We re-graded the MLB board before the Wild Card: 59% farming-flagged, a clean cohort still underwater, and 1,429 pre-game calls that hit 21 games fewer than their prices implied

The Wild Card round opens September 29. We re-graded our MLB board eleven days out, and the board that will carry October's copy-trading money looks like this: 1,663 graded wallets, 58.6% of them farming-flagged against 35.3% site-wide, a clean cohort that is still net negative on $55.6M of turnover, and 15 positive verdicts with $31,745 traded between them, not one observed for 60 days. New since the August read: a per-wallet ledger of every pre-game moneyline position these wallets held at first pitch. Over its first three weeks, the graded population's 1,429 settled calls hit 727 games against 747.7 expected at the prices they paid — twenty-one games short of the market's own forecast.

All wallet figures are a point-in-time pull from CopyGrade's live scored set via the read-only database, dated 2026-09-18; the game-call figures are from the same database's wallet_game_calls ledger, which settles pre-kickoff moneyline positions against each game's final. The method is documented here. Nothing below names a wallet.

Three in five, not two in three

When we planned this post in July, the graded MLB board was 67.8% flagged and the working headline was two in three. We re-derived every figure today, as we do for every data cut, and it did not hold: the board has grown from 419 wallets in July to 1,120 in August to 1,663 now, and the flagged share on today's board is 58.6% (975 wallets; 43.8%, or 729, at the severe level). That is roughly three in five, against roughly one in three across the 10,553 graded wallets site-wide.

MLB is the tenth-highest flag share of our 26 category boards. Nine boards sit above it — cricket (78.7%), the WNBA (78.5%), college football (74.9%), UFC (71.2%), esports (68.2%), tennis (67.9%), IPOs (65.9%), boxing (64.5%) and commodities (61.3%) — so MLB is not the dirtiest anything. It is, though, the fourth-largest board by attributed volume ($372.7M, behind soccer, esports and tennis). And a caveat that belongs in every one of these posts: all 26 boards sit above the site-wide flag rate, the lowest (Formula 1) at 39.7%, because a board requires five trades in one category while the site-wide figure counts every graded wallet. The MLB gap over the site-wide rate is real; a reader who does not know that every board clears it will over-read it.

The causal story is the one we expected and it is honest: 162 games a team means a game almost every day, daily repetition is what the farming detectors read, and the market structure that makes baseball the best vetting data in sports is the same structure that manufactures records. What we cannot say is whether the share is rising or falling. The board grew 48% in five weeks because our discovery widened — 320 of today's 1,663 wallets were first seen in the last 30 days — and farming signals are recomputed from scratch on every sync, so the series has no memory. Levels, dated; never trends.

The clean cohort is still net negative

Five weeks ago the clean cohort was 475 wallets and −$53,721; today it is 688 wallets and −$89,438. The board's entire realized profit still sits with the two flagged cohorts, and the cohort a copier could actually consider — 41% of the wallets, 15% of the volume — finished underwater on the season's tape so far.

The last two columns carry the pattern the NFL board taught us, and they have to be read together: clean wallets are underwater more often than severe ones (91.6% against 79.0%) and by far less (a median −0.50% against −4.90%). Clean wallets lose small and often; flagged wallets win often on paper and lose large in the estimate of what a copier keeps. Quote either half alone and the finding inverts. The board's negative-edge share, 84.4%, is the sixth-highest of the 26, behind Formula 1, golf, the NFL, the NBA and the NHL — the boards whose wallets look best by headline measures keep turning out to be the ones with the least copyable edge.

The fifteen positive verdicts are the same young wallets, five weeks older

In August all fifteen positive verdicts were brand new and collectively down; that was the post's headline. Today the count is fifteen again, the combined MLB turnover is $31,745, the combined MLB result is −$1,431, and the count observed for 60 days is still zero — the median wallet among them was first seen 40 days ago. They are largely the same accounts: eleven of today's fifteen already carried a positive verdict in mid-August, three were not yet graded then, and one was on the board with a different verdict. Meanwhile, of the wallets on today's board that carried a positive verdict in mid-August, eight have since lost it. Ten of the fifteen are profitable on MLB, on a median of 28 MLB trades each. A positive verdict on a wallet with a month of tape is where diligence starts; on baseball, with hundreds of settled games available to a wallet that has been at it all season, it is a thin reason to start.

