Track Record

House creed: we publish the misses as loudly as the hits. Every leg below is graded against a box score — wins, losses, voids and the ones we couldn't verify.

Counter audit, Sep 26: 237 hits and 215 misses across 452 decided legs on 53 cards (Sep 16's leg-by-leg recount of 187/172/359, Thursday's 2–5, Friday's 5–1, Saturday's 15–10, Sunday's 11–11, Monday's 7–7, Tuesday's 2–5, Thursday Sep 24's 5–2, Friday Sep 25's 3–2) — the figure above is asserted against that total on every grading run now, and the run fails loudly if they disagree. 12 voids and 2 unverified legs are excluded from the denominator and counted separately below. Last 30 days equals all-time because it does: the leg-level history before August 16 was lost in the site rebuild (documented at the foot of this page), so every decided leg we can show falls inside the window. We would rather say that than print two numbers that imply a longer record than we can prove.

September 25 — 3 of 5 (60%), graded September 26

UAB 24, Navy 20 · Clemson 24, California 10 · Indiana 29, Northwestern 23. The Friday college card went 3–2: both team legs cashed, the Northwestern volume under cashed, and the two Cal passing overs — which the card itself printed as needing the same thing — missed together. HIT — UAB +6.5 (+102): the Blazers won outright, 24–20, after Navy turned it over twice in the fourth quarter. The leg was published on every money layer and every model layer pointing the same way (93% of the market-maker's dollars, a reverse move from −10 to −6.5, three ratings inside the number), and the number closed exactly where we took it — flat closing-line value, a hit on the result and a wash on the process. HIT — Clemson–Cal Under 51.5 (−104): 34 points. Clemson went 24–10 in a game its own third-down conversion rate promised, and the total closed at 52.5 — a full point of positive closing-line value on a total the public's Over tickets had pushed our way.

HIT — Aidan Chiles Under 29.5 passing attempts (−125): 29, by half an attempt. Northwestern trailed 12–0 at the half and 22–7 in the fourth, scored 16 in the final quarter, and still threw only 29 times — Indiana allowed 11 first downs a game coming in and the play-count read held even in the trailing script the card named as its dissent. That is a hit we grade honestly: the margin was one attempt. MISS — Jaron-Keawe Sagapolutele Over 35.5 passing attempts (−107): 29 attempts, 16 completions, 119 yards and two interceptions. Cal ran 57 plays and 27 of them were rushes; Clemson ran 49 times and took the clock with it. The card's dissent was two straight games under the number (34, 33), and it was the right read. MISS — Chase Hendricks Over 54.5 receiving yards (−114): 4 catches for 37. Cal threw for 145 as a team; nobody on the roster reached 45. The leg needed a quarterback who averaged 304 over his previous seven to do something like it, and he threw for 119.

By star tier: ★★★★ 0–2 (Sagapolutele 8.0, Hendricks 7.0 — the card's two highest-scored props were its two misses, again), ★★★ 1–0 (Chiles 6.5). Team legs 2–0. Closing-line value: UAB +6.5 closed +6.5 (flat); the Under closed 52.5 (+1.0, POSITIVE); no prop carries a close — our data platform was dark Friday night and dark again at 2:15 PM Saturday, and the Friday games had rolled off the second opinion's tracker by the time we graded. Two of two team legs recorded; zero of three props. We print that rather than a flat. Agree vs dissent: both team legs were ALIGNED (money and models on our side) and went 2–0; the three props were Confirms/Neutral panel reads and went 1–2. By game state: the card's ledger said the Cal passing legs "need Cal to move the ball" and the Under needed the opposite — the Under won and both Cal overs lost, which is the state conflict resolving one way, not three independent results. PLAY vs WATCH: PLAY tiles 3–2. Northwestern +20.5, printed not published, would have hit (lost by 6). WATCH tiles were not bet and would have gone 1–6: C. Robinson Under 17.5 receiving yards hit (no catches); Preston Over 19.5 receiving yards missed (18), Marsh Over 65.5 missed (52), Richard Over 88.5 rushing yards missed (68), Sagapolutele Over 259.5 passing yards missed (119), Hendricks Over 4.5 receptions missed (4), Moore Over 5.5 receptions missed (3). The NO PLAY tiles: Davidson Under 15.5 attempts would have missed (22), Komolafe Under 12.5 receiving yards would have hit (3), T. Brown Under 22.5 receiving yards would have missed (29) — the declines were right five times in seven on this card.

September 24 — 5 of 7 (71%), graded September 25

Atlanta 35, Green Bay 14 · Lynx 86, Fever 66 · Aces 100, Mercury 82 · Mystics 101, Sky 93. The two Thursday cards went 5–2, and both losses were the team legs — the props went 5–0. Falcons @ Packers: MISS — Packers −4.5 (−107). The card published the models over the money and printed the dissent in full: sharp dollars on the dog at both books, a reverse line move from +6.5 to +4.5, four experts on Atlanta and none on Green Bay. The dissent was the game. Penix's first start in ten months went 18 of 25 for 256 and the Falcons ran for 242. HIT — Drake London Over 69.5 receiving yards (+105): 9 catches on 10 targets for 194 — the "wrong pie" read (his 2025 games with Penix, not his 2026 games with the backups) was the whole leg, and it cleared the number by 124 yards. HIT — Jahan Dotson Under 17.5 receiving yards (−104): 1 catch on 3 targets, 11. HIT — Bijan Robinson Over 119.5 rush + rec yards (−113): 29 carries for 194 and two catches for 19, 213 — the largest rushing night of his career on a leg that needed 120.

WNBA finale: MISS — Fever −4.5 (−104). The handicap said Minnesota had nothing to play for and had sat four starters the entire second half two nights earlier; Minnesota started Collier, Howard, McBride and Williams and won by 20. A coach's rest plan is not a lineup card, and the leg was priced on a plan. The tile board's own WATCH verdicts on Clark and Boston sat in the same script and were not bet. HIT — A'ja Wilson Over 46.55 fantasy score (−115): 34 points, 10 rebounds, 4 assists, a steal and one turnover — 54.0 on the DFS operator's scoring, the one favorite on the slate whose stars had to play. HIT — Kiki Iriafen Over 9.5 rebounds (−101): 15 (7 offensive) in 32 minutes with three fouls — the foul-trouble cap (6.0) was applied in print and the minutes held this time.

By star tier: ★★★★ 3–0 (London 8.0, Robinson 7.5, Wilson 8.0), ★★★ 2–0 (Dotson 6.5, Iriafen 6.0-capped). Team legs 0–2. Closing-line value: not recorded on any of the seven legs — our data platform was dark from Thursday evening until Friday at 4:47 PM ET and the Thursday-night public-book close capture the report promised was not made; the publish read is the only read and we say so rather than print a flat. Agree vs dissent: the one leg published against the money at both books (Packers) MISSED; the Fever leg carried the market-maker's agreement (+34 dollars over tickets) and MISSED anyway; all five props were Confirms/ Neutral reads and went 5–0. PLAY vs WATCH: PLAY tiles 5–2. WATCH tiles were not bet: Love Under 224.5 would have missed (312 on 53 attempts), Golden Under 59.5 would have missed (100), Pitts Under 39.5 would have hit (5), Penix Over 204.5 would have hit (256), Zaccheaus Under 9.5 would have hit (9), Lloyd Under 24.5 would have hit (11), Clark and Boston on the Fever board were moot.

September 22 — 2 of 7 (29%), graded September 23

WNBA Tuesday, five games: 2 of 7 — the number bet and one rebound leg cashed, and four scoring overs did not. Sparks +17.5 HIT (89–79 — a ten-point Aces win against a number the public book closed at 17.5; the market-maker's thin lean to Las Vegas, printed as dissent, was right about the side and wrong about the margin). Harrison Over 7.5 rebounds HIT (13, with 21 points — Toronto's center out, Chicago last in the league against the glass, the read did exactly what it said). MISS: Iriafen Over 9.5 rebounds (5 in 18 minutes — two fouls before the first quarter ended, five by the finish; the minutes never existed), Amoore Over 12.5 PRA (8-0-3 = 11 in 26 minutes — the leg needed a 17-point blowout to pull the bench into the fourth quarter and Washington won by 10), Howard Over 11.5 points (6 in 12 minutes — Minnesota's coach sat four starters for the entire second half of the regular-season finale with the playoff seed already set; a coach's decision is not an injury and it grades a MISS), Burrell Over 14.5 points (11 in 29 minutes — Atkins played, so the 6.0 cap stood and the leg simply fell short on 4-of-11), Young Over 21.5 points (14 on 6-of-16 and 1-of-8 from three — with 11 assists; the usage went to the pass on a night Wilson took 15 shots).

By star tier: ★★★★ 0–3 (Amoore 8.0, Howard 7.0, Young 7.5 — all missed), ★★★ 1–2 (Harrison hit; Iriafen and the capped Burrell missed). Team legs 1–0. Closing-line value: the Sparks leg is flat against our published number and +1 against the public book's 16.5 at publish (POSITIVE vs the market); no prop carries a close — our data platform was dark Tuesday evening after the card shipped and dark again Wednesday afternoon, so the publish read is the only read. Agree vs dissent: the one team leg carried the market-maker's dissent and HIT; all six props were Confirms/Neutral reads and went 1–5. PLAY vs WATCH: PLAY tiles 2–5; WATCH tiles were not bet — Miles (GTD) did not play, Clark went for 27-7-9 on the tile we declined for the matchup rank, Austin's 25-13 cleared the 29.5 PRA we declined on the H2H window.

