostics on public on-chain records, not investment advice; a past read is not a forecast. The full scan, including every interval and the one uncorrected positive record, is at the top-50 skill scan. How the bar is defined: methodology. The tests we have run that came back null: negative results.
Paste any Polymarket wallet into the free analyzer. It returns the same skill-vs-luck read with its interval, not a profit screenshot.
Check a wallet| +0.8 [-1.3, +2.9] |
| too concentrated to read |
| ligmaaaaa | interval entirely below zero | 81 | -11.7 [-18.5, -6.0] | too concentrated to read |
| nbafan88 | interval entirely below zero | 127 | -18.9 [-23.3, -13.7] | not separable from chance |
| Michelangelo02 | not separable from chance | 82 priced of 83 resolved | +0.3 [-5.5, +5.7] | not separable from chance |
| anon_1d8a377c | too concentrated to read | 92 | +4.1 [-5.4, +13.1] | not separable from chance |
| Zarvantis | too concentrated to read | 90 | -1.8 [-6.0, +1.6] | not separable from chance |
| Drakonis | too concentrated to read | 90 | +4.2 [+1.0, +8.2] | not separable from chance |
| Vaedrix | not separable from chance | 82 | -3.7 [-10.0, +0.5] | not separable from chance |
| beachboy4 | too concentrated to read | 190 | +1.5 [-5.6, +8.1] | not separable from chance |
| Green | too concentrated to read | 43 | +8.1 [+0.9, +16.7] | too concentrated to read |
| Zorin-472 | too concentrated to read | 62 | -1.7 [-7.4, +2.8] | too concentrated to read |
| Evandor-363 | too concentrated to read | 61 | -1.8 [-6.9, +1.4] | too concentrated to read |
| Vaelric | too concentrated to read | 54 | +0.4 [-4.6, +5.8] | too concentrated to read |
| Caelthar | too concentrated to read | 53 | +1.4 [-3.2, +6.7] | not separable from chance |
| lejcles | too concentrated to read | 41 | +0.0 [-7.7, +4.6] | too concentrated to read |
| Myndor | too concentrated to read | 52 | +4.1 [-0.4, +9.1] | not separable from chance |
| ggggggggggggggggggggggggggg | too concentrated to read | 53 | -7.7 [-20.7, +6.1] | too few resolved positions |
| Uryxen | too concentrated to read | 52 | +1.8 [-4.9, +7.4] | too few resolved positions |
| 0x4128Be7113DBca57DD75e7CD2cd508Dd9FEDdC3c-1767821982593 | interval entirely below zero | 32 | -8.8 [-15.9, -1.2] | not separable from chance |
| Xerria | not separable from chance | 46 | -1.0 [-9.4, +6.7] | too few resolved positions |
| Jexley | too concentrated to read | 43 | -0.8 [-7.5, +7.2] | too few resolved positions |
| Velenza | too concentrated to read | 45 | +4.0 [-1.9, +11.3] | too few resolved positions |
| Lomyn | too concentrated to read | 39 | -1.1 [-7.8, +4.1] | too few resolved positions |
| Strenik | too concentrated to read | 41 | +0.2 [-6.6, +6.0] | too few resolved positions |
| Sytherin | not separable from chance | 38 | -0.7 [-7.4, +5.0] | too few resolved positions |
| c66 | too concentrated to read | 37 | -3.6 [-14.4, +5.7] | too concentrated to read |
| PiderPen | too concentrated to read | 33 | -2.5 [-11.1, +7.6] | too concentrated to read |
| Trevio | too concentrated to read | 34 | +4.8 [-2.1, +12.8] | too few resolved positions |