GLIZZY GOD dropped 138.24 and made it three straight wins, so Viet has earned the right to be a pain in the ass this week. Cossum won by 71.30 while leaving Kyle Monangai’s 27.50 on the bench, an incredibly rude way to help Burns earn his second shotgun.
Five teams are 3–1, and Mason is still looking for win number one after a team-name change accomplished absolutely fuck all. Huff leads the league in points and has 91.1% playoff odds, because apparently escaping Bergquist by 6.50 was enough to make the computer cocky again.
GLIZZY GOD rolled 138.24–104.02, with CeeDee Lamb dropping 32.80 and Viet collecting his third straight win. AK got the week’s best player, Tetairoa McMillan at 38.20, and still lost by 34.22. Justin Jefferson’s zero and Romeo Doubs’s 20.80 on the bench didn’t help, but even AK’s perfect lineup loses this one.
The Treyliban won 116.40–108.06 behind Tee Higgins’s 21.20 and a lineup with only 2.00 points left to squeeze out. Longoria left Drake Maye’s 26.16 on the bench, apparently committed to making the team name a cry for help. Starting him still wouldn’t have flipped it, so this week the panic is mostly decorative.
drunk drafting smoked Golden Showered in Ws 136.88–65.58, and Bijan Robinson, Javonte Williams and Nico Collins alone combined for 83.30. Cossum even left Kyle Monangai’s 27.50 on the bench and won by more points than Burns actually scored. Burns got outscored by three dudes and still has more wins than Mason.
Monday Night Mustardson rolled 120.28–96.12 with Kenneth Walker III doing 30.40 points of the heavy lifting. Mason renamed his team Turnaround Time and immediately went to 0–4, so the rebrand is working about as well as the lineup. Zay Flowers’s 21.80 deserved a better home than this.
Hit and Rice cruised 110.66–84.72 behind Sam LaPorta’s 18.40 and a 16.00-point kicker, a deeply annoying way for Valentine to reach 3–1. Calvin got 22.20 from Jonathan Taylor, then asked Rashee Rice’s zero to help finish the job. T.J. Hockenson had 18.40 on the bench, although even the perfect lineup still loses.
Gibbs Me Liberty or Death escaped JB Unit 115.38–108.88 thanks to Kyren Williams’s 31.70. Denzel Boston outscored Ja'Marr Chase 10.90–4.20, and swapping them would’ve given Bergquist a 0.20-point win. Nobody’s pretending that was the obvious call before kickoff, but Huff should still send Boston a thank-you card.
Top three in each division make it after Week 17.
| # | Team | Record | PF | Odds* |
|---|---|---|---|---|
| 1 | 3–1 | 483.24 | 86% | |
| 2 | 3–1 | 463.06 | 82% | |
| 3 | 2–2 | 468.94 | 79% | |
| CURRENT PLAYOFF CUT | ||||
| 4 | 2–2 | 298.42 | 6% | |
| 5 | 1–3 | 405.40 | 29% | |
| 6 | 1–3 | 390.74 | 18% | |
| # | Team | Record | PF | Odds* |
|---|---|---|---|---|
| 1 | 3–1 | 508.02 | 91% | |
| 2 | 3–1 | 486.38 | 81% | |
| 3 | 3–1 | 436.36 | 58% | |
| CURRENT PLAYOFF CUT | ||||
| 4 | 2–2 | 418.56 | 40% | |
| 5 | 1–3 | 396.68 | 23% | |
| 6 | 0–4 | 371.42 | 7% | |
The line shows the current top three.
We played out the actual remaining schedule 60,000 times. Results through Week 4 stay fixed. The top three in each division qualify after Week 17. Record comes first, then points for, then head-to-head.
When a complete roster projection is available, each weekly scoring forecast blends 50% roster projection and 50% recent scoring strength. A missing next-week lineup total falls back to recent scoring strength alone. The scoring half uses this season’s results, weighted toward recent weeks, plus two weeks at the 2025 league average. Our estimate of weekly volatility now includes this season too.
Next week uses ESPN’s weekly projections with unavailable players and byes removed. Later weeks use season-outlook averages, known byes and current roster depth. A player marked OUT this week is not automatically out all season. Positive outlooks beyond next week assume availability outside byes, without claiming a return date. IR slots stay excluded. An empty position on the current roster contributes zero before blending.
The real fantasy schedule already accounts for strength of remaining opponents. We don’t add the same schedule adjustment twice. All-play record describes results so far and does not count as extra independent evidence.