September 21 — 7 of 14 (50%), 1 void, graded September 22

Monday night, NFL: 1 of 4, one void. Rams −6.5 HIT (28–6 — the card's one side, 54, and the market-maker's dollars were on it at both books by kickoff). Dart Over 31.5 rush MISS — he was hurt on the opening drive (3-of-5 for 20, no carries) and Jameis Winston played the rest of the game; Lesson 5 says an injury exit is a MISS, not a void, and that is how it grades. Skattebo Over 49.5 rush MISS (12 carries, 36 — the Giants trailed 21–0 at half with a backup quarterback and ran 21 times as a team). Nabers Over 64.5 rec MISS (1 catch, 1 yard on 4 targets — Winston went 11-of-27 for 111). Stafford Over 238.5 VOID — Nacua was inactive, the leg printed "VOID if he is inactive" at publish and at the 4:29 PM gameday pass, and it voids even though Stafford threw for 327 and four scores. A rule that only binds when it helps is not a rule.

WNBA, two games: 6 of 10. Dream −1.5 HIT (95–84 on the road — the SHARP-at-both-books side, printed against all three ratings). DAL–PHX Under 173.5 HIT — by half a point (87–86, 173), with the public book closing 174.5. Bonner Over 11.5 PRA HIT (17-6-2 = 25), Bueckers Over 23.5 P+R HIT (30 points, 5 boards = 35), Bueckers Over 8.5 R+A HIT (5 + 7 = 12), Thomas Over 7.5 assists HIT (12 — the card's one Conflicts leg, capped at 6.0 for Copper's GTD tag; Copper started and played 34). MISS: Reese Over 32.5 PRA (18-10-2 = 30 in 34 minutes — the card's highest-scored leg, 8.0, missed again), Gray Over 2.5 assists (1), Gray Over 1.5 threes (0-for-3), Ogunbowale Over 5.5 R+A (1 + 2 = 3 in 31 foul-shortened minutes; 2-of-8 from the floor).

By star tier: ★★★★ 2–4 (Bonner, Bueckers P+R hit; Reese, Gray assists, Ogunbowale, Dart missed), ★★★ 2–3 (Bueckers R+A, Thomas hit; Gray threes, Skattebo, Nabers missed). Team legs 3–0. Closing-line value: 1 of 3 team legs POSITIVE (the Under, +1), 1 marginal NEGATIVE (Dream, 3 cents), 1 flat; no prop carries a close — our data platform was dark Monday afternoon and again Tuesday, so the publish read is the only read and the row says so. Agree vs dissent: the Dream side, published against all three ratings with the money layer for it, HIT; the one Conflicts prop (Thomas) HIT; Confirms props 4–4. PLAY vs WATCH: PLAY tiles 7–7; WATCH tiles (J. Jones, Copper, Adams, Parkinson) were not bet — Copper's 31 points and Adams's 195 yards were both on WATCH tiles.

September 20 — 11 of 22 (50%), graded September 21

The Week 2 Sunday card: 22 legs, 11 cashed — the sides carried it and the props gave it back. Sides and totals 6–3: Vikings +4.5 HIT (won 9–3 outright — the card's top side at 57), Bengals +2.5 HIT (won 20–6 with Burrow active), Broncos −2.5 HIT (20–13), Raiders +6.5 HIT (won 26–14 outright), LV–LAC Under 43.5 HIT (40 points — the market-maker's 97%-dollar Under, ALIGNED), PIT–NE Under 41.5 HIT (23). MISS: Commanders +4.5 (lost 37–20 — both books' dollars were on Dallas and we printed it), Falcons +2.5 (lost 34–3 at home — the lowest-confidence side on the card at 48, and the one both books' dollars argued against), MIA–SF Under 44.5 (48 — an ALIGNED total that lost to a 35-point 49ers half).

Props 5–8. HIT: Shough Over 231.5 (252 — 27-of-34), Higgins Over 55.5 (95), Pollard Over 41.5 (64), Odunze Over 39.5 (43 — four yards of room), Achane Over 59.5 (74). MISS: Michael Wilson Over 37.5 (18) and McBride Over 59.5 (41) — Arizona threw for 95 yards as a team in a 31–7 loss to a backup quarterback; the "sharp Arizona money" script we printed was the side's, not the passing game's. Henry Over 93.5 (68 on 16 carries in a 24–17 loss), Dobbins Over 59.5 (10 carries, 36 — Harvey OUT did not move the volume to him), Taylor Under 85.5 (92 on 24 carries in a 33–30 shootout), Kupp Over 24.5 (2 for 20 — the leg the card itself had flagged WATCH-if-published-today after Darnold and Charbonnet went OUT), Etienne Over 41.5 (8 carries, 25 — the Saints threw 34 times and won), Hurts Under 206.5 (264 in a 24–20 game the Eagles had to throw to win).

By star tier: ★★★★ 3–4 (Pollard, Odunze, Achane hit; Wilson, McBride, Etienne, Hurts missed), ★★★ 2–4 (Shough, Higgins hit; Henry, Dobbins, Taylor, Kupp missed). The three 8.0-scored legs went 1–2 — the card's highest-scored props missed again (Lesson 7, now the seventh grading day of eight). Overs 5–6, Unders 0–2 on props; the two unders were both volume-player unders in games that went over the number. Closing-line value: 9 of 22 POSITIVE (41%), 2 NEGATIVE, 11 flat — the two negatives were both legs where a better number was posted somewhere in Nevada at publish (Falcons +3 at Wynn; MIA–SF 45 at six books). Hurts (+3 yards), Taylor (+6 yards) and Henry (+1 to +2) all closed with value and all lost — the process grade and the pick grade disagreed on three of eight prop misses. Agree vs dissent: the two legs published with printed dissent split (Shough HIT, Kupp MISS); the two Neutral sides split (Raiders HIT, Falcons MISS); Confirms sides 5–2. PLAY vs WATCH: PLAY tiles 5–8; WATCH/NO PLAY tiles were not bet.

The experimental Sunday Top 50 (its own page, not this record): 19 of 50 (38%) on the line. Ranks 1–10 went 2–8, 11–20 6–4, 21–30 5–5, 31–40 3–7, 41–50 3–7 — the confidence ranking did not rank. Ja'Marr Chase Over 74.5 cashed by half a yard (75). The full verdict is on the card.

September 19 — 15 of 25 (60%), graded September 20

The Week 3 Saturday card: 25 legs, 15 cashed. The props carried it — 8 of 11 — and the sides went 7–7 on a day the SEC's road teams did whatever they wanted. Sides, HIT: Louisiana Tech +19.5 (lost by 17), Charlotte +17.5 (lost by 5), UConn −3 (won by 28 — the card's cleanest read: a ground game against a defense that could not stop one), Sacramento State +26.5 (lost by 21 with the backup quarterback the platform's tag had not caught up to), Fresno State −6.5 (won by 16 — the top side at 57 delivered). Sides, MISS: Kansas +5.5 (lost by 7 in London — the RLM tag is now 1-for-10 on our sheet), Texas A&M −16.5 (lost outright to Kentucky, 31–21), Auburn +2.5 (lost by 5 in an 83-point game), Virginia −10.5 (lost outright to West Virginia), Louisiana −7.5 (won by exactly seven — the hook), San Diego State −1.5 (lost by 13). Totals: CHAR–APP Under 52.5 HIT (47), FRES–SJSU Under 49.5 HIT (36), MSST–SC Under 58.5 MISS (75).

Props, HIT: Byrum Brown Over 275.5 pass + rush (336), Singleton Over 3.5 receptions (7 for 99 and a score), Noah Kim Over 133.5 (140 — six yards of room), DeAndre Moore Over 45.5 (56), Stockton Under 240.5 (147 — pulled in the fourth as scripted), Nate Frazier Under 76.5 (75 — a hit by a yard and a half; logged as lucky), Kewan Lacy Under 17.5 receiving (−2), Lloyd Avant Under 66.5 (27). Props, MISS: Da'Shawn Martin Over 22.5 (1 catch, 13 — the West Virginia shootout ran through other hands), Fluff Bothwell Under 69.5 (115 on 21 carries — an under on a lead back in a 75-point game), Malik Washington Under 267.5 (359 — Maryland threw 43 times chasing Virginia Tech).

By star tier: ★★★★ 5–2 (Brown, Kim, Moore, Stockton, Lacy hit; Bothwell, Washington missed), ★★★ 3–1 (Singleton, Frazier, Avant hit; Martin missed). Overs 4–1, Unders 4–2 — the exceed-side reads keep cashing (Sep 14 directive, now 7–1 over two Saturdays and a Friday). Closing-line value, sides and totals: 7 of 14 POSITIVE (50%), 5 NEGATIVE, 2 flat, against the last eight-book Nevada read before the evening kicks — under the 60% target; the two biggest movers were both ours (Fresno −6.5 → −5.5, Sac State +26.5 → +25.5) and they split. Prop closes are not recorded: the data platform went dark at 10:15 AM ET Saturday and stayed dark through the evening kicks, so eleven legs have no closing read. Agree vs dissent: all 25 legs published Confirms/Neutral — no Conflicts leg on the card; the two Neutral legs (SDSU, Sac State) went 1–1. PLAY vs WATCH: PLAY tiles 8–3; the 22 WATCH/NO PLAY tiles are on the Insights board and were not bet.

