Model output, not advice.

More pick tools

Model predictions Book feed as of Wed, Sep 23, 2:07 AM EDT

Picks · this week

Picks: one play, the price, the heat — and why

The slate is a heat map of calibrated confidence — glow is earned by the model's gate, not marketing. Each play opens into its attribution: which lenses fired, what the sharp money did, where the key numbers sit.

This week’s picks · NFL 2026 Week 3

16 house plays across 16 games · first kickoff in 1 day

as of Sep 23, 2:23 AM ET · n=16

This week’s picks · NFL 2026 Week 3 · 16 plays above the floor, 0 NO PLAYs shown so you know we looked

The side the model favors and how far past a coin flip it leans (bar from 50, the 0.55 play floor ticked), the matchup's side/total/top-prop reads · best captured price · published edge · the side each lens likes on this line

? · what this block is

tonight table

What

One matchup card per game on the slate, three stacked facts (Grant 2026-09-04): the side the model likes and how far past a coin flip it leans, its total read the same way, and up to 3 player-prop reads ranked by heat — plus the house play, best captured price, published edge and per-lens reads underneath.

Why it is here

The whole slate in one glance, with the NO PLAYs shown so you know we looked.

Example (illustrative rows, not measured)

Play	Game	Side	Total	Top props	Best price	Edge	Lens reads
NE −2.5	SEA @ NE · Thu 8:20 PM ET	0.62 · NE favored ▮▮▮	0.55 · Over favored ▮	Metcalf Rec yds O68.5 · 0.66	−108 DK	+2.1%	Divisional Dog NE ✓ · 360-314 · Rest Edge NE ✓ · 208-206 · +1
OVER 44.5	KC @ LAC · Sun 4:25 PM ET	no side published yet	0.57 · Over favored ▮	(no props block: no prop confidence published yet)	−105 Pinnacle	—	no graded lens reads this game

Honesty

Every number is read from the picks and odds tables; an uncaptured price prints a dash and is never filled with −110; a market with no published pick for a game prints an honest "no side/total published yet" instead of a fabricated 50/50 read; props is [] and the whole props block does not render when the sport has no prop confidence yet — never a resting "no reads" line (Grant: cut empty widgets).

Source

picks + odds_snapshots through the /picks loader (projectTonightTable)

Checked live

rows_match_countsprice_names_its_sourceheat_and_edge_in_rangeedge_absence_is_explainedfavored_side_matches_decisiongraded_voices_are_well_formedbar_ends_name_both_sidesgame_ids_uniquematchup_fact_honestymatchup_props_bounded

