How the published grades have done · 2026-07-29 – 2026-10-28
How to read this. One row per game and market — the final pre-pitch grade if there was one, otherwise the first. Every grade is frozen before first pitch and settled from the box score, and none is ever removed. Win % leaves pushes out; rows under 60 games are marked insufficient sample.
THE PICKS· every A, B and C grade, by market
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total | 352 | 179 | 154 | 19 | 53.8% | |
| A grades | 58 | 30 | 26 | 2 | 53.6% | insufficient sample |
| B grades | 114 | 56 | 53 | 5 | 51.4% | |
| C grades | 180 | 93 | 75 | 12 | 55.4% | |
| First-5-innings total | 411 | 203 | 184 | 24 | 52.5% | |
| A grades | 48 | 22 | 23 | 3 | 48.9% | insufficient sample |
| B grades | 151 | 82 | 61 | 8 | 57.3% | |
| C grades | 212 | 99 | 100 | 13 | 49.7% | |
| All graded picks | 763 | 382 | 338 | 43 | 53.1% | |
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| No lean — the market's right | 955 | 461 | 435 | 59 | 51.5% | |
Picks are every card where the model claimed an edge. No lean is every card where it agreed with the market — not a track record, since nothing there was ever a recommendation: its W and L are against the model's nominal side, bookkeeping, not a call.
BY MODEL· each model's own picks, newest first
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| v8 — the October rebuild (current) · Oct 7 – Oct 28 | 15 | 8 | 5 | 2 | 61.5% | insufficient sample |
| Full-game total | 7 | 4 | 2 | 1 | 66.7% | insufficient sample |
| First-5-innings total | 8 | 4 | 3 | 1 | 57.1% | insufficient sample |
| v5 — the First-5 umpire fix · Sep 4 – Oct 7 | 311 | 153 | 139 | 19 | 52.4% | |
| Full-game total | 163 | 74 | 78 | 11 | 48.7% | |
| First-5-innings total | 148 | 79 | 61 | 8 | 56.4% | |
| v2 — the August refit · Aug 8 – Sep 3 | 351 | 182 | 152 | 17 | 54.5% | |
| Full-game total | 160 | 93 | 61 | 6 | 60.4% | |
| First-5-innings total | 191 | 89 | 91 | 11 | 49.4% | |
| v1 — the first model · Jul 29 – Aug 7 | 86 | 39 | 42 | 5 | 48.1% | |
| Full-game total | 22 | 8 | 13 | 1 | 38.1% | insufficient sample |
| First-5-innings total | 64 | 31 | 29 | 4 | 51.7% | |
Each model's own picks, over the slates it was the one publishing — added together they are the season table above. A new model starts from zero, so read its sample size before its win rate. Shadow grades never appear on this page.
AGAINST THE CLOSE· the morning grade's line vs the same book's close
Did the line move our way? Each pick's line at the morning grade against the same book's close; a close from another book is not a move and is left unpriced. Beat % counts only the lines that moved, and avg move is runs toward the pick — the two markets are never added together.
