# David Harrison's 52.02 Week, Built Almost Entirely on the 2026 Draft | MissedTakes Weekly

*Published 2026-08-18 — MissedTakes*

Week 2026-W33 was the week the grader emptied its 2026 draft backlog. Of the 4,044 takes graded in the last seven days, 1,611 were mock-draft picks — nearly 40% of the week's output in one prediction type. That shaped both awards.

## The awards

**Pundit of the Week (MissedTakes Score): [David Harrison](/pundits/david-harrison), 52.02.** Sample: 127 graded takes across 19 articles.

**Pundit of the Week (Accuracy): [Gregg Rosenthal](/pundits/gregg-rosenthal), 93.1.** Sample: 8 graded takes across 3 articles.

Harrison also sits first on the [overall leaderboard](/leaderboard) at a 54.76 MissedTakes Score and 71.1 accuracy — the same 127 graded takes, which the site still flags as limited confidence. The score is mean skill against a naive baseline, where 0 means no better than the naive guess; the [methodology](/methodology) spells out the baselines.

![Pundit of the Week, 2026-W33](/posts/visuals/6a20a03a-12bd-4351-8ef1-0c5896474cb7.png)

## What Harrison's week actually consisted of

Almost all of it is draft work, and almost all of it landed on August 14 and 15. His record splits into 120 graded mock picks (65.62 accuracy, 43.32 score), 4 position-order calls (100 accuracy, 82.38 score), 2 Super Bowl picks (100/100) and one team draft-slot call.

The single best-scoring document is [his only full first-round mock of the cycle](/opinions/5cc44fc1-4ff9-447a-9c32-5f715fa6e9ba), published April 23:

- [Fernando Mendoza](/nfl/players/fernando-mendoza) first overall to the [Raiders](/nfl/teams/raiders) — exact, 100.
- [Olaivavega Ioane](/nfl/players/olaivavega-ioane) to the [Ravens](/nfl/teams/ravens) at 14 — exact, 100.
- [Jeremiyah Love](/nfl/players/jeremiyah-love) at 4 (went 3), [David Bailey](/nfl/players/david-bailey) at 3 (went 2), [Caleb Downs](/nfl/players/caleb-downs) at 10 (went 11), [Caleb Lomu](/nfl/players/caleb-lomu) at 27 (went 28) — each 95.76.

The position-order calls are where the skill numbers get large, because naming one player out of a 20- or 36-man class at random scores badly by default. [Mansoor Delane](/nfl/players/mansoor-delane) as CB1 was worth +87.36 against a 12.64 baseline; [Jordyn Tyson](/nfl/players/jordyn-tyson) as the second receiver off the board, +88.55 against 11.45; [Spencer Fano](/nfl/players/spencer-fano) as the first tackle, +82.95. Add two graded Super Bowl calls for the [Seahawks](/nfl/teams/seahawks) — "If my money is good, then it's the Seattle Seahawks" — each 100 accuracy and +100 skill, and the arithmetic isn't mysterious.

### And the part the accuracy column hated

Harrison's worst grades are all Day 3 Commanders mocks of players who never got drafted. [Roman Hemby](/nfl/players/roman-hemby) at 223 was graded 0 three separate times; so were [Vincent Anthony](/nfl/players/vincent-anthony), [Cade McDonald](/nfl/players/cade-mcdonald), [Diego Pounds](/nfl/players/diego-pounds) and [Jermott McCoy](/nfl/players/jermott-mccoy), the last of them mocked to Washington at 18. [Kaelon Black](/nfl/players/kaelon-black) at 223 (actual 90) scored 0.31; [Cade Klubnik](/nfl/players/cade-klubnik) at 187 (actual 110) scored 3.56.

Notice what those rows did to his *score*, though: the Hemby zero still carried a +50 skill delta, because the ordering of the mock it sat inside graded 100. That gap is the honest caveat on a 52.02 week. Harrison's accuracy of 71.1 is the number that absorbs the whiffs; the score largely doesn't.

That pattern isn't personal to him. The site's [standing mock-draft study](/nfl) has 2,072 graded picks: 76.6 average accuracy on picks mocked in round 1, and 26.9 on picks mocked at 176 or later, where average miss distance is nearly 50 slots.

![Mock-draft accuracy by the round a player was mocked in](/posts/visuals/751d11f1-81fc-4c12-968a-1f12525c6317.png)

## Rosenthal's 93.1, with the sample-size warning attached

Eight graded takes is one podcast segment's worth of claims, and a weekly accuracy leader drawn from eight rows is close to noise. For scale: Rosenthal's standing record is 325 graded takes at 59.85 accuracy and a 17.65 score — established confidence, and 33 points below his award week. His strongest types are mock picks (73.77 on 108) and depth-chart calls (92.57 on 19, though at a score of −2.82, because "the starter stays the starter" is what the naive baseline already says). His weakest: conference champions (25 on 10), division winners (0 on 3).

The same seven-day results window also lists a Rosenthal item among the week's worst: [his Week 1 slate](/opinions/9443503e-89b7-461f-8b67-e99f60c8b5d2) — Giants at Washington, Lions over Packers, Jaguars over Panthers — graded 0 accuracy and −200 skill on game outcomes. Both things are in the record.

## The rest of the week

Beyond the mock picks: 622 game-outcome picks, 351 division winners, 296 sleeper calls, 187 conference champions, 124 playoff picks, 117 position-order calls, 110 depth-chart calls. 1,864 new takes came in.

![What got graded in the last seven days](/posts/visuals/f9725618-74f7-4b76-a314-289fc3d040f6.png)

The week's extremes came from the same few documents in both directions. [Andrew Erickson](/pundits/andrew-erickson) appears in the best list at 100/+200 on fantasy bust calls and in the worst at 0/−200 on sleeper calls, from [one article](/opinions/3809e587-6fa2-464b-ab98-fd96d353c1bd). [Nick Wright](/pundits/nick-wright) and [John Middlekauff](/pundits/john-middlekauff) posted 100s; [Jason McIntyre](/pundits/jason-mcintyre) (Chiefs over Chargers) and [Cynthia Frelund](/pundits/cynthia-frelund) took −200s on game outcomes.

On the consensus side, nothing new resolved, but the standing 2025 misses remain instructive: the field was 6.3 wins low on Kansas City across 13 calls and 5 wins high on New England, Seattle and Denver. And on 101 MVP ballots, 27 had Matthew Stafford — the winner — which is as close to a collective hit as that market gets.