V2 is experimental. Future pickups, late injury news and shared NFL game effects remain uncertain. We preserve the forecasts and check them against next week’s results.
Model V2. Current rosters, recent form, and a fresh chance to embarrass the computer.
| Rank | Team | Record | Model Pts / Wk |
|---|---|---|---|
| 01 | Gibbs Me Liberty or Death | 3–1 | 115.39 |
| 02 | GLIZZY GOD | 3–1 | 111.63 |
| 03 | Monday Night Mustardson | 3–1 | 111.32 |
| 04 | drunk drafting | 2–2 | 110.56 |
| 05 | The Treyliban | 3–1 | 108.59 |
| 06 | Hit and Rice | 3–1 | 104.96 |
| 07 | JB Unit | 1–3 | 104.09 |
| 08 | Mayeday, Mayeday, Mayeday | 1–3 | 103.75 |
| 09 | Big Boy 97 | 2–2 | 103.70 |
| 10 | Hoo Kim | 1–3 | 100.12 |
| 11 | Turnaround Time | 0–4 | 99.15 |
| 12 | Golden Showered in Ws | 2–2 | 88.05 |
Model Pts / Wk is the average scoring forecast for the remaining regular-season weeks. Half comes from the current roster’s ESPN projections, half from recent scoring. This year’s results get more weight than before, with a modest boost for recent games.
Next week uses weekly player projections. Later weeks use ESPN’s season-outlook averages and known byes, reselecting the lineup each week. That is a roster forecast, not a prediction of future waiver moves or exact injury return dates.
Fantasy opponents and current record do not set this ranking. They do affect playoff odds. NFL byes affect available scoring. V2 is an experiment, and the weights have not been tuned to make the last three weeks look good.
What if you played everybody each week?
Burns is 2–2 with a 4–40 record against the rest of the league. Two wins from four weeks of this shit is an incredible return on absolutely nothing.
| Team | Record | Vs. Everyone | Win % |
|---|---|---|---|
| 3–1 | 33–11 | 75.0% | |
| 3–1 | 33–11 | 75.0% | |
| 3–1 | 32–12 | 72.7% | |
| 3–1 | 31–13 | 70.5% | |
| 3–1 | 26–18 | 59.1% | |
| 2–2 | 25–19 | 56.8% | |
| 2–2 | 20–24 | 45.5% | |
| 1–3 | 18–26 | 40.9% | |
| 1–3 | 18–26 | 40.9% | |
| 1–3 | 13–31 | 29.5% | |
| 0–4 | 11–33 | 25.0% | |
| 2–2 | 4–40 | 9.1% |
Each score faces all 11 other teams. Through Week 4, that’s 44 comparisons per team. Ties count as half a win for Win %.
Bruderer has faced the lowest-scoring opponents so far. His own 32–12 all-play record says he’s good too, which makes the extra help even more annoying.
Mason still has the toughest scoring schedule. His own 11–33 all-play record suggests the schedule isn’t the only thing that needs fixing.
Compare the actual record with how the same scores would have fared against every opponent. This describes results so far, not future wins. It supplements the scoring-based power rankings without changing the league standings.
The biggest playoff swings, plus a score trend worth watching.
Playoff odds: 71.4% to 85.8%
Three straight wins and the week’s high score. Viet is becoming a problem, and unfortunately the computer agrees.
Playoff odds: 59.1% to 39.8%
Calvin dropped 84.72 and watched his playoff odds fall 19.3 points. That Week 2 high score is starting to feel like a vacation photo.
80.92, then 99.78, then 136.88 over the last three weeks. Cossum remembered where the points button was and took it out on Burns.
Odds movement since last week’s published recap. Same V2 weights, fresh results and roster snapshots. These are estimates, not clinching scenarios.
You drafted the points. Did you actually start them?
Valentine has captured 95.2% of his available lineup points, best in the league. Annoying how much easier this gets when you start the right guys.
AK has left 108.84 points behind across four weeks, most in the league. His bench has practically submitted an application for its own franchise.
| Team | Captured | Points Left |
|---|---|---|
| 95.2% | 22.20 | |
| 94.9% | 24.80 | |
| 91.9% | 41.30 | |
| 90.9% | 51.10 | |
| 90.9% | 48.58 | |
| 90.2% | 52.60 | |
| 88.7% | 51.66 | |
| 85.6% | 70.24 | |
| 84.7% | 71.80 | |
| 83.3% | 74.20 | |
| 78.2% | 108.84 | |
| 76.7% | 90.90 |
Captured = total starting points divided by the total highest-scoring lineup available on each week’s roster. Points Left is the difference. This is hindsight, excludes IR, and may require multiple swaps. It doesn’t mean those decisions were obvious before kickoff.