Play	Model	Best price	Edge	Lens reads
ATL +6.5	0.80 ATL favored ATL GB Side ATL GB 0.80 · ATL favored Total UND OVR 0.60 · Under favored	−105ticket	+11.9	Primetime Dog ATL ✓ · 221-196 · ATS Streak Regression GB ✗ · 410-388 · Line Value ATL ✓ · 389-358 Primetime Dog ATL ✓ · 221-196 ATS Streak Regression GB ✗ · 410-388 Line Value ATL ✓ · 389-358
NYG +5.5	0.66 NYG favored TEN NYG Side TEN NYG 0.66 · NYG favored Total UND OVR 0.71 · Over favored	−110ticket	+8.3	ATS Streak Regression TEN ✗ · 410-388 · Line Value TEN ✗ · 389-358 ATS Streak Regression TEN ✗ · 410-388 Line Value TEN ✗ · 389-358
CLE −2.5	0.79 CLE favored CAR CLE Side CAR CLE 0.79 · CLE favored Total UND OVR 0.61 · Over favored	−110ticket	+7.9	Line Value CLE ✓ · 389-358 · ATS Streak Regression CAR ✗ · 410-388 Line Value CLE ✓ · 389-358 ATS Streak Regression CAR ✗ · 410-388
DET +7	0.78 DET favored NYJ DET Side NYJ DET 0.78 · DET favored Total UND OVR 0.57 · Over favored	−110ticket	+1.3	Rest Edge DET ✓ · 212-206 · Line Value NYJ ✗ · 389-358 · ATS Streak Regression NYJ ✗ · 410-388 Rest Edge DET ✓ · 212-206 Line Value NYJ ✗ · 389-358 ATS Streak Regression NYJ ✗ · 410-388
MIA −8.5	0.90 MIA favored KC MIA Side KC MIA 0.90 · MIA favored Total UND OVR 0.61 · Under favored	−110ticket	+15.3	Line Value MIA ✓ · 389-358 · ATS Streak Regression MIA ✓ · 410-388 Line Value MIA ✓ · 389-358 ATS Streak Regression MIA ✓ · 410-388
SEA −2.5	0.86 SEA favored SEA WAS Side SEA WAS 0.86 · SEA favored Total UND OVR 0.77 · Over favored	−110ticket	+10.5	ATS Streak Regression WAS ✗ · 410-388 · Line Value WAS ✗ · 389-358 ATS Streak Regression WAS ✗ · 410-388 Line Value WAS ✗ · 389-358
BUF +4.5	0.58 BUF favored LAC BUF Side LAC BUF 0.58 · BUF favored Total UND OVR 0.72 · Under favored	−110ticket	+7.1	Line Value LAC ✗ · 389-358 · Rest Edge BUF ✓ · 212-206 · ATS Streak Regression LAC ✗ · 410-388 Line Value LAC ✗ · 389-358 Rest Edge BUF ✓ · 212-206 ATS Streak Regression LAC ✗ · 410-388
IND −2.5	0.63 IND favored HOU IND Side HOU IND 0.63 · IND favored Total UND OVR 0.69 · Over favored	−110ticket	+1.8	Line Value IND ✓ · 389-358 · ATS Streak Regression HOU ✗ · 410-388 · Divisional Dog IND ✓ · 366-319 Line Value IND ✓ · 389-358 ATS Streak Regression HOU ✗ · 410-388 Divisional Dog IND ✓ · 366-319
PIT −2.5	0.90 PIT favored CIN PIT Side CIN PIT 0.90 · PIT favored Total UND OVR 0.71 · Over favored	−110ticket	+13.4	Divisional Dog PIT ✓ · 366-319 · Line Value PIT ✓ · 389-358 Divisional Dog PIT ✓ · 366-319 Line Value PIT ✓ · 389-358
NE +2.5	0.60 NE favored NE JAX Side NE JAX 0.60 · NE favored Total UND OVR 0.56 · Over favored	−110ticket	−1.8	Line Value NE ✓ · 389-358 Line Value NE ✓ · 389-358
ARI +8.5	0.67 ARI favored ARI SF Side ARI SF 0.67 · ARI favored Total UND OVR 0.61 · Over favored	−106Pinnacle	+12.0	Divisional Dog ARI ✓ · 366-319 · ATS Streak Regression ARI ✓ · 410-388 Divisional Dog ARI ✓ · 366-319 ATS Streak Regression ARI ✓ · 410-388
MIN +1.5	0.67 MIN favored MIN TB Side MIN TB 0.67 · MIN favored Total UND OVR 0.60 · Under favored	−110ticket	+3.9	Line Value TB ✗ · 389-358 Line Value TB ✗ · 389-358
DAL −3	0.75 DAL favored BAL DAL Side BAL DAL 0.75 · DAL favored Total UND OVR 0.60 · Over favored	−110ticket	+5.2	ATS Streak Regression DAL ✓ · 410-388 · Line Value DAL ✓ · 389-358 ATS Streak Regression DAL ✓ · 410-388 Line Value DAL ✓ · 389-358
NO +3	0.74 NO favored LV NO Side LV NO 0.74 · NO favored Total UND OVR 0.73 · Under favored	−110ticket	+6.4	Line Value LV ✗ · 389-358 · ATS Streak Regression LV ✗ · 410-388 Line Value LV ✗ · 389-358 ATS Streak Regression LV ✗ · 410-388
DEN −2.5	0.75 DEN favored LA DEN Side LA DEN 0.75 · DEN favored Total UND OVR 0.59 · Over favored	−110ticket	+8.1	ATS Streak Regression LA ✗ · 410-388 · Line Value DEN ✓ · 389-358 · Primetime Dog DEN ✓ · 221-196 · Home Dog Primetime DEN ✓ · 86-74 ATS Streak Regression LA ✗ · 410-388 Line Value DEN ✓ · 389-358 Primetime Dog DEN ✓ · 221-196 Home Dog Primetime DEN ✓ · 86-74
CHI +1.5	0.64 CHI favored PHI CHI Side PHI CHI 0.64 · CHI favored Total UND OVR 0.55 · Under favored	−105ticket	+4.2	Home Dog Primetime CHI ✓ · 86-74 · ATS Streak Regression PHI ✗ · 410-388 · Line Value CHI ✓ · 389-358 · Primetime Dog CHI ✓ · 221-196 Home Dog Primetime CHI ✓ · 86-74 ATS Streak Regression PHI ✗ · 410-388 Line Value CHI ✓ · 389-358 Primetime Dog CHI ✓ · 221-196

