We graded the college football wallets: 82% of the 169 graded wallets on Polymarket's CFB games carry a farming-risk flag
Two Saturdays into the college football season, 169 graded Polymarket wallets have traded the cfb-2026 game events, and 138 of them — 81.7% — carry a farming-risk flag, against 36.5% across everything we grade. Those 169 wallets put $4.77M of buy volume through 194 of the 196 college football game events Polymarket has listed since Week 0, and 95% of their fills came from the 113 wallets carrying a severe flag. Exactly one wallet holds a positive copy verdict, and we have observed it for 13 days.
College football runs roughly 120 game events a week against the NFL's 16 — 123 events and $12.5M of volume in the week of August 31 alone, against 14 NFL game events listed for the week of September 7 — and the graded wallets working it are flagged at a higher rate than either sports population we have published a rate for: 54.3% on the NFL board on 2026-08-28 and 70% of the graded World Cup wallets on 2026-07-24. This post is the dated baseline we will measure the season against, the same way the NFL board got one before kickoff.
All wallet figures are a point-in-time pull from CopyGrade's live scored set via the read-only database, dated 2026-09-06. "Real edge" is our post-fee, post-latency estimate of what a copier would have kept across a wallet's whole record — not its college football record alone — per the published methodology. The College Football board and the weekly graded slate refresh from the same data on every sync rotation — scores re-grade on that rotation, and copier outcomes mature twice a day — so treat the numbers here as the early-September snapshot they are.
How big is college football on Polymarket right now?
Big enough to be its own season. Polymarket listed the first cfb-2026 game event on Thursday, August 27 and had 196 by September 6, of which 187 have settled. Lifetime volume across them is $15.9M. The week-by-week shape:
| Week (Mon–Sun, UTC) | Game events | Volume |
|---|---|---|
| Week of Aug 24 (Week 0) | 69 | $3.34M |
| Week of Aug 31 (Week 1) | 123 | $12.48M |
| Week of Sep 7 (listed so far) | 4 | $55k |
Ranked by stored lifetime volume on 2026-09-06, the top three game events were San José State vs. Eastern Michigan ($1.33M), Colorado vs. Georgia Tech ($899k) and North Carolina vs. TCU ($824k) — a reminder that college football volume does not follow the television schedule. Each game event carries 108–156 markets (moneyline, spreads, totals, team totals, quarter and half lines).
For scale, the NFL's nfl-2026 series has 14 regular-season game events listed for the week of September 7 and five for the week after. College football produces roughly eight times the weekly game count, and — as the season-versus-bracket guide argues — resolution count is what turns a record into evidence.
Who is trading it?
Of the 170 wallets with a fill on a college football game event, 169 are in our graded universe (9,393 wallets on 2026-09-06). Split by farming level:
| CFB game wallets (2026-09-06) | Wallets | Fills | Buy volume | Share of fills before kickoff |
|---|---|---|---|---|
| Passes farming check | 31 | 164 | $313k | 57.9% |
| Watch-level flag | 25 | 325 | $618k | 54.8% |
| Severe farming flag | 113 | 10,139 | $3.84M | 35.6% |
| All graded | 169 | 10,628 | $4.77M | — |
Two things stand out. First, the flag rate: 81.7% flagged, 66.9% severe, against 36.5% and 25.8% site-wide. Second, when each cohort trades. The clean cohort makes most of its buys before kickoff — 57.9% of its fills, $198k of its $313k — while the severe cohort trades mostly in play: 64.4% of its 10,139 fills land after the game has started. The largest cohort by volume is not forecasting games; it is trading them live, and live trading is the pattern that generates the repetition, round-tripping and volume our farming detectors read.
One caveat belongs beside every board-versus-site comparison, and it cuts harder here than usual: selecting wallets by activity on one topic selects for flagged wallets. Every board we grade sits above the site-wide flag rate, because concentration on one market family is exactly the repetition the detectors read. An 81.7% flag rate on college football is a finding about a young, in-play-heavy population as much as about the sport. It is also not a superlative — we did not re-rank the full board table for this post, and this population (wallets on game events) is narrower than a tag board's.