| SEGMENT | N | BEAT | FLAT | WORSE | BEAT % | AVG MOVE |
|---|
| Full-game total | 331 | 43 | 219 | 69 | 38.4% | -0.05 runs27 unpriced |
| First-5-innings total | 370 | 56 | 264 | 50 | 52.8% | +0.01 runs40 unpriced |
MORE DETAIL· each model by band · if C were dropped · the close by band · over and under · starting pitchers · data gaps
EACH MODEL BY BAND· the model table, split by band
v8 — the October rebuild (current) · Oct 7 – Oct 28
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · B | 4 | 2 | 1 | 1 | 66.7% | insufficient sample |
| Full-game total · C | 3 | 2 | 1 | 0 | 66.7% | insufficient sample |
| First-5-innings total · A | 1 | 1 | 0 | 0 | 100.0% | insufficient sample |
| First-5-innings total · B | 2 | 1 | 1 | 0 | 50.0% | insufficient sample |
| First-5-innings total · C | 5 | 2 | 2 | 1 | 50.0% | insufficient sample |
| All graded picks | 15 | 8 | 5 | 2 | 61.5% | insufficient sample |
v5 — the First-5 umpire fix · Sep 4 – Oct 7
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · A | 28 | 14 | 13 | 1 | 51.9% | insufficient sample |
| Full-game total · B | 50 | 23 | 26 | 1 | 46.9% | insufficient sample |
| Full-game total · C | 85 | 37 | 39 | 9 | 48.7% | |
| First-5-innings total · A | 16 | 7 | 8 | 1 | 46.7% | insufficient sample |
| First-5-innings total · B | 55 | 27 | 24 | 4 | 52.9% | insufficient sample |
| First-5-innings total · C | 77 | 45 | 29 | 3 | 60.8% | |
| All graded picks | 311 | 153 | 139 | 19 | 52.4% | |
v2 — the August refit · Aug 8 – Sep 3
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · A | 22 | 13 | 8 | 1 | 61.9% | insufficient sample |
| Full-game total · B | 46 | 26 | 18 | 2 | 59.1% | insufficient sample |
| Full-game total · C | 92 | 54 | 35 | 3 | 60.7% | |
| First-5-innings total · A | 25 | 10 | 13 | 2 | 43.5% | insufficient sample |
| First-5-innings total · B | 67 | 38 | 26 | 3 | 59.4% | |
| First-5-innings total · C | 99 | 41 | 52 | 6 | 44.1% | |
| All graded picks | 351 | 182 | 152 | 17 | 54.5% | |
v1 — the first model · Jul 29 – Aug 7
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · A | 8 | 3 | 5 | 0 | 37.5% | insufficient sample |
| Full-game total · B | 14 | 5 | 8 | 1 | 38.5% | insufficient sample |
| First-5-innings total · A | 6 | 4 | 2 | 0 | 66.7% | insufficient sample |
| First-5-innings total · B | 27 | 16 | 10 | 1 | 61.5% | insufficient sample |
| First-5-innings total · C | 31 | 11 | 17 | 3 | 39.3% | insufficient sample |
| All graded picks | 86 | 39 | 42 | 5 | 48.1% | |
IF C WERE DROPPED· A and B only, per market
The open band question. The A and B rows added together, market by market — the record if C stopped publishing. Nothing is removed: C still publishes and keeps its own rows in the pick table.
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · AB combined — C set aside | 172 | 86 | 79 | 7 | 52.1% | |
| First-5-innings total · AB combined — C set aside | 199 | 104 | 84 | 11 | 55.3% | |
| All picks · AB combined — C set aside | 371 | 190 | 163 | 18 | 53.8% | |
AGAINST THE CLOSE — BY BAND· and the no-lean cards
By morning band, which can differ from the final band, so these rows do not add up to the pick tables and are not meant to. Measured from the final grade instead — written about an hour before first pitch, with the close captured 35 minutes before — 770 of 820 full-game total picks and 810 of 855 first-5-innings total picks closed at the very same line, so that anchor can only say the line stops moving once lineups are in. The morning grade is the one with hours to move.
| SEGMENT | N | BEAT | FLAT | WORSE | BEAT % | AVG MOVE |
|---|
| Full-game total · A | 46 | 8 | 27 | 11 | 42.1% | -0.04 runsinsufficient sample7 unpriced |
| Full-game total · B | 94 | 9 | 63 | 22 | 29.0% | -0.11 runs14 unpriced |
| Full-game total · C | 191 | 26 | 129 | 36 | 41.9% | -0.03 runs6 unpriced |
| Full-game total · all graded picks | 331 | 43 | 219 | 69 | 38.4% | -0.05 runs27 unpriced |
| First-5-innings total · A | 57 | 6 | 39 | 12 | 33.3% | -0.05 runsinsufficient sample8 unpriced |
| First-5-innings total · B | 101 | 16 | 74 | 11 | 59.3% | +0.03 runs13 unpriced |
| First-5-innings total · C | 212 | 34 | 151 | 27 | 55.7% | +0.02 runs19 unpriced |
| First-5-innings total · all graded picks | 370 | 56 | 264 | 50 | 52.8% | +0.01 runs40 unpriced |
| SEGMENT | N | BEAT | FLAT | WORSE | BEAT % | AVG MOVE |
|---|
| Full-game total · no lean | 437 | 77 | 275 | 85 | 47.5% | -0.01 runs65 unpriced |
| First-5-innings total · no lean | 407 | 42 | 314 | 51 | 45.2% | -0.01 runs43 unpriced |
OVER VS UNDER — PICKS· every graded pick, by market
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · Over | 165 | 91 | 64 | 10 | 58.7% | |
| Full-game total · Under | 187 | 88 | 90 | 9 | 49.4% | |
| First-5-innings total · Over | 315 | 150 | 145 | 20 | 50.8% | |
| First-5-innings total · Under | 96 | 53 | 39 | 4 | 57.6% | |
OVER VS UNDER — A AND B ONLY· C set aside
The picks the model rates highest — A and B — over and under: the block above minus its C grades.
