# Cutdown Week Floods the Queue, Nothing Grades, and the Win-Total Field's One-Way Misses | MissedTakes Daily

*Published 2026-08-30 — MissedTakes*

Four hundred and thirty-two new claims landed on the site in the last seven days. Zero of them were graded, and neither was anything else: the results ledger for the week is empty. That is not a bug, it is the calendar. Late August is when pundits talk most and settle least.

![The week in one line](/posts/visuals/82697b4a-430e-4499-a95d-fec14c9307f3.png)

## What the new pile is made of

The 25 most recently logged pending takes are all dated August 28, and every one is either a roster-spot call or a depth-chart call — Jeff Kerr sorting the [Philadelphia Eagles](/nfl/teams/eagles) rookie class by cut risk, Alain Poupart declaring Malik Willis "the unquestioned starter" in [Miami](/nfl/teams/dolphins), Shanna McCarriston leaning veteran on the [Raiders](/nfl/teams/raiders) quarterback job ("Cousins hasn't struggled enough to hand the rookie the" job, in the excerpt we hold on [Kirk Cousins](/nfl/players/kirk-cousins)).

That is worth knowing because of how those two categories grade. Across the whole database, [depth-chart calls](/predictions/depth-chart) average 93.5 accuracy — the highest of any type with a real sample — and a skill delta of just +1.82. [Roster-spot calls](/predictions/roster-spot) average 81.6 accuracy and a skill delta of **-4.08**, meaning the graded ones have collectively come in slightly *worse* than the naive baseline despite looking like near-misses on the accuracy scale. Saying the starter will start is correct and costs nothing.

![Accuracy is not skill](/posts/visuals/138200e1-4dc4-41eb-b2f2-0fcb4085f848.png)

Compare that to [mock draft picks](/predictions/nfl-draft-position) (64.4 accuracy, +36.09 skill) or [position order calls](/predictions/draft-positional-order) (76.9, +56.36), where being right is genuinely hard. High accuracy is not the same as saying something.

There is also a backlog problem worth stating plainly. We hold 3,438 depth-chart takes and have graded 415 of them. We hold 1,888 roster-spot takes and have graded 49. Total pending across all types: 12,373. Cutdown weekend will settle some of the roster calls; the depth-chart claims wait on a season of usage, which is the right way to settle them and also the slow way.

## Nobody changed their mind

No reversals were flagged this week — no pundit on record moving off a previously logged position. In a week with 432 new claims and no results to react to, that is what you would expect. Minds change when the scoreboard argues with them.

## Consensus: 2025's misses only went one direction

The graded season is still 2025, and the [win-total consensus](/predictions/season-win-total) there has a shape. The field missed by more than four wins on six teams, and at the extremes it missed unanimously, not on average.

New England won 14. Eleven pundits put a number on the Patriots and the highest was 10 — the field was 5 wins light with no dissenter. Seattle won 14; seven calls, highest 11. Denver won 14; six calls, highest 11. Jacksonville won 13; ten calls, highest 11. Four teams reached 13 wins and not one of the 34 numbers logged against them cleared 11.

The other direction is Kansas City, where the field was 6.3 wins too high across 13 calls against an actual 6. The *lowest* call on the board was 10. Rich Eisen alone accounts for five of those 13 numbers, ranging from 10 to 15.

![2025 win totals: where the field was furthest off](/posts/visuals/b6ed86a6-1767-46d7-a201-308cb7f66cec.png)

Season win totals grade at 49.2 accuracy site-wide with a +6.6 skill delta — barely better than the naive number. The 2025 board is why.

On the awards, one line each: the MVP field got it right, with Matthew Stafford the plurality pick at 26 of 100 ballots logged and the actual winner; Offensive Rookie of the Year was a coin flip the field half-won, with Tetairoa McMillan tied for the most picks (5 of 29) alongside Ashton Jeanty.

## The board

[David Harrison](/pundits/david-harrison) still leads at a 48.77 MissedTakes Score on 131 graded takes, but nine of the top ten are flagged limited confidence. The exception is [Daniel Jeremiah](/pundits/daniel-jeremiah) at ninth — 23.05, 65.23 accuracy, 277 graded takes, the only established record up there. Nothing moved this week, because nothing graded. Full standings on the [leaderboard](/leaderboard), and the [methodology](/methodology) explains why accuracy and skill disagree as often as they do.