Do the clean wallets have an edge?
Mostly not, and the sign-versus-magnitude split runs the opposite way to the NFL board's. Read both columns together:
| Cohort (2026-09-06) | Wallets | Negative real edge | Median real edge | 75th pct. real edge | Median score | Score 60+ | Median days observed |
|---|---|---|---|---|---|---|---|
| Passes farming check | 31 | 74.2% | −28.2% | −0.75% | 56 | 12 | 27 |
| Watch-level flag | 25 | 80.0% | −13.6% | −1.40% | 53 | 8 | 24 |
| Severe farming flag | 113 | 77.0% | −11.4% | −0.50% | 30 | 0 | 34 |
| All | 169 | 76.9% | −15.2% | 30 | 20 |
On the NFL board the clean cohort was underwater more often and by less. Here the clean cohort is underwater slightly less often (74.2% against 77.0% for the severe cohort) but by far more at the median — a −28.2% whole-record real edge, against −11.4%. The reason is age and size. The clean wallets are small (164 fills between 31 of them, a median of 358 trades across their whole records) and young (median 27 days observed), so a few bad weeks dominate the estimate; the severe wallets carry thousands of fills each (median 3,175), which pulls their median toward the fee drag of a churned book rather than the outcome of any one bet. Neither column is a compliment. The fourth column is the one to read: even at the 75th percentile, every cohort is still negative after fees.
What survives the cut:
| Gate (2026-09-06) | Wallets |
|---|---|
| Traded a college football game event | 169 |
| Passes farming check | 31 |
| …and scores 60 or better | 12 |
| …and real edge above 0 | 3 |
| …and real edge above 1% | 2 |
| Positive copy verdict | 1 |
That one positive verdict is farming-clean and has been under observation for 13 days. It holds the verdict because its overall record clears the bar, not because it has proven anything on college football — the same young-wallet pattern we found at the top of the NFL and MLB boards. Of the 169, 76 have been observed for under 30 days and only 31 for 90 days or more. Verdicts overall: 1 positive, 35 caution, 133 negative.
College football and the NFL are the same wallets
93 of the 169 college football game wallets have also traded an nfl-2026 game event — more than half of the college cohort, and about a third of the 265 graded wallets on NFL games. For a copier this matters in both directions. A wallet's Saturday record and its Sunday record are one book, so a college football win streak that sits inside an NFL losing record is not a reason to copy the Saturdays; and a wallet that only surfaces on one of the two boards is telling you something about where its attention actually is. The College Football board and the NFL board rank the same wallets on different markets — check both before either.
What this baseline commits us to
We will re-grade this population as the season runs and publish what changed — the flag share, the clean cohort's size and edge, the positive-verdict count — as dated levels, not trends, because the graded universe is still widening and a moving denominator makes any "farming rose" or "fell" claim unsupportable. The weekly graded slate does the game-by-game version every week: which side each cohort leaned before kickoff, and whether the lean settled right — how to read it.
If you are vetting a college football wallet today, three habits from the numbers above:
- Read the age before the score. Seventy-six of these wallets are under a month old. A score earned in two Saturdays is a rumor.
- Check when it trades. A record built in play — thousands of fills after kickoff — is a market-making or churn record, not a forecasting one, whatever the profit line says. The fill timing on a verdict page shows it.
- Run the full checklist, then the farming veto. The vetting checklist applies unchanged; on this population the veto removes four wallets in five before the checklist starts. The college football guide covers the board-specific mechanics.
Caveats
These figures cover CopyGrade's covered universe, not all of Polymarket, and describe wallets with a fill on a cfb-2026 game event — a narrower population than the tag-derived College Football board, which also counts futures and props. Every number is a dated snapshot; board standings refresh on each sync rotation and scores re-grade with it. Real edge is a whole-record estimate from public trade history against a published methodology, not a college-football-specific one; realized profit on open positions is not counted. A farming-risk flag is an algorithmic risk assessment, not an accusation of conduct. A "positive" verdict is where diligence starts, not a recommendation — our own validation of whether the score predicts copier returns is still null, and the track record page carries the latest read.
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.