We re-graded the NFL board after Week 1: realized profit went five-fold, and 64% of it is one wallet first seen eight days ago

Five weeks before kickoff we graded Polymarket's NFL board and promised to grade it again once real games had been played, whatever the answer turned out to be. Week 1 settled on Tuesday. Here is the re-grade.

The short version: the board's realized profit went from $524,005 on August 28 to $2,801,094 today, a five-fold move in nineteen days — and $1,786,670 of it, 64%, is one wallet that we first observed eight days ago, that carries a watch-level farming-risk flag, and that put $2.29M on the Chiefs before Monday night's kickoff at an average of 55¢. The clean cohort — the wallets that pass our farming-risk check — finished Week 1 with less than it started with: its realized take fell from 0.82% of volume five weeks before kickoff to 0.55% now, and the clean wallets that were already graded on August 28 are, between them, sixteen dollars off where they stood then.

All wallet figures are a point-in-time pull from CopyGrade's live scored set via the read-only database, dated 2026-09-16; the Week 1 game figures are from the Graded Slate as of the same day. The method is documented here. Nothing below names a wallet.

The board, three dated readings

Two of those rows moved for reasons that have nothing to do with football, and the earlier posts said so: the graded universe grew as ingestion widened, and the board grew with it. The rows that moved because of football are volume and profit. The sixteen Week 1 game events carry $48.7M of volume between them — the entire pre-season board of open NFL events carried $68.1M on August 5 — and the champion futures market that dominated the pre-season board is now at $55.9M, up from $42.9M five weeks ago.

Where the new money came from

Realized profit on the board rose by $2.28M between August 28 and today. It did not rise evenly.

The watch-level cohort's $2.31M is almost entirely one account. That wallet has been under observation for eight days, scores under the slate's clean floor — though its flag would keep it out of the clean cohort regardless — and in that time has traded $19.6M across every board we grade for +$3.93M realized. On the NFL board it is $7.14M of volume and $1,786,670 of profit: 77% of its cohort's profit and 64% of the board's. Its Week 1 was seven pre-kickoff fills on the Chiefs moneyline totalling $2,291,636 at an average of 55¢ — $2.29M of the $2.8M the entire flagged cohort put on that game. The Chiefs won.

Take that one wallet out and the board reads 1,523 wallets, $49.9M of volume and +$1,014,424 of profit: a 2.0% take, still about double the pre-kickoff level. The remaining watch-level wallets hold +$525k, and the severe cohort's +$395k is concentrated too — its five most profitable wallets hold $367k of it.

Did the clean cohort survive kickoff?

This is the question the August 5 post was written to set up. Then, NFL was the only one of the fifteen sports and esports boards where the clean cohort both finished positive and out-earned the flagged cohort in dollars, and we said plainly that a slow futures board might flatter patient wallets in a way live games would not. The August 28 read recorded the dollar lead already gone on a board four times the size, and could not say whether that was the market or our reach.

Week 1 lets us separate the two, because every wallet's grade is snapshotted daily and today's board can be split by what each wallet was on August 28.

Read the first row against the August 28 post. The clean cohort held +$118,321 then; the wallets that were already graded clean then hold +$118,305 now (the two populations differ by about a dozen wallets that crossed the five-trade board minimum in between). Across sixteen games and $48.7M of game volume, the clean cohort that existed before kickoff made, net, nothing. The cohort's slide from +$118k to +$94k is entirely the 74 clean wallets that arrived with the season and are down $24,155 between them. Meanwhile 31 wallets that arrived with the season carrying a watch-level flag are up $2.02M, and $1.79M of that is the account above.

So the answer to August's question: the clean cohort did not stop passing the check when the games started — it stopped earning. Its take fell from 0.82% five weeks before kickoff (0.78% on August 28) to 0.55% because its volume grew by $1.8M while its profit fell by $24k. The per-wallet edge shape did not move at all: 92.6% of clean wallets have a negative real edge (93.2% on August 28), the median is still −0.40%, the median score is still 51. What did move is the severe cohort's median real edge, from −1.45% to −2.40%: the flagged tail is fatter now than it was, in both directions, which is the shape the detectors exist to flag.