September 18 — 5 of 6 (83%), graded September 19

Miami 33, Wake Forest 20 · Texas Tech 28, Houston 26. The Friday card was the first full-apparatus football card since the Sep 12 Saturday, and it went 5–1 with five of six legs beating the close. HIT — Wake Forest +21 (−115): lost by 13; the ALIGNED tag (public 87% on Miami, the market-maker's dollars on Wake, a line that moved 22.5 → 20 toward the dog) did what the doctrine says it does. HIT — Under 56 (−110): 53 points, and the number closed 55.5 — we held the better side of the total all week. MISS — Texas Tech −7 (−115): won by two, Houston's two-point try for the tie short in the last minute. The SHARP +21 read was right about the side and wrong about the margin; the argument against we printed (CB Boyd out, Trey White questionable, Carter a game-time call) was the game — Houston threw for 282 against that secondary. HIT — Mark Fletcher Jr. Over 76.5 rush yds (−119): 16 for 85 and a score. HIT — Carlos Hernandez Over 63.5 rec yds (−114): 10 catches, 69 — the card's one Conflicts leg (Miami 1st vs receiving yards on the season) cashed against its printed dissent, on the exact mechanism we wrote: Miami's two starting corners out, Lopez throwing 46 times. HIT — Amare Thomas Over 71.5 rec yds (−114): 7 for 75 and a touchdown, after the board had chased the line to 80.5.

Closing-line value: 5 of 6 positive (83%) — Wake +0.5/+1.0, Under +0.5, Tech +0.5 (the positive-CLV loser), Hernandez +1.0, Thomas +9.0; Fletcher −1.0 the lone negative. Prop closes are the last platform board read (Thu 6:18 PM ET — the platform was dark at Friday's 2:31 PM pre-flight); the sides and totals are the four-book Nevada read at 2:40 PM ET Friday. By star tier: ★★★★ 1–0 (Fletcher 8.0), ★★★ 2–0 (Hernandez 6.5, Thomas 6.5). Agree vs dissent: the Conflicts leg hit; the two ALIGNED/SHARP team legs went 1–1. PLAY vs WATCH: PLAY tiles 3–0; the WATCH tile with a stated direction (Overmyer, 39.5 rec yds, Carter's status the reason it stayed WATCH) went 8 for 72 — the WATCH was right to wait for Carter and wrong about what Carter playing meant.

September 17 — 2 of 7 (29%), graded September 18

Buffalo 41, Detroit 31 · Pittsburgh 27, Syracuse 13. The two Thursday cards went 2–5, and the shape of the loss is more instructive than the number. In Orchard Park we published three legs that needed the same thing — Detroit keeps it close and runs — and two that needed the opposite, and the game picked a side: the two opposite-script legs cashed and the three same-script legs died together. In Pittsburgh a card built on "Syracuse has to throw" watched Syracuse throw 32 times for 205 yards, bench its quarterback for the last drive, and lose a game Pitt ran on 37 times for 245.

Leg by leg — Lions @ Bills. MISS — Lions +5.5 (−106), lost by 10; a 14–0 first quarter and a 27–10 half. MISS — Under 54.5 (+104): 72 points — the printed argument against ("these offenses scored 90 the last time they met and Detroit's secondary is missing its two best players") was the game. HIT — Jared Goff Over 35.5 pass attempts (−120): 38, 26 of 38 for 327 and four scores, the trailing script we said was the hedge against the side — and the card's one Conflicts leg (Buffalo faced the 2nd-fewest attempts in 2025) cashed against its own printed dissent. MISS — Jahmyr Gibbs Over 89.5 rush yds (−118): 16 carries, 52, 3.3 a carry. The role was real (16 carries, 8 targets, a touchdown); the trailing script and 3.3 a carry were not the bet. HIT — James Cook Over 78.5 rush yds (−114): 21 for 135 and a score, the leading-script leg, cashed with room to spare.

Leg by leg — Syracuse @ Pitt. MISS — Steve Angeli Over 250.5 pass yds (−114): 13 of 25 for 161, two interceptions, pulled for Amari Odom (7 of 7, 44) late. The panel's 306 median was three Notre Dame games and two Syracuse games; we discounted the windows to 1.0 for the program change and left the cushion at 2.0 on the same mixed sample — the cushion was the transfer mirage the windows rule already knows about. MISS — Umari Hatcher Over 71.5 rec yds (−114): 3 catches, 58 (plus a 28-yard run that does not count). "The targets concentrate" with two receivers out was true — and irrelevant when the quarterback completes 13 passes.

Closing-line value: 3 of 7 positive (43%), 1 flat, 3 negative — below the 60% target for the second graded card in a row. Goff +1.0 and Cook +1.0 (both hit); Angeli +8.0 (the best close on either card, and it lost by 90 yards); Lions +5.5 flat; Under 54.5 −0.5 to −1.0; Hatcher −1.0; Gibbs −10 yards — the main line was cut from 89.5 to 79.5 at 5:17 PM Thursday and we flagged it in the ledger before kickoff. Closes are the last verified market read (Thu 6:16–6:18 PM ET, about two hours before the NFL kick, one hour before the ACC kick): the platform was dark at grading and the tracker had rolled to Week 3, so true closes are not recorded and that is stated rather than filled in. By star tier: ★★★★ 0–2 (Angeli 8.0, Hatcher 7.0), ★★★ 2–0 (Goff 6.0, Cook 6.0-capped), ★★ 0–1 (Gibbs 4.5). The card's highest-scored prop lost again. Agree vs dissent: the one leg published against a printed Conflicts (Goff) hit; the two team legs published against the money at both books (71–78% of the dollars on Buffalo) both lost.

September 15 — 7 of 8 (88%), graded September 16

The first full-apparatus MLB card in six runs went 7–1, and the one that lost was the one we sized smallest. Milwaukee 5, Pittsburgh 1 (6 runs, HIT — Under 7.5). Texas 4, Boston 2 (HIT — Rangers ML +/−130, the line that flipped from Boston favored to Texas favored with the market-maker's dollars, and it held). Yankees 8, Twins 1 (9 runs, HIT — Over 7.5; HIT — Cody Bellinger Over 1.5 H+R+RBI: 3-for-4, a homer, 1 run, 2 RBI — six against a line of one and a half). Detroit 10, Toronto 1 (HIT — Tigers ML +133, the ALIGNED tag at plus money). Angels 2, Mariners 1 (3 runs, HIT — Under 8.5). Washington 6, Philadelphia 3 (HIT — Nationals ML +199, the loudest split on the board — 94% of the market-maker's moneyline dollars on a +199 dog — and the biggest price we have cashed on a team leg). MISS — Braves @ Cubs Under 8.5 (+101): Atlanta 6, Chicago 3, nine runs, over by half a run. The Braves lineup that was "14th of 15 in runs since Aug 1" scored six.

What the win says and what it doesn't. Six of the seven wins were splits-doctrine legs — the market-maker's dollars against the public's tickets — and every one carried a printed dissent. One night is one night; the same doctrine's reverse-line-move cousin is still 1-for-9 on our record, and we are not promoting anything off an 8-leg sample. The miss was the leg with the thinnest case: no model, no system and no ratings edge fired on the ATL–CHC total; it published on two money layers alone at the card's lowest conviction (54), and the "argument against" we printed — Gausman's FIP better than his ERA, the number rated fair — was the argument that won. It gets a ledger row.

Closing lines: not recorded. This run graded the card the following afternoon and neither the platform's board nor the network's tracker still carries Tuesday's closes, so CLV on all eight legs is printed as a gap, not estimated. That is the cost of a grading run that happens a day late, and it is stated rather than filled in.

September 14 — 5 of 10 (50%), graded September 15

Kansas City 31, Denver 10. The Monday card beat the market on five of the six legs where we can measure it and still graded at a coin flip, because a 31–10 blowout is not a pricing event — it is a game-script event, and this card had one script bet six different ways.

Six legs, one thesis. Denver +2.5, Dobbins Over 8 fantasy, Nix Over 15 fantasy, Harvey Over 8 fantasy, Lutz Over 7.5 kicking points and Walker III Under 16.5 carries all needed the same thing: a competitive game in which Denver has the football. Denver ran 15 times for 61 yards, went 2-for-12 on third down, finished with 192 total yards and lost by 21. Five of those six lost. The sixth, Harvey, survived by one-tenth of a fantasy point on garbage-time checkdowns. We wrote on September 14 that a slate whose legs all need the same thing to go wrong is one bet; this is the same failure inside a single game, and nothing in the process flagged it before it shipped. It is the ninth through twelfth appearance of the same-script correlation candidate and it is now the largest open item in the ledger.