Best price = the best captured book price at the play’s own line, across books (ticket = the price at publish when no book quotes that exact line yet; — = no price at all, never −110). Edge = published edge vs the opening no-vig fair price, pts. Lens reads: each GRADED canonical lens’s side on this line’s market and its win-loss record · ✓ agrees with the house · ✗ opposes · persona voices never render here. n=16 games · as of Sep 23, 2:23 AM ET

Game spotlight · NFL

? · what this block is

game spotlight

What

Everything about one game: the market by book, the player prop lines grouped per player with the model’s read when it has one, the released house pick, and the open → now movement.

Why it is here

A pick is a claim about a game; the spotlight shows the whole game so the claim can be judged.

Example (illustrative rows, not measured)

Player	Market	Line	Model read
D. Maye	Passing yards	245.5	no model read yet
S. Darnold	Passing TDs	1.5	no model read yet

Honesty

Prop lines are the books’ own captured lines and carry no price; a prop without a model read says so instead of drawing a bar; each section degrades on its own and prints its cure.

Source

GET /api/game-spotlight (odds_snapshots, underdog_props_latest, picks, odds_opening_lines)

Checked live

game_identity_presentready_sections_carry_rowsconfidence_in_unit_intervalplayers_unique

Loading upcoming games…

ATL @ GB · the why

? · what this block is

play why

What

The house play with its calibrated confidence against the floor, and one row per lens voice: agrees, opposes or no read, with the stored signal strength.

Why it is here

Attribution, not vibes — which lenses fired and how hard is what makes a play defensible.

Example (illustrative rows, not measured)

Lens	Mark	Strength	Take
Sharp money divergence	✓	84%	Line moved NE −1 → −2.5 against 38% of tickets.
Contrarian	—	no read	Nothing to fade.

Honesty

Strengths are stored lens_signals values printed as-is; a lens without a fired signal prints “no read” and no number; a NO PLAY still lists what was weighed.

Source

picks + lens_signals through projectPlayWhy

Checked live

agree_count_matches_lensesno_read_has_no_strengthheat_in_unit_intervallens_ids_unique

ATL +6.5

calibrated 0.80 vs the 0.55 floor · attribution, not vibes

Fan no read —
Casual bettor no read —
Sharp money 100% ✗
Modeler 100% ✓
Contrarian 100% ✓

2 of 5 voices agree

as of Sep 23, 2:23 AM ET

Where the money is going

? · what this block is

money flow

What

Where the money is going on each game: the side it leans to, tickets versus dollars, the open → now line and a one-word read.

Why it is here

A line that moves against the ticket count is the sharpest public signal there is; the read names it.