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · Over | 74 | 34 | 37 | 3 | 47.9% | |
| Full-game total · Under | 98 | 52 | 42 | 4 | 55.3% | |
| First-5-innings total · Over | 156 | 78 | 69 | 9 | 53.1% | |
| First-5-innings total · Under | 43 | 26 | 15 | 2 | 63.4% | insufficient sample |
OVER VS UNDER — NO LEAN· not picks
Not picks. The model's nominal side on cards where it saw no edge — which way it leaned while making no call. It was never a recommendation.
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · Over | 261 | 123 | 125 | 13 | 49.6% | |
| Full-game total · Under | 246 | 118 | 114 | 14 | 50.9% | |
| First-5-innings total · Over | 273 | 136 | 121 | 16 | 52.9% | |
| First-5-innings total · Under | 175 | 84 | 75 | 16 | 52.8% | |
BY STARTING-PITCHING AGREEMENT· did the starters argue for the pick?
A watch, not a rule. On 2021-2024 history, picks the starting pitchers argued for won about 54% and picks they argued against 47-49%. The season-end read found no clear gap on this year's numbers, so it stays a watch. Nothing is filtered on it; a pick whose top five drivers leave out the starters is counted on its own.
| SEGMENT | N | W | L | PUSH | WIN % | NOTES |
|---|
| Full-game total · starters argued for the pick | 203 | 105 | 88 | 10 | 54.4% | |
| Full-game total · starters argued against it | 42 | 21 | 20 | 1 | 51.2% | insufficient sample |
| Full-game total · starters not among the five drivers | 107 | 53 | 46 | 8 | 53.5% | |
| First-5-innings total · starters argued for the pick | 314 | 151 | 146 | 17 | 50.8% | |
| First-5-innings total · starters argued against it | 66 | 34 | 29 | 3 | 54.0% | |
| First-5-innings total · starters not among the five drivers | 31 | 18 | 9 | 4 | 66.7% | insufficient sample |
Data gap · Jul 29 – Jul 31. Graded while the boxscore-context collector was dark — no fielding row had landed since 2026-07-12, so the 30-game fielding fold stayed frozen there. The 2026-08-01 re-parse refilled the fold; 2026-08-01 is the first slate graded against it. Nothing re-grades, so those grades are counted as written.
Data gap · Jul 29 – Aug 2. Graded while the IL collector was dark — no stint had been fetched since 2026-07-07, and a stint that was never fetched reads as no stint, so a starter just back from the injured list scored as healthy. It misread 11 of the 71 games on these dates — 12 starter-sides that should have carried an IL-return flag and did not. The 2026-08-03 repair refilled the stints; 2026-08-03 is the first slate graded against them. Nothing re-grades, so those grades are counted as written.
SERIES_UNDER — THE PERSONAL RULE· not a Spotter grade, and not a bet log
The owner's own rule. When a game finishes at least 4 runs under its closing total, take the under in the same series' next game. The model does not make it and no grade is written for it — the pipeline sends a nightly note and stops.
| POPULATION | N | UNDER | OVER | PUSH | UNDER % | ROI |
|---|
| Every trigger | 68 | 35 | 30 | 3 | 53.8% | +4.6% |
| of which severe — 5+ runs under | 38 | 20 | 15 | 3 | 57.1% | +10.6% |
68 triggers since 2026-08-05 · 68 graded. Each row is the under at that game's closing total, one flat unit at the closing price. They are not bets. Nothing in the bet log is a SERIES_UNDER position. On 65 decided triggers the 95% range around 53.8% runs 42–65% — too wide to tell a coin flip from the screened edge, which took 1,261 triggers to show 54.61% under and +5.63% ROI against a 52.38% break-even. Tracking the screen so far is not evidence the rule works.