Leg by leg. MISS — Broncos +2.5 (+100), lost by 21. MISS — J.K. Dobbins Over 8 fantasy score: 8 carries, 36 yards, no receptions, 3.6. MISS — Bo Nix Over 15 fantasy score: 131 yards, 1 TD, 1 INT, 1 lost fumble, 7.44. MISS — Kenneth Walker III Under 16.5 carries: 23 carries, 173 yards — the exact blowout script our own Filter 3 describes, running the other way. MISS — Wil Lutz Over 7.5 kicking points: 1 field goal, 1 extra point, 4. HIT — Harrison Butker Under 7.5 kicking points: 7, on the nose of the read, and it hit against the printed dissent (Denver ranked 30th in kicking points allowed, scored a zero). HIT — Jaylen Waddle Under 55.5 receiving yards: 1 catch, 2 yards. HIT — Patrick Mahomes Over 15.5 fantasy score: 22.66. HIT — Travis Kelce Over 10 fantasy score: 10.1. HIT — RJ Harvey Over 8 fantasy score: 8.1.

Two of the five wins cashed by a tenth of a point and are logged as lucky hits: Kelce needed 10.0 and produced 10.1; Harvey needed 8.0 and produced 8.1. A card that grades 5–5 would have graded 3–7 on two more incompletions. We are saying so rather than banking the record.

Closing-line value: 5 of 6 measurable legs positive (83%), and four of those five lost. We published Denver +2.5 into a Denver +1.5 market (+1.0), Dobbins Over 8 into a 9 (+1.0), Nix Over 15 into a 17.5 (+2.5), Mahomes Over 15.5 into a 17 (+1.5) and Walker Under 16.5 at −119 into +100 (+19 cents on the same number). The one negative was Waddle Under 55.5 into a 59.5 (−4.0) — and Waddle is the one that won, on two yards. This is the cleanest illustration we have of the doctrine: CLV grades the process, the result grades the pick, and on a single game the two can point in opposite directions all night. Four legs have no recorded close — Butker, Kelce, Harvey and Lutz — because our data platform was dark when this card graded and Kelce's fantasy market was pulled before kickoff. We are printing the gap rather than estimating into it.

By Prop Score tier: ★★★★ 1–3, ★★★ 2–0, ★★ 2–1, team legs 0–1. The card's highest-scored leg — Dobbins at 8.0 — missed, which is the seventh straight grading day the top-scored prop on a card has lost. The promoted early-season cushion cap was applied to it: the tile said the cushion was built on 2025 tape and capped the leg at ★★★. The cap held the rating down and the leg published anyway, and that is the hole. A cap that changes the star count but not the decision is not a filter. That goes to the study session today.

September 13 — 49 of 95 (52%), 4 no-action, graded September 14

The biggest card we have ever published graded almost exactly at a coin flip, and the way it got there is the lesson. Ninety-nine legs, thirteen games, one direction. Every game total on the card was an Under, and all four of them lost: TB–CIN Under 50.5 (60 points), NO–DET Under 49.5 (61), BAL–IND Under 47.5 (64) and WAS–PHI Under 44.5 (46). That is not four bad reads. That is one bad read, sold four times, on a Sunday the league scored in. We are writing it that way because pretending those were independent legs is how a 51% card starts calling itself a system.

Team legs went 4–5. The sides carried it: Bears −2.5 in a 59–37 track meet, Jaguars −7.5 (34–10), Vikings −1.5 (39–22) and Cardinals +9.5, who did not need the points at all in a 26–14 win at Los Angeles. Dolphins +3.5 was the side that died, 27–13. Worth saying out loud: three of the four winning sides came with positive closing-line value — we had Chicago at −2.5 into a −3 close, Jacksonville at −7.5 into −8.5, Minnesota at −1.5 into −2.5 — so the sides were priced right and the totals were priced wrong.

Props went 45–41, and the pattern inside them is worse than the number. We bet against the same game script twice and lost twice, in two different games. On Baltimore–Indianapolis we held the Under, Lamar Jackson Under 219.5 passing and Jackson Under 36.5 rushing — one thesis, three tickets — and Jackson threw for 324 with 40 on the ground in a 41–23 win. On New Orleans–Detroit we held the Under, Jahmyr Gibbs Under 17.5 carries and Gibbs Under 82.5 rushing; Gibbs ran 29 times for 156. Two games, six legs, zero wins, and in both cases the correlation was visible before kickoff.

We also hedged ourselves three times and went 1–1 each time, which is the most expensive way to be right. Chase Brown Over 15 fantasy (18.8) cashed while Brown Under 14.5 carries (16) lost. Bryce Young Over 15 fantasy (32.4) cashed while Young Under 209.5 passing (361) lost badly. De'Von Achane Under 31.5 receiving (30) cashed while Achane Over 17 fantasy (10.6) lost. The house rule already says hedged pairs mean stay away; the card broke its own rule three times.

The single worst leg was Ja'Marr Chase Over 19 fantasy, which returned 3.2 — two catches for twelve yards on four targets in a game his team won 33–27. A WR1 with four targets in a sixty-point game is a usage collapse, and with our data platform's pricing layer dark tonight we cannot confirm the cause from the panel. It is logged as unexplained rather than explained away.

Four legs graded no-action. Tua Tagovailoa, Brock Bowers and Darius Slayton never appeared in a box score and are VOID, not losses. Khalil Shakir Over 9 fantasy landed on exactly 9.0 — a true push, recorded as VOID because that is the only no-action bucket the grader has, and stated here so the 9.0 is not read as a duck. We also audited five zeroes before grading them as wins: Pitts, Bateman, Loveland, Addison and Jennings all carried targets, so they played and produced nothing. Those unders are real wins, not DNPs wearing a win's clothes.

Also graded today: September 11 and September 7, both already written up below, now synced leg-by-leg in the source files.

September 12 — 18 of 31 (58%), graded September 13

Saturday's forty-five-game monster grades out as the best day the prop desk has had, and a coin-flip afternoon everywhere else. The tile grid went 9–3 and the three 8.0-scored legs went 3–0 — Underwood Under 28.5 attempts (17 in a 17–10 rock fight), Reed Under 249.5 (163 on fifteen completions in a blowout that never needed him), and Williams Jr. Under 338.5 pass-plus-rush (305 in a 16–14 escape). After six straight grading days of the card's top-scored leg missing, the streak is over — and it broke on the first card built under the early-season cushion cap, which is exactly what promotion was supposed to buy. The supporting cast held: Mestemaker Over 208.5 sailed (317), Hawkins Over 14.5 attempts hit (23), Davison Under 15.5 attempts hit (8), Laughery Under 10.5 survived by half a carry (10 — the leg we capped at 6.0 for the Forster tag), Bo Jackson Over 54.5 ground out 74, Roebuck Under 77.5 landed at 59.

The prop misses: M. Washington Under 262.5 P+R (350) — Maryland hung 38 on UConn and the trailing script we bet against on the side is the same script that fed him 41 attempts; we held UConn +11.5 AND the QB under, one competitive-game thesis priced as two legs, and both died together. Raynor Over 151.5 (117) — the Cy-Hawk game did Cy-Hawk things, 13–16, and our Over 40.5 died in the same rubble. Carr Under 251.5 (253, by yards) — the blowout script we wanted arrived exactly as drawn, Notre Dame 52–0, Carr done early at twenty throws — and 12.7 a throw beat the number anyway. An attempts market wins that bet; a yards market left it to efficiency.

The sides and totals went 8–8, and the losses cluster where our dogs simply didn't show: Georgia Tech (by 21), Old Dominion (by 23), Eastern Michigan (by 28), UConn (by 24) — four double-digit dogs that lost by three touchdowns-plus. The wins were the market-facing ones: Duke +6.5 won outright 31–27, Michigan +4.5 won outright 17–10 with the Under 43.5 (27 points), Temple +24.5 covered a 27–9 game, Boise handled Memphis by 18, Sam Houston covered by a touchdown, Hawai'i and South Alabama both cashed. Alabama–Kentucky Under 48.5 was never alive (62), and Duke–Illinois Under 51.5 died at 58 while its side leg cashed.

MLB went 1 of 3 on the restore card: Brewers −174 cashed 13–9; the Cardinals at −103 lost a one-run game to the White Sox, and Arizona −122 got handled 6–2 by a Rangers team we graded as fading.

By star tier: ★★★★ 7–3, ★★★ 2–0. CLV is unavailable for this entire batch and we say so rather than guess: the data platform was dark at grading time and the Nevada tracker drops finished games, so no closing number could be verified on any of the 31 legs. Running record moves to 192 of 373 (51%), 60 cards.

September 11 — 1 of 4 (25%), graded September 12

Friday's card is on the books, and it's a rough page. Boston College 28, Rutgers 21 — the dog covered nothing: Rutgers +3.5 (−120, conviction 55) is a loss by exactly the margin the hook doesn't save. The public book's 27-point dollar gap on the dog, the full point of line move, two ratings calling it a one-point game — all of it lost to the layer we printed as dissent and overrode: the market-maker's moneyline money was 90% Boston College at publish, and Boston College won by seven behind four Rutgers interceptions. Lesson 6 takes another scalp: sharp-money signals are tiebreakers, never theses, and the market-maker's dissent is starting to look like a veto.