Example (illustrative rows, not measured)

Game	Side	Open	Now	Read
SEA @ NE	NE	NE −1	NE −2.5	sharp side · 38% tickets · 61% money
KC @ LAC	KC	KC −3	KC −3	split · 52% tickets · 49% money

Honesty

Percentages are the splits feed rounded to whole numbers; the read is a closed set (public pile-on, sharp side, split, no read) derived only from those splits and the line; an empty slate prints its own empty text.

Source

spread splits + odds_snapshots through projectMoneyFlow

Checked live

percentages_are_integers_in_rangerow_text_presentempty_slate_has_its_textgame_ids_unique

open → now · tickets vs dollars · 10 books

Game	Open	Now	Read
ATL @ GB	ATL +7	ATL +5.5	49% tickets · 59% money
TEN @ NYG	TEN +6	TEN +2.5	56% tickets · 51% money
CAR @ CLE	CAR −2.5	CAR −3	79% tickets · 82% money
NYJ @ DET	DET −6.5	DET −6.5	72% tickets · 63% money
KC @ MIA	KC −10.5	KC −11.5	86% tickets · 93% money
SEA @ WAS	SEA −6.5	SEA −7	86% tickets · 94% money
LAC @ BUF	BUF −7	BUF −7.5	75% tickets · 67% money
HOU @ IND	IND +2.5	IND +2.5	58% tickets · 56% money
CIN @ PIT	CIN −3.5	CIN −3.5	71% tickets · 86% money
NE @ JAX	JAX −3	JAX −3	35% tickets · 51% money
ARI @ SF	SF −8	SF −8.5	76% tickets · 83% money
MIN @ TB	MIN −2.5	MIN −1.5	64% tickets · 57% money
BAL @ DAL	BAL −2.5	BAL −3.5	46% tickets · 77% money
LV @ NO	LV +3	LV +3	57% tickets · 60% money
LA @ DEN	LA −2.5	LA −2.5	60% tickets · 72% money
PHI @ CHI	PHI −4.5	PHI −4.5	64% tickets · 77% money

as of Sep 23, 2:20 AM ET

Steam · live rail

? · what this block is

steam rail

What

Timestamped line moves on the slate, largest first: the game, the market, from → to, the books behind it and the cause when the data names one.

Why it is here

Steam is the market talking; a rail that says who moved and when beats a chart nobody reads.

Example (illustrative rows, not measured)

Time	Game	Move	Books	Cause
3:41 AM ET	SEA @ NE	NE −2 → −2.5	4 books in 90 s	sharp origin
2:58 AM ET	BUF @ MIA	BUF −3 holding at key	6 books	no action

Honesty

Magnitude is the absolute points moved between captures; a cause is a tag from the data or “unknown”, never inferred from prose; moves under the threshold are not shown.

Source

odds_opening_lines vs the latest odds_snapshots capture through projectSteamRail

Checked live

rows_largest_move_firstmagnitude_nonnegative_finiterow_text_presentempty_slate_has_its_text

open → now on the slate, largest first · cause tags arrive with the splits read

2:07 AM ET TEN @ NYG · spread · NYG −6 → −2.5 2 books through a key number
2:07 AM ET ATL @ GB · spread · GB −7 → −5.5 2 books through a key number
2:07 AM ET KC @ MIA · spread · MIA +10.5 → +11.5 2 books cause not tagged yet
2:07 AM ET MIN @ TB · spread · TB +2.5 → +1.5 2 books cause not tagged yet
1:57 AM ET BAL @ DAL · spread · DAL +2.5 → +3.5 22 books through a key number
2:07 AM ET CAR @ CLE · spread · CLE +2.5 → +3 2 books through a key number
2:07 AM ET SEA @ WAS · spread · WAS +6.5 → +7 2 books through a key number
2:07 AM ET LAC @ BUF · spread · BUF −7 → −7.5 2 books through a key number
2:07 AM ET ARI @ SF · spread · SF −8 → −8.5 2 books cause not tagged yet

as of Sep 23, 2:23 AM ET

Book screen · best number lit

every book on the same line — nothing to chase

? · what this block is

book screen

What

One row per house PLAY, one column per book quoting the home spread; the cell that gives the house side the best number is lit. Closed by default: the summary counts the games where books disagree on the line and glows only then.