Right is more points for you. Up is more points for the other guy.
Averages through Week 4. Dashed lines = league average (106.82).
| Team | Avg. scored | Avg. opponent |
|---|---|---|
| 127.00 | 97.39 | |
| 121.59 | 92.05 | |
| 120.81 | 106.97 | |
| 117.23 | 106.64 | |
| 115.77 | 104.76 | |
| 109.09 | 104.11 | |
| 104.64 | 114.92 | |
| 101.35 | 100.54 | |
| 99.17 | 117.41 | |
| 97.69 | 112.45 | |
| 92.86 | 128.47 | |
| 74.61 | 96.09 |
League rank, week by week
Record, points for, then head-to-head.
COSS 136.88 · BRNS 65.58
DAWG
AK
Points From One Swap
BRNS
COSS
Largest margin of victory
Burns, 65.58 gets you shotgun number two and sole possession of the league lead nobody wanted. Brian Robinson Jr. put 25.20 on your bench while Saquon gave you 1.50, so grab a beer and explain why the points keep ending up in the wrong fucking place.
The count sticks. Sending the video clears what you owe.
AK — 1 owedWeeks 3
Burns — 1 owedWeeks 4
It roasted your lineup. Now we check its homework.
Five right, one wrong. V1 and V2 finally earned their beer, although they picked the exact same six favorites. V2 missed team scores by 12.17 points on average against V1’s 12.83, but V1 had better probability scores. A mixed first test, so nobody gets to call the new model a genius yet.
Original V1 record: 4–8 before Week 4, now 9–9 across Weeks 2–4. V2’s separate prospective record begins in Week 4. No earlier picks are credited to V2.
| Pick | Win Odds | Winner | Result |
|---|---|---|---|
| 60.0% | Hit | ||
| 51.9% | Hit | ||
| 70.9% | Hit | ||
| 59.4% | Hit | ||
| 50.8% | Miss | ||
| 68.4% | Hit |
Week 4. Same games, three forecasts, nowhere to hide.
| Model | Picks | Avg. Score Error | Brier |
|---|---|---|---|
| V1 | 5/6 | 12.83 | 0.1572 |
| V2 | 5/6 | 12.17 | 0.1665 |
| ESPN Only | 4/6 | 13.10 | 0.2081 |
Lower score error and Brier are better. Six games are a preliminary check, not proof of lasting improvement.
Average score error is MAE in fantasy points. RMSE gives larger score misses more weight. Brier and log loss measure probability accuracy. Picks exclude 50/50 forecasts and tied games. ESPN Only uses our probability calculation from optimized ESPN lineup totals, not ESPN’s win percentages.
V1: RMSE 16.70 · Log loss 0.5024 · 6 games and 12 team scores.
V2: RMSE 16.14 · Log loss 0.5201 · 6 games and 12 team scores.
ESPN Only: RMSE 16.86 · Log loss 0.6052 · 6 games and 12 team scores.
Week 5 picks are saved. Same weights, another test. Nobody gets to edit their picks on Monday.
| Game | V1 | V2 | ESPN Only |
|---|---|---|---|
| Matt v. DAWG | DAWG · 61.5% | DAWG · 56.4% | Matt · 53.8% |
| AK v. COSS | COSS · 61.3% | COSS · 65.8% | Unavailable |
| GFY v. BRNS | GFY · 72.4% | GFY · 79.3% | GFY · 78.9% |
| URQ v. MAGA | MAGA · 60.1% | MAGA · 54.7% | URQ · 54.6% |
| MKM v. UNIT | UNIT · 53.6% | UNIT · 56.6% | UNIT · 56.3% |
| HR v. HUFF | HUFF · 60.6% | HUFF · 63.3% | HUFF · 63.5% |
Favored team and win odds. V1 is the old model. V2 adds roster forecasts and recent form. ESPN Only uses the optimized lineup totals with our uncertainty model, not ESPN’s own win percentages.
Next Tuesday we’ll compare score errors, pick accuracy, Brier score and log loss using these saved forecasts. Lower errors are better. Six games are a first check, not proof that the new model is better.
ESPN’s best projected lineups. Our model’s picks. Plenty to argue about.
AK has no available D/ST because Kansas City is on bye. COSS has no available kicker because Harrison Butker is on bye. Their full ESPN lineup totals are unavailable. For this matchup, the unchanged V2 fallback uses recent scoring strength alone. Future pickups are not assumed.