The props went 1–2, and both losses argue with our screens. The hit: M. McKenzie Over 57.5 rush yds (−125) — 94 yards and a rushing score on the read we wrote, a running quarterback meeting a 115th-ranked front. The misses: K. Duff Under 105.5 rec yds (−115) got buried — 8 catches, 137 yards and two touchdowns, because the first-career-start quarterback we priced as a downgrade did what first starts do: threw to the alpha. The 43% Week 1 target share we printed as dissent was the floor, not the anomaly, and the books' 97.5 → 106.5 march was information we scored as value. K. Kubecka Under 20.5 rec yds (−119) — the card's highest-scored leg at 8.0 — lost to a single 21-yard catch: 3 receptions, 38 yards. A two-season median of 10.5 on a 9% share, windows in the nineties, and one chunk play beat the whole file. The card's top-scored leg has now missed on six straight grading days, and this one was built on exactly the 2025-tape cushion the newly promoted cap was written for.

Star tiers on the day: ★★★★ 1–2 (all three props scored in the 7.0–8.0 band) · team 0–1. CLV, with an honesty note: our data platform was dark at re-verification and at close, so prop closes are partly unverifiable — printed rather than guessed. What we can verify: Rutgers +3.5 flat (−120 board / +3 −115 Nevada at the 4:08 PM ET re-check), Kubecka flat (20.5 on the consensus board at 4:08 PM), Duff one point against on the last observed number (the board dealt 106.5 Friday morning; we published 105.5 Thursday's number), McKenzie unobserved. Second-opinion ledger: the market-maker's BC dissent won; our stability-chart fade lost; the Border War pass verdict was the right call — Missouri covered −4.5 easily and the +6.5 stale number we named would have lost too. Running record moves to 174 of 342 (51%), 58 cards.

September 10 — 5 of 11 (45%), graded September 11

The Melbourne game is on the books: 49ers 27, Rams 7 — and the sharp side we rode from Sunday cashed without a sweat. San Francisco +3.5 (published at −104) won outright behind three Purdy touchdown throws; the market-maker's 77%-of-money print on the dog at lock was the loudest signal of the week and it was right. One honesty note goes with the win: the board closed +3.5 at +103, nineteen cents better than our Sunday price — a winning side, bought early, at the worse number. It goes in the ledger anyway; that's what the ledger is for.

The props went 4–6, and the total told us why in advance. The hits: K. Williams Under 13.5 rush attempts (−119 as published) — eleven carries, exactly the Shanahan-front read; K. Williams Over 11.55 Fantasy (UD) cashed at 14.0 on the goal-line touchdown; D. Adams Under 4.5 receptions (−119) came home on three catches — a leg we ourselves re-scored to Conflicts on Tuesday, so it grades as a win against our own printed dissent, 1–1 with Purdy on the card's two wrong-way-opponent legs; and C. McCaffrey Over 4.5 receptions (−119) landed on the number, five catches on eight targets in exactly the dog-checkdown script we priced — except the dog never trailed.

The misses: B. Purdy Under 33.5 attempts (−119) died by half an attempt — 34 — in a game his team led wire-to-wire; the printed Conflicts dissent (LA 26th vs attempts) decided a leg for the second time in three days. M. Stafford Over 18.35 Fantasy (UD) produced 5.1 — 155 yards, no scores, a pick, and a fourth-quarter seat for Stetson Bennett. P. Nacua Over 14.85 Fantasy (UD) managed 9.9 on five catches. G. Kittle Over 10.5 Fantasy (PP) finished 3.2 — two catches on five targets — and the books told us: cleared Thursday morning, and the line got cut anyway, 10.5 → 9 → 8.5. A market that keeps selling a cleared player was pricing a snap plan, we printed that sentence on the card, and we left the leg a PLAY anyway. K. Williams Over 53.5 rushing (−109) got 41 — Blake Corum's ten carries took the committee share our 52% number said didn't exist — and C. Parkinson Over 19.5 receiving (+105 at lock, graded −119) got one target's worth: six yards, the three-man room we scored as a committee doing committee things.

Star tiers on the day: ★★★★ 3–3 · ★★★ 1–2 · ★★ 1–0 — and all three 8.0s on the card missed. Stafford, Nacua and Williams' rushing yards were the three highest-scored legs we published, they went 0–3, and Lesson 7 takes its fifth straight scalp at the top of a card. The common thread is printed in our own Crowd-vs-Cash section: 84% of the sharp book's total money was on the Under 47.5, we called it "the market's loudest opinion this week," played none of it, and then lost four passing-volume overs in a 34-point game. The dissent we printed and didn't connect went 4–0 against us. CLV: positive on 2 of 10 props (Nacua +2 points, Williams rushing +3 — both losers; good process, bad result), flat 2, negative 6 — 20% positive is the worst CLV day on the books and the second straight under the 60% bar. Second-opinion ledger: the sharp-side agreement on the 49ers won; the Under-total dissent went 4–0 against our overs. Running record moves to 173 of 338 (51%), 57 cards.

September 9 — 4 of 9 (44%), graded September 10

The first regular-season NFL card is on the books: Seahawks 13, Patriots 10 — and the side we refused to touch pushed on the number. SEA −3 won by exactly three; the stay-away verdict we held through four passes cost nothing and would have earned nothing, which is what a fair number looks like. The total sailed under 44.5 by three touchdowns. The game itself broke strange: Sam Darnold injured his hip on the opening drive (third-down sack, two pass attempts, gone), Drew Lock played the rest, and Drake Maye threw three fourth-quarter interceptions to hand the champion the win.

The hits: J. Smith-Njigba Over 17 Fantasy (−119) was never in doubt — 8 catches, 122 yards and the game's only receiving touchdown for 26.2; the 29% target share thesis survived a backup quarterback without blinking. R. Stevenson Over 12 Fantasy (−119) cashed at 14.5 on lead-back volume (18 carries, 5 catches) — the Henderson-out role read, bought at 12 before the market moved the line to 15. R. Shaheed Under 2.5 receptions (−119 as published) came home on one catch. And S. Darnold Under 15.5 Fantasy (−119) "won" at 0.52 — two pass attempts before the injury cart. We are counting it, by the same rule that made Jonquel Jones a loss on a six-minute exit: the player took the field and the market settled. But it goes in the ledger as a lucky hit — the script-and-defense read we published was never tested, and we know the difference between being right and being paid.

The misses: D. Maye Under 31.5 attempts (−118) — the card's highest-scored leg at ★★★★ 7.0 — died on 33 dropbacks; the printed dissent (Seattle 28th vs pass attempts) was the layer that failed, and Lesson 7 takes its fourth straight scalp at the top of a card. D. Maye Over 17.5 Fantasy (−119) finished 12.82 — three interceptions are minus-three, and there is a correlation confession owed here: attempts under plus fantasy over on the same quarterback is one bet on efficiency, not two bets, and we priced them as two. R. Stevenson Under 11.5 attempts (+100) was the dead read riding at 40 since Henderson went out — 18 carries, graded at the published +100 against a board that closed 15.5. H. Henry Over 8 Fantasy managed 5.6 on three catches, the contradicting 52% season window doing exactly what contradicting windows do. And R. Doubs Over 8.5 Fantasy posted a zero — three targets, no catches, a 71% window earned entirely in Green Bay meeting its first Patriots Wednesday.

Star tiers on the day: ★★★★ 0–1 · ★★★ 4–2 · ★★ 0–2. The rubric's middle went 4–2 while its top and bottom went 0–3 — the fourth grading day in a row the highest-scored leg on the card missed. CLV, against the final platform capture (Wed 5:07 PM ET; the true close is unavailable — the platform board was dark on the grading morning): positive on 3 of 9 (Maye attempts +1, Maye fantasy +0.5, Stevenson fantasy +3 — the best number on the card), flat on 3, negative on 3 (Stevenson attempts −4 on the Henderson news, Doubs −0.5, Shaheed 11 cents of juice given back). 33% positive is under the 60% bar and the number says what it says: half this card was priced before the news cycle finished. Second-opinion ledger: the one leg the Vegas layer's AI model dissented on (Maye passing volume, projecting OVER) — the model's direction won. Dissent record on legs it argued against: still unbeaten. Running record moves to 168 of 327 (51%), 56 cards.

September 7 — 4 of 10 (40%)

Two cards graded on the Tuesday run: the Labor Day one-gamer in Tallahassee and the restored MLB slate. Both went 2 of 5, and both losses columns have something to teach.

The NCAAF card went 2 of 5, and the half points went the wrong way twice. SMU won 27–24 on a field goal with 21 seconds left in a game delayed nearly two hours by a stadium power outage — so Florida State +2.5 (+112) lost by exactly the hook. The number we spent three days demanding, +3, arrived at the market-maker on Monday afternoon and would have pushed; we had priced in at +2.5 on Sunday morning under printed protest and the half point decided the grade. The market closed SMU −3, a half point of negative CLV that turned out to be the whole result. The dissent column called it: the Vegas expert room went 8-of-9 on SMU against our lean, and the room was right while all three rating systems were wrong. On the props: K. Jennings Over 270.5 passing cashed emphatically (430 yards, three scores) and Y. Knight Under 49.5 receiving survived by half a yard at 49 on six catches — the full sweat. D. Robinson Over 67.5 receiving, the card's highest-scored prop at 8.0, caught two balls for 25 yards and lost a fumble; Lesson 7 takes its third straight scalp at the top of a card. K. Jennings Under 10.5 rushing died on 33 scramble yards in a game SMU's backfield couldn't run in — the pocket escapes we priced as contained showed up as first downs.

