LONGSHOT LEDGER
Circa Survivor · 2026 · Private Updated Wed, Sep 30 · 10:35 AM PT
Week 4 · Locks Saturday, Oct 3

Week 4: 2 of 5 entries still standing.

One team to win outright each week, each team usable once, ties lose. A 25,017-entry, $25,017,000 winner-take-most pool — and a model that published its season targets before kickoff. This page is the running scoreboard of how those predictions hold up.

2 / 5
Our entries alive
6,292
of 24,999 field entries still alive
$25.0M
Prize pool
0%
Chance of a quiet season — massacre weeks are coming

This week

Week 4 — picks shown before Saturday's lock. Spectators only, play nice.

The field walks into a crowded trade

About 58% of the field is projected on the Ravens at home against the Titans, 86% favorites. That's a strong pick with a very crowded exit.

Our two entries split across two doors: Seahawks, Ravens. Seahawks is the leverage play — only 4% of the field is there, so a win buys real separation from the crowd.

Pre-registered prediction №4: 57.7% of the field picks the Ravens this week. Graded the moment Circa publishes the official Week 4 pick report.

Where the field is going

Model-projected share of all 25,017 entries, week 4
Our picks
Rest of the field
BAL
OURS
57.7%
MIN
29.2%
SEA
OURS
3.7%
BUF
1.7%
IND
1.7%
CHI
1.4%
GB
1.2%

Our entries

Four run by the model. One decided by the group chat — same scoreboard, no mercy.
DOC BROWN 1
OUT
Week pick
—
Teams burned: DET, TB
DOC BROWN 2
OUT
Week pick
—
Teams burned: JAX, BAL
DOC BROWN 3
OUT
Week pick
—
Teams burned: JAX, TB
DOC BROWN 4
LIVE
Week pick
SEA · 77% to win
Teams burned: PHI, KC, SF
MCMAHON & RODGERS
LIVEGROUP-DECIDED
Week pick
BAL · vote closes Sat
Model suggests BAL — the group has the wheel

Season outlook

What the simulator sees from here — thousands of full-season runs, refreshed three times a week.
How this likely ends
~10 entries
out of 25,017 are expected to split the pot when the music stops. That's the game: outlast 25,007 strangers.
Model average across simulated seasons; the swing is enormous by design.

Danger weeks, called in advance

Where the model says the field gets hurt — judge us on these in December
WEEK 5
The Bengals trap. The model projects 38% of the surviving field onto one team — with a 23% chance that team loses. That's how thousands of entries die in one afternoon.
THANKSGIVING
Short menu, forced hands. A restricted holiday slate with everyone's best teams already spent; the model has 45% of the field massing on the Rams. First of two holiday choke points.
CHRISTMAS
The season's biggest crash risk. By Christmas the menus are thin — 45% of the field funnels onto the Bears, and the model gives that favorite a 32% chance of losing. A coin flip carrying a huge slice of the pool.

The machine's record

The scorecard we asked to be judged on — four targets, pre-registered before leg 1, graded all season even after our entries die.
Crowd prediction error
1.19%
Target: ≤ 2.65% · 3 weeks graded. On the three most-picked teams: 7.0 pts avg. miss (model's norm ~11)
Our survival vs the field
+11.9 pp
Ours 75%, field 63% over 3 weeks — target: beat the field
Danger weeks called
3 flagged
Week 5, Thanksgiving & Christmas — called before kickoff
Operations
—
A monitoring check needs attention; details stay internal