What the pre-game calls say

This is the section we could not write in August. Since 2026-08-27 the site has kept a per-wallet ledger of game calls: for every graded wallet and every MLB game, the side it was net long on the moneyline at first pitch — pre-kickoff buys minus pre-kickoff sells — settled hit or miss against the final. Nothing after first pitch counts; a wallet with no net position at first pitch made no call. The recent-form guide has every rule; the one that matters here is expected hits: the sum of the prices a wallet paid, which is how many hits the market itself predicted.

Over its first 23 days the ledger recorded 2,838 MLB calls by 278 graded wallets across 287 games; 1,429 of them, by 140 wallets, cleared the $250 stake floor a form is read on. Those 1,429 calls hit 727 times against 747.7 expected at an average price of 52¢. The graded population, taken together, called MLB games about twenty-one hits worse than the market it was trading in.

Three things to read off that table, in order of how much they should move you.

The clean cohort barely exists. One call in three weeks from a farming-clean wallet in the Candidate band — the same structural absence the Graded Slate shows every game day, where the season ledger on 2026-09-18 read clean 0–0, flagged 84–72. The wallets that clear both of a copier's bars do not, as a population, hold MLB moneylines into first pitch. That is a finding about the population, not a gap in the instrument.

The count-based result is the honest one. Hits against expected is immune to position size, and on that measure every cohort with a sample came in at or under its prices: the clean-under-65 cohort dead on (70 against 69.8), the watch cohort twelve short, the severe cohort nine short. Split by price instead of by cohort, the whole population's underdog calls (under 40¢) hit 28.3% against 35.4% expected, its near-coin-flip calls (40–60¢, nearly two-thirds of the ledger) 47.3% against 50.1%, and its favourite calls (60¢ and up) 68.4% against 64.4%. Graded MLB wallets, as a group, were worse than the market on underdogs and slightly better on favourites, which is what a population with no net information over the closing line looks like.

The dollar result is one position. The ledger's −$2.97M net is 79% a single call: one severe-flagged wallet held a pre-game position of about $2.35M on one side, at close to even money, and lost all of it; its only other call also missed. Take that wallet out and the other 139 wallets are down about $445,000 between them on roughly $16M staked, with the severe cohort at about −$201,000 rather than −$2.72M. Sixty-eight of the 140 wallets were net positive over the window; the median call staked $4,515 and one call in ten staked more than $27,400. We publish the dollar figure because the ledger is public and it is what the tape says; we lead with the count because a copier sizes positions to their own capital, and a copier mirroring that wallet at any sane size would have inherited its call, not its loss. We do not describe the position further than that: this site names no wallet, and a fill-by-fill description of a single on-chain position would be a name in all but form.

What to do with the eleven days

The money October will be traded on is already in place: the World Series champion futures event stood at $40.7M of volume across 31 team markets on our tracked events on 2026-09-18, up from $36.8M when we first sized it in late July, and the bracket's series and single-game markets will be layered on top of it from the 29th. Three moves before then:

1. Build the shortlist from sample, not verdict pills. Fifteen month-old positive verdicts with $31,745 between them are not a shortlist. Use the board sorted by volume to find wallets with a season of MLB tape, then run the full vetting checklist and demand weeks of observation.
2. Read every candidate's recent form against its expected hits. The form card on each verdict — here is how to read it — lists the wallet's last fourteen days of calls with the price paid on each. A 9–3 built on 80¢ favourites is the market; a 9–3 built on coin-flips is the beginning of evidence. On today's read, no MLB wallet that also passed the farming check and the Candidate band was hot.
3. Backtest before you commit, and set exits before the bracket. The Copy Simulator replays a candidate's real fills under your capital, fees, slippage and latency — an honest backtest, step by step — and October compresses samples and sharpens prices, so write the stop rules now, while nothing is on fire. The postseason guide covers the short-series and bracket-exposure traps in detail.

Caveats

These figures cover CopyGrade's covered universe, not all of Polymarket, and are a dated snapshot — the board recomputes with every sync and the call ledger settles twice a day, so both will have moved by the time you read this. Board membership is at least five trades on MLB-tagged markets; realized profit and volume are the board's own, computed over each wallet's tracked window from public trade history against a published methodology, and understate wallets holding unresolved positions. "Real edge" is our post-fee, post-latency estimate across a wallet's whole record, not its MLB record alone. Game calls are pre-kickoff net positions on the moneyline only; in-play trading, spreads, totals and futures are outside the ledger, and a wallet's call is scored on the side it held at first pitch even if it traded out of it later. The ledger's dollar figures are the tape's, and one position dominates them, as stated. Board attribution is per category, so wallets active elsewhere appear on other boards too. 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.

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.