The MLB card went 2 of 5, and it earned the ledger a new entry. The wins: WSH–SD Under 8.5 (3–2 at Petco, the card's top lean and the cleanest thesis on it) and Chase Burns Over 4.5 Ks — five strikeouts in three no-hit innings before the Reds pulled him at 44 pitches on his innings limit. The losses: Braves ML +150 (1–0 Phillies on a Schwarber homer off the bullpen — the sharp-money shape at both books bought a shutout loss), CLE–BAL Under 7.5 (6–4, ten runs, two Coby Mayo homers into the 96%-money under), and Nick Pivetta Over 4.5 Ks — two strikeouts in five shutout innings on a 63-pitch leash in his first start back from the injured list. Pivetta and Burns were the same bet on the same night with the same flaw: both arms were capped and neither cap was in the leg note. One cap arrived after five strikeouts, one before three. That is Lesson 9 now, printed above, and pitcher K overs carry a workload check from today.

Star tiers on the day: ★★★★ 0–1 · ★★★ 3–2. Second-opinion ledger: legs the Vegas layer dissented on went 0–1 against us being right (FSU), and legs it agreed with went 2–3 — the layer's dissent record stays perfect at small sample, which is exactly the kind of sentence we write down so we can laugh at it later if it stops being true.

September 6 — 4 of 14 (29%)

Two cards graded on the Monday run, and neither one is getting framed. The Sunday NCAAF card went 1 of 7. Team legs 0–2: Wisconsin +20.5 died 41–13 — the 135th-ranked offense we docked the leg for showed up exactly as ranked — and Ole Miss −6.5 won by three, 41–38, in a shootout that covered for everyone except us. Props 1–5, and the order of finish is the lesson: C. Joseph Over 180.5 pass+rush cashed at 220 (217 through the air, 3 on the ground — Notre Dame's 10th-ranked run front took the legs and the arm delivered anyway), while D. Williams Jr. Under 299.5 — the card's only ★★★★★ at 8.5 — went over by 29 on 268 passing and 61 rushing in a game Washington won by only 14; the blowout-throttle script never arrived. CJ Carr Under 229.5 missed by 9.5 on 239. I. Brown Over 71.5 got 16 carries and turned them into 36 yards — the 8.8-a-carry profile met an Ole Miss front that decided otherwise — and B. Foley Over 18.5 caught one ball for 3. The two legs the rubric liked most, Williams and Carr, both lost; the one it scored 8.0 with a printed Conflicts flag won. Lesson 7 stands reinforced.

The MLB game-lines card went 3 of 7. The wins: Braves ML (5–4 at Philadelphia — the coin-flip price on the better team, exactly the thesis), D-backs ML (3–2 in Houston), and Christian Scott Over 5.5 Ks — seven strikeouts in 4.2 innings, the short-hook risk we printed never got the chance to matter. The losses: Cubs ML −14 units of ugly in a 10–3 Marlins rout, Rays ML 6–8 in Texas, and the Angels +163 flyer lost 1–0 — Ureña was every bit as real as the case said, and the game's only run was a Henry Davis solo homer, which also killed our Davis Under 0.5 Total Bases leg on the same card. One swing of a .169 hitter's bat, two legs. That is a correlation we sold ourselves as a script agreement and should have priced as shared risk: the friendliest game on the slate for a bases under was only friendly if Pittsburgh's bottom of the order stayed quiet, and our other leg needed Pittsburgh to lose. Star tiers on the day: ★★★★★ 0–1, ★★★★ 1–2, ★★★ 1–2, ★★ 0–1. Running record moves to 160 of 308 (52%).

September 5 — 12 of 31 (39%), 1 void

The worst day of the young season, printed as loudly as the best one. The Saturday Week 1 card carried thirty-two legs — six team positions, twenty-six panel-scored props — and graded 12 of 31 against final box scores on Sunday morning, with J. Nicholas voided (ruled out minutes before kickoff, never active).

Team legs went 1–5. The lone cover was the one no splits existed for: Western Michigan +27.5 lost 13–12 and nearly won at the Big House. Everything else broke: Clemson +10 was a 51–10 rout the wrong way — the reverse-line move we sided with (70% of spread money on LSU, line falling toward Clemson) walked into the most lopsided result on the slate; Tulane +7.5 (83% of moneyline money on 20% of tickets) lost by 14 in a 17–3 rock fight; Auburn −7.5 won by one, 17–16, and the model dissent we filed on the card (EV+ on Baylor's moneyline) was closer to the truth than our side; ECU–Alabama Under 52.5 went over on 48–10 — one team cleared the number alone; Liberty +6 lost by exactly seven, 20–13. Two of the day's five team misses were sharp-money reads that had just cashed twice on Friday. The signal is a tiebreaker, and Saturday is why it stays one.

The props went 11–14 with one void — and the Prop Score rubric's first grading day ran exactly backwards: ★★★★★ 0–3, ★★★★ 4–6, ★★★ 5–5, ★★ 2–0. The three five-star misses — Byrum Brown Over 70.5 rushing (21 carries, 27 yards), Byrum Brown Over 323.5 pass+rush (286), T.J. Moore Over 44.5 receiving (31 in the rout) — were all maxed on median cushion and hit-rate windows, both built on 2025 samples. Week 1 broke the continuity those components assume: Auburn scored 17 on the 122nd-ranked run defense, and Clemson's garbage-time script never fed Moore. Meanwhile the 2–0 two-star tier was both plus-money interception overs (Lagway, Leavitt) — legs the rubric scored thin on cushion and we played for the price. Lesson filed: cushion and windows overweight last season's tape in Week 1, and the component analysis of the misses says so — that is what storing the components was for.

The wins that did land were mostly game-state unders and role plays: Underwood's pick and his passing under (170 yards), Goodwin (zero catches on one target), Z. Marshall (zero targets — graded a HIT with the judgment printed: no availability flag anywhere, presumptive TE1, we grade the zero and say so), Lowry Over 28.5 rushing (35), Moulton Over 85.5 rush+rec (183 rushing alone), D. Hill and A. Gray's rushing unders, T. Green's four catches, Lagway and Leavitt interception overs. The narrow agony award goes to E. Dickens Over 11.5 rush attempts: eleven carries — one handoff short.

September 4 — 6 of 11 (55%)

The Week 1 Friday card, graded against final box scores on Saturday morning. The two sharp-money reads both cashed: Toledo +10.5 covered by exactly half a point (MSU won 30-20 — the line closed 9.5, we held 10.5, and every fraction of that closing-line value turned out to be the whole bet), and San Jose State +2.5 won the game outright, 27-21 — the sixteen-minute line move we chased on Wednesday closed EMU −1.5 and the Spartans never needed the points. The raise that leg earned and was refused on the kickoff rule would have been the right raise; the rule stays anyway, because rules that flex when they would have helped are not rules.

The Miami game was the card's teacher. Miami 45, Stanford 6. Stanford +23.5 (docked twice to 48 on the fifteen-out injury report, the torn-up offensive line and the negative CLV — every reason printed in advance) lost by 39: the dock was right, the leg was still wrong, and the injury class we flagged as the one that breaks big-dog covers did exactly that. Mensah Under 236.5 — the card's flagged Conflicts leg, docked 52→45 — missed by 165 yards (27-of-30, 401, five touchdowns, a school record). Stanford's dead-last pass-defense ranking and his own 71% over-window, the two data points the platform used to argue with us, were both correct and we said so before kickoff. That is what the flag is for. On the same field, Toney Over 6 receptions cashed by six (12 catches, 234 yards — another school record) and Warren Under 0.5 passing TDs cashed clean (21-of-41, zero scores behind the patchwork line). The game script we feared for Stanford was the game script that paid us twice.

The blowout-script props split. Robinson Over 54.5 rushing MISSED (7 carries, 30 yards — OU led 51-0 and still never fed the backup; the second-half-volume thesis died of too much passing efficiency), while Trell Harris Over 47.5 receiving HIT (2 catches, 61 yards and a score — the flat line next to the crashed one was indeed the stale number). Maiava Under 263.5 MISSED badly (349 yards in a 39-0 rout — USC never throttled down, and a 23-of-25 night beats any script), but Hines Under 47.5 HIT (2 for 18 — the Questionable tag resolved to a quiet night, and Fresno's game never produced the garbage-time volume that voids unders). McCray Under 50.5 MISSED (8 catches, 69 yards — the role expansion the books priced at twenty yards over his median actually showed up on opening night; we faded a promotion that was real).

The sizing column had its best night of the young football season. The two lowest-confidence legs on the card (Stanford 48, Mensah 45) both lost; the two highest (Toledo 58, Toney 57) both won. Six of the seven legs sized 52 or better cashed. The Conflicts gate is now 1-for-1 on flagged football legs kept for grading — kept, sized down, and beaten exactly as the flag warned.