Eleven positive verdicts, $9,170 of NFL volume

Same shape as August 28 (1,295 → 592 → 85 → 15 → 9), same lesson. The 570 clean wallets scoring under 60 net out to −$28,462 on $9.6M of turnover; the clean cohort's profit is again carried by its best-graded 109. All eleven positives also clear the canonical copyable bar (positive verdict, clean farming check, positive real edge, at least 20 trades in the last 90 days), and between them they have traded $9,170 of NFL volume. Ten of the eleven have been observed for under 60 days (median 30; the youngest, five). The wallet that was the board's only positive verdict on August 5 is still here and still positive, at $7,966 of NFL volume and the same +$720 — and it alone accounts for 87% of the eleven positives' combined NFL turnover.

What the Graded Slate saw

The tables above are lifetime records. The Graded Slate is the week itself: for each Week 1 game, the buys graded wallets made on the moneyline before kickoff, split into a clean cohort (passes the check and scores 65 or higher, as graded at that game's kickoff) and a flagged cohort, with a call counted only when at least three wallets, $250 and a 60% majority sat behind a lean.

• 109 graded wallets bought $3.7M before kickoff across the 34 Week 1 game and prop events.
• The clean cohort bought $30. Two games, one wallet each: $20 on the Patriots, $10 on the Commanders. No call on any game, so its Week 1 record is blank — as its pre-season profile predicted, since only 13 wallets on the whole NFL board are clean and score 65 or higher, and they have traded $9,378 of NFL volume between them, lifetime.
• The flagged cohort had a call on 15 of 16 games and went 7–8. Its largest call was the $2.8M on the Chiefs at an average 55¢, a hit; its largest misses were $162k on the Colts at 40¢ and $159k on the Texans at 48¢. Its one no-call was Cowboys–Giants, where $150k across 24 wallets leaned Giants at just under 60%, a hair under the majority bar.

The slate's clean floor is 65, the score at which a wallet enters our Candidate band, and we decided to revisit it after Week 1 rather than before. So here is the counterfactual, computed against wallets' grades today rather than at kickoff: at a floor of 55 the clean cohort would have had six calls in Week 1 and gone 2–4 (hits on the Eagles and the Giants; misses on the Rams, the Texans, the Chargers and the Broncos), on $108k of pre-kickoff buys from 19 wallets. With no floor at all — every wallet that passes the check — it would have had nine calls. Whether the floor moves is a decision we have not made; if it does, the slate guide will say so, and settled weeks will not be rewritten.

What to do with this

1. Read the age before the score. The wallet carrying 64% of the board's profit has been visible for eight days. A four-day-old wallet can score 86; an eight-day-old one can be up $3.9M. Neither is a track record.
2. Treat the clean cohort's Week 1 as the base rate, not the exception. Across sixteen games the wallets that pass the check made nothing net and bought $30 at the Candidate band. That is what a clean NFL record has looked like on every reading since August: a veto passed, not an edge.
3. Read the season ledger, not the week. Week 2 is already open on the slate; the flagged cohort's 7–8 is fifteen coin flips. We will publish these tables again when the ledger has enough weeks to mean something.

Caveats

These figures cover CopyGrade's covered universe, not all of Polymarket, and are a dated snapshot — the board recomputes with every sync and will have moved by the time you read this. Realized profit is computed from public trade history against a published methodology and understates wallets holding unresolved positions; on a board that is still mostly futures, that is a real limitation. The August 28 cohort split reads each wallet's grade from our daily snapshots and is exact for the wallets we had graded by then; it can say nothing about a wallet's activity before we first observed it. The floor counterfactual uses today's grades, not grades at kickoff, so it is an estimate the live slate would not print. Board attribution is per category, so wallets active elsewhere appear on other boards too. A farming-risk flag — watch-level or severe — is an algorithmic risk assessment, not an accusation of conduct, and nothing here asserts anything about any wallet's intent. A "positive" verdict is where diligence starts, not a recommendation — our own validation of whether the score predicts copier returns is still null.

CopyGrade is analysis-only — it never executes trades, holds funds, or custodies keys, and a CopyGrade Score is a documented research opinion, not financial advice.