September 3 — 4 of 8 (50%, 1 UNVERIFIED)

The Week 1 Thursday card, graded against final box scores in the early hours of September 4. The headline cashed: Colorado +6.5 at +106 won the game outright, 14-13 — the reverse line move, the model's double EV+ badge and the 64%-money/28%-tickets split all pointed the same way, and Georgia Tech's missed field goal turned a cover into a win. Julian Lewis' clean-game prop cashed at plus money too (zero picks on 27 attempts, against the defense that couldn't take the ball away in 2025). The other side of the ledger: Utah won 66-14 — the blowout we scripted — but so fast the script inverted. Dampier threw three first-half touchdowns and still stayed under his passing number (our under HIT even though our own platform dock argued against it), while Parker missed his rushing over by 3.5 yards because he scored twice as a receiver instead. Mendoza completed 17 against our under 16.5 — a half-completion miss on a night the auto-trend feed had flagged 17.5 as the safer line. The matchup-gate note for the file: both Conflicts-flagged legs (Bray, Pegan) lost exactly as the re-mark warned — two vindications — while the Dampier dock was wrong. Shoels recorded no statistics of any kind and his participation could not be confirmed: UNVERIFIED, excluded from the count. Running total moves from 134/244 to 138/252 — 55% all-time.

September 2 — 4 of 6 (67%, 1 VOID)

Two cards, seven legs, graded against final box scores the morning of September 3. The headline is the leg that died: Dylan Cease's nineteen-start strikeout streak ended at four punchouts — six shutout innings, 83 pitches, an 11−0 laugher that removed every incentive to chase whiffs. Cleveland's low-whiff lineup profile, the exact thing the market was pricing at plus money, was the whole story; we printed that argument against our own #1 and then lost to it. Meanwhile the card's flagged Conflicts layer-test — Nathan Church into Yamamoto — CASHED (2-for-4, two runs), which is now a second data point against over-trusting the matchup gate on hot bats. The game-lines card went 2-1: the Tigers hangover buy paid in twelve innings, the Brewers ace tax cleared, and the Padres getaway-day ambush we flagged in print showed up on schedule. Running total moves from 130/237 to 134/244 — 55% all-time.

August 29 — 16 of 22 (73%)

Three cards, twenty-five legs — the twenty-one-leg NCAAF Week 0 flagship plus two clean two-leg cards — all graded against final box scores the morning of September 2, after three dark days. The Week 0 card went 12-of-18 with two voids and one unverified; the WNBA and MLB game-lines cards each went a perfect 2-of-2. The running number moves from 114/215 to 130/237, which lifts the all-time rate from 53% to 55% — the site's best grading day yet, delivered by the season we said we couldn't stat-model.

NCAAF Week 0 card — 12 of 18 (2 VOID, 1 UNVERIFIED)

The method verdict first, because it's the story: team leans went 4-4 — a coin flip, priced like one (average confidence 51) — while the role-based player props went 8-2. The market pre-paid for role expansions the tape never showed, and the tape won: Trader (2 for 19 against a 45.5 line bid up from 38.5), Thomas (12 carries under 14.5, the committee exactly as scripted), Hall (36 yards under a 47.5 line that had been bid up 22 from its open), Hedden (26 completions in catch-up mode), Barfield (one carry), Maiava (286 under 304 with the fourth-quarter exit the spread promised). The two prop misses were both rushing volume in blowouts — Miller and Payne — the exact script the card flagged as the over case. The UNLV -3.5 lean we publicly declared NO PLAY at 30 confidence lost by six: the kill-signals were right, and declining to bet it was the best call on the card.

WNBA team-calls card — 2 of 2

MLB game-lines card — 2 of 2

Two web-primary cards, six team legs, all graded off final box scores the morning of August 29. The WNBA card went a clean 2-of-2 — both schedule-and-injury reads cashed without sweat. The MLB game-lines card went 1-of-4 on its probation night: three reverse-line-move leans died, one lived. The running number moves from 111/209 to 114/215 — the all-time rate holds at 53%.

WNBA team-calls card — 2 of 2

MLB game-lines card — 1 of 4

August 25 — 6 of 17 (35%)

Three cards, seventeen legs, all seventeen verified against final box scores — graded the afternoon of August 28 after two dark days. The prop card bounced back from its 0-for-4 with a 4-of-6, the WNBA card went 2-of-6, and the game-lines card posted the site's first 0-for-5 — the money-divergence layer's worst night on record. The running number moves from 105/192 to 111/209, which drops the all-time rate to 53%.

MLB Edge Gap card — 4 of 6

WNBA card — 2 of 6

MLB game-lines card — 0 of 5

What the day taught the layers: the matchup gate got its cleanest split yet — the Plum Conflicts test lost to the exact defense the gate pointed at, while the Buxton Conflicts test won on a mechanism (runs-not-RBIs) the gate doesn't model; layer tests now stand 1-1. The money-divergence layer went 0-for-5 on game lines and its season record no longer clears a coin flip — it stays published, demoted from thesis to input. And the injury-cascade rule graduates from suggestion to gate: Rivers is the second leg this month where a named cascade risk materialized and beat us.

August 24 — 3 of 9 (33%)

Two cards, nine legs, all nine verified against final box scores — graded the morning of August 25. The prop card went 0-for-4, the worst single-card day this site has published, and the game-lines card went 3-for-5 and carried the whole date. The running number moves from 102/183 to 105/192, which drops the all-time rate to 55%.

MLB Edge Gap card — 0 of 4

MLB game-lines card — 3 of 5

August 23 — 6 of 12 (50%)

Four cards, twelve legs, all twelve verified against final box scores. No voids and nothing unverified among the published legs. The running number moves from 96/171 to 102/183, which leaves the all-time rate at 56%. The cards published mid-afternoon rather than at 6:30 AM because the automation fired late, and the shadow ledger below puts a number on what that cost.

MLB Edge Gap card — 1 of 4

MLB game-lines card — 1 of 4

WNBA card — 2 of 2

NFL preseason card — 2 of 2

Edge Gap ranking had its first clearly good day since we started counting. Six player props ranked by Edge Gap: the top three went 2-for-3, the bottom three went 1-for-3. Across Aug 16–23 the largest edges are now 12-for-25 (48%) and the smallest 18-for-28 (64%). One day does not un-invert a ranking, and the honest summary is unchanged: we still cannot say Edge Gap ranking works.

The matchup layer had a bad day where it mattered most. Every leg on the MLB card was Confirms — the platoon gate approved all four — and the card went 1-for-4. Player props overall: Confirms 2-for-5, Neutral 1-for-1. Running totals: Confirms 8-for-17 (47%), Neutral 21-for-34 (62%). The gate now trails Neutral by fifteen points on seventeen legs. What we cannot see is the other half of the ledger: the gate also removed twenty-two conflicts from that same board, and we have not graded the drops since Aug 16. That audit is overdue and we are saying so here rather than quietly hoping the question goes away.

Team picks went 1-for-4 and are now 17-for-34 (50%) all-time. Both wide-divergence follows lost — the 27-point Coors Over and the 31-point Minnesota moneyline — while the one lean where the money simply agreed with the fundamentals (Dodgers −1.5) cashed. The pattern across two weeks is uncomfortable for the divergence thesis and we are watching it say so in the numbers.

Confidence: the derived figure was flat and the hand-set figure inverted again. All twelve legs ranked by published confidence split top six 3-for-6, bottom six 3-for-6. Cumulatively across Aug 18–23 the top halves are 21-for-36 (58%) and the bottom halves 20-for-37 (54%) — still nothing. The six hand-set numbers alone (four team picks, two football legs) split top three 1-for-3, bottom three 2-for-3: the two highest hand-set numbers on the day both lost, and the two lowest both won. Playing-time mode itself went 2-for-2 — the reads were right; the confidence ordering of them was backwards, again.

The projection layer earned its keep on the one card that used it. Both WNBA legs carried a recorded projection verdict (7.1 against a 7.5 line, 4.35 against 4.5 — margins of 0.4 and 0.15), both were on our side of the line, and both legs won. Two legs is not a sample, but this is the first day the per-leg projection instrumentation we promised on Aug 22 existed and could be graded at all.

August 22 — 17 of 32 (53%)

Three cards, thirty-two legs, all thirty-two verified against final box scores. No voids and nothing unverified. The running number moves from 79/139 to 96/171, which leaves the all-time rate at 56%. This was by some distance our largest published day, and the size of it is itself the story: the MLB game-lines post carried thirteen team picks on a fifteen-game slate, which is close to the number of leans we told ourselves on this very page we should not be manufacturing.

MLB game-lines card — 7 of 13

WNBA card — 4 of 8 player props, 2 of 3 team picks

NFL preseason card — 4 of 8

Confidence tiers: the first day the top half finished ahead, on the biggest sample we have had. Rank all thirty-two graded legs by the published confidence figure and split them: the top sixteen went 9-for-16, the bottom sixteen went 8-for-16. That is a one-leg difference and it is nothing, but it is the first time the split has pointed the right way on a sample of any size. Cumulatively across August 18–22 the top halves are 18-for-30 (60%) and the bottom halves 17-for-31 (55%). We flagged this figure as the number to watch four days ago. It has now been inverted once, correct twice and flat once. We still cannot say the confidence figure carries information.

Football confidence pointed the right way for the first time in five cards. The eight preseason legs split top four 3-for-4, bottom four 1-for-4 — the exact opposite of the High/Moderate inversion we recorded across August 13–15. That older label scheme is retired and stays on the record at High 2-for-7, Moderate 4-for-4; the numeric figure that replaced it has one good day.

Playing-time intel converted on four of eight, and the failures were all the same failure. The two reads that said "this quarterback inherits the bulk" both hit (Winston 150, Milton 179). The two reads that said "a two-man split caps both quarterbacks" went 1-for-2, because Spencer Rattler took twenty-three attempts and 141 yards out of a split we had described as capping him at 67.5. And the Mayfield leg was the cleanest miss of the day: we read a confirmed "appearance" with Godwin active as two real series and he threw four passes. The pattern is that we are good at identifying who plays and poor at estimating how much.

Edge Gap on the WNBA card was perfectly flat. Eight player props: ranked by Edge Gap, the top four went 2-for-4 and the bottom four went 2-for-4. Across August 16–22 the largest edges are now 10-for-22 (45%) and the smallest 17-for-25 (68%). Six days, forty-seven legs, and the ranking is still upside down. We are not going to keep calling this variance without saying plainly that it has now had enough legs to start looking like something else.

The matchup layer had a fine day and is still behind Neutral overall. Player props only: Confirms 2-for-3, Neutral 2-for-4, Conflicts 0-for-1. Running totals: Confirms 6-for-12 (50%), Neutral 20-for-33 (61%).

August 20 — 5 of 8 (63%), plus 1 void

Three cards, nine legs. Eight graded, one voided because the player never appeared. The running number moves from 74/131 to 79/139, which leaves the all-time rate at 57% before August 22 pulled it back to 56%.

MLB Edge Gap card — 3 of 4, 1 void

WNBA card — 1 of 1

MLB game-lines card — 1 of 3

Edge Gap inverted again on the MLB card. Five graded props ranked by Edge Gap: the top two went 1-for-2 and the bottom three went 3-for-3, and the largest edge on the board lost for the fifth time in six days. Ramos lost by one total base.

The matchup layer: Confirms 0-for-1, Neutral 4-for-4. One Confirms leg lost and the other voided. Every Neutral leg on the day won. Eight legs across the two verdicts is not a sample, and it is the fourth consecutive card on which Neutral has finished at or above Confirms.

Confidence tiers pointed the right way, mildly. All eight graded legs ranked by confidence: top four 3-for-4, bottom four 2-for-4.

August 19 — 6 of 12 (50%)

Three cards, twelve legs, all twelve verified against final box scores. No voids and nothing unverified. The running number moves from 68/119 to 74/131, which drops the all-time rate from 57% to 56%. An even split is the least interesting headline we have posted; what happened underneath it is not.

MLB Edge Gap card — 3 of 6

WNBA card — 1 of 1

MLB game-lines card — 2 of 5

Edge Gap: the largest edge lost again, and the overall inversion keeps narrowing. Seven player props graded. Ranked by Edge Gap, the top three went 2-for-3 and the bottom four went 2-for-4 — a marginal win for the method, the first time in four days the top half has finished ahead of the bottom. But the single largest edge on the board lost for the fourth day in five: Chris Bassitt at +35.4, the biggest gap we have ever published, struck out six against an L10 median of 2.5. Across Aug 16–19 the largest edges are now 7-for-16 (44%) and the smallest 12-for-18 (67%). The gap has closed from 33 points to 23. We still cannot say Edge Gap ranking works.

The matchup layer had another poor day and is now clearly behind Neutral. Player props only: Confirms went 1-for-2, Neutral went 3-for-5. Running totals: Confirms 4-for-8 (50%), Neutral 14-for-25 (56%). Eight legs is not a sample, and we would note that the one Confirms leg that lost — Jake Bauers, a left-handed bat with the platoon edge against a right-hander — lost by going 0-for-4 with three strikeouts, which no gate detects. But the layer has now been at or below Neutral for three consecutive days and we are recording that rather than explaining it away.

Playing-time confidence tiers: unchanged at High 2-for-7, Moderate 4-for-4. No football card has run since Aug 15; the next is Aug 22.

August 18 — 7 of 9 (78%), plus 2 voids

Three cards, eleven legs. Nine graded, two voided because the player never took the field. The running number moves from 61/110 to 68/119, which lifts the all-time rate from 55% to 57%. This was our best graded day, and the reason it was our best day is the part of the card we have been most sceptical about.

MLB Edge Gap card — 2 of 3, 1 void

WNBA card — 1 of 2, 1 void

MLB game-lines card — 4 of 4

Edge Gap: the largest edge finally won, and the inversion softened. Five player props graded. Ranked by Edge Gap, the top two went 1-for-2 and the bottom three went 2-for-3 — still no separation, but nothing like the clean reversals of Aug 16 and 17. Notably the single biggest edge on the board (Ramos, +28.1) hit, ending a three-day run of the top-ranked leg losing. Across Aug 16–18 the largest edges are now 5-for-13 (38%) and the smallest are 10-for-14 (71%). The gap is narrowing and the conclusion has not changed: we cannot yet say Edge Gap ranking works, and we are no longer able to say it is failing outright either.

The matchup layer had a bad day and is now behind Neutral. Player props only: Confirms went 0-for-1, Neutral went 3-for-4. Running totals since we started recording the verdict: Confirms 3-for-6 (50%), Neutral 11-for-20 (55%). The gate is now marginally worse than no gate. Six legs is not a sample, and we would note that both WNBA Confirms legs on Aug 18 were decided by availability rather than by the matchup — one void, one injury exit at 3:31 of the first quarter — which is a thing a defensive ranking cannot be blamed for.

Playing-time confidence tiers: unchanged at High 2-for-7, Moderate 4-for-4. No football card has run since Aug 15 and none runs until Aug 22.

August 17 — 5 of 12 (42%)

Two cards, twelve legs, all twelve verified against final box scores. No voids and nothing unverified. The running number moves from 56/98 to 61/110, which drops the all-time rate from 57% to 55%. This was our worst graded day.

MLB Edge Gap card — 3 of 5

MLB game-lines card — 2 of 7

The matchup layer had a flat day. Player props only: Confirms went 2-for-4, Neutral went 3-for-6 — 50% either way, no separation at all. Running totals since we started recording the verdict: Confirms 3-for-5 (60%), Neutral 8-for-16 (50%). Five legs is not a sample. What the layer did do well on August 17 is documented on the cards themselves: it removed a Brandon Young strikeout Over into the lowest-strikeout lineup in baseball, six Dodgers Unders at Coors, and a Brandon Lowe leg on handedness. We have not gone back to grade those drops and we should — that is the honest test of a gate, and it is on our list.

August 16 — 6 of 11 (55%)

Two cards, eleven legs, all eleven verified against final box scores. No voids and nothing unverified. The running number moves from 50/87 to 56/98, which rounds to 57% either way.

MLB — 5 of 9

WNBA — 1 of 2

Is the method actually working? (written August 17, on August 16 data)

Three findings, and only one of them is good. Everything in this section is superseded in part by the August 17 results above — the Edge Gap inversion described here happened again the following day, and the encouraging note about the matchup layer did not hold. We are leaving the section as written rather than quietly editing it, because a track record that revises its own reasoning after the fact is not a track record.

The matchup layer had its first clearly good day. On August 16 it approved one leg (Caitlin Clark, Confirms) and that leg won. More usefully, it forced three drops, and we checked all three afterwards: Ezi Magbegor Under 13.5 Pts+Reb would have lost (21), Jade Melbourne Under 7.5 Reb+Ast would have lost (11), and Megan DiLeo Under 1.5 Assists would have won (1). Two saves, one bad drop. Running verdict record: Confirms 1-for-1 on August 16, Neutral 5-for-10. Prior to this the layer was 0-for-3 on its drops. It is now roughly break-even and still nowhere near a sample.

Playing-time confidence tiers remain inverted. Across August 13, 14 and 15 the football pick labelled High confidence lost every time and the Moderate pick won every time — cumulatively High 2-for-7 (29%), Moderate 4-for-4 (100%). No football card has run since, so this number is unchanged. We read "High confidence" as how much research went in, not how likely it is to win, and we are treating the label as a defect rather than as variance.

Known corrections

September 14 — the counter was resynced to the legs, and it went down. Until today the headline counter read 373 legs across 60 cards. That number could not be reproduced from the graded legs in our own source files, which carry 341 graded legs (175 hits, 51%) across 40 cards, plus 11 voids and 2 unverified. The counter has been reset to the number we can rebuild leg by leg. The gap is a bookkeeping failure on our side, not a change to any result: six cards that are written up on this page in full — September 6, 9 and 10, and the MLB cards around them — were graded in prose but never had their per-leg results written back into the source files, so they do not yet count toward a number that is computed from those files. Back-filling them is the next run's first job, and when it lands the counter will go up for that reason and we will say so here. We would rather publish a smaller number we can defend than a larger one we cannot.

Implied probability was being calculated by hand and was wrong on six of eight legs on August 16. A −185 line implies 64.9%, not 58%. That inflated Edge Gap by up to 6.9 points and changed the ranking order. From the August 16 rebuild onward, implied probability and Edge Gap are computed in code from the odds, so the error cannot recur. The Edge Gap figures in the August 16 table above are the corrected ones.