Why it is here

The same play at a better number is free money; the screen makes the best book obvious.

Example (illustrative rows, not measured)

Game	Play	DK	FD	MGM	Pinny
SEA @ NE	NE −2.5	−2.5	−2.5	−2 ★	−2.5

Honesty

Captured prices only; a book that did not quote a game leaves its cell empty rather than printing a dash as if it were a number; a total or moneyline play lights no cell.

Source

odds_snapshots per book through the /picks loader (TonightGame.lines.books)

Checked live

cells_follow_booksbest_cell_has_a_lineno_best_without_spread_sidegap_rows_counts_rows_with_a_gapgame_ids_unique

2 books, one glance — never take a worse number

Play	Pinnacle	Bovada
ATL @ GB ATL +6.5	−5.5	−5.5

home spread per book · live pregame · n=2 quotes · captured Sep 23, 2:23 AM ET. Totals and moneylines have no home-spread number to light.

Value board · ranked by edge · edge vs the opening no-vig fair price

? · what this block is

edge board

What

Every house PLAY ranked by the published edge over the opening no-vig fair probability, with the captured price for the play’s side.

Why it is here

Edge is the reason to bet; ranking by it puts the strongest claim first.

Example (illustrative rows, not measured)

Game	Play	Edge	Price	Confidence
SEA @ NE	NE −2.5	+3.4%	−108	0.63 vs 0.55 floor
KC @ LAC	KC −3	edge not published	−110	0.58 vs 0.55 floor

Honesty

The edge is read from picks.metadata as published, never recomputed here; a play whose writer published no edge is listed after the ranked rows as “edge not published”, never given 0.

Source

picks.metadata.edge_vs_opening_fair_probability + the captured play-side price (resolveParlayOdds)

Checked live

ranked_rows_sorted_by_edgeunranked_rows_have_no_edgeplay_count_is_row_countpriced_count_matches_rowsrelease_ids_unique

16 house plays · This week’s picks · NFL 2026 Week 3

16 ranked by published edge · 0 without a published edge · 16 priced

#	Game	Play	Edge	Ticket price	Model p
1	KC @ MIA	MIA −8.5	+15.3 pts	−110	90%
2	CIN @ PIT	PIT −2.5	+13.4 pts	−110	90%
3	ARI @ SF	ARI +8.5	+12.0 pts	−110	67%
4	ATL @ GB	ATL +6.5	+11.9 pts	−105	80%
5	SEA @ WAS	SEA −2.5	+10.5 pts	−110	86%
6	TEN @ NYG	NYG +5.5	+8.3 pts	−110	66%
7	LA @ DEN	DEN −2.5	+8.1 pts	−110	75%
8	CAR @ CLE	CLE −2.5	+7.9 pts	−110	79%
9	LAC @ BUF	BUF +4.5	+7.1 pts	−110	58%
10	LV @ NO	NO +3	+6.4 pts	−110	74%
11	BAL @ DAL	DAL −3	+5.2 pts	−110	75%
12	PHI @ CHI	CHI +1.5	+4.2 pts	−105	64%
13	MIN @ TB	MIN +1.5	+3.9 pts	−110	67%
14	HOU @ IND	IND −2.5	+1.8 pts	−110	63%
15	NYJ @ DET	DET +7	+1.3 pts	−110	78%
16	NE @ JAX	NE +2.5	−1.8 pts	−110	60%

Swipe the table for price · model p

Edge is the number the model release published against the opening no-vig fair price — never derived here. Ticket price = the play's price at publish (the best book price at the play's line is on the tonight table); a missing capture reads as its words, never as −110.

Unlock the full toolkit

This week's slate above is free to read. Go Research Pro for alerts, CSV export, the DFS optimizer and unlimited model training.

$1 for your first week, then $5.99 per week.

Unlock for $1

Alerts · wake me when it matters

push · email · SMS · the slate moves while you work

Sunday digest 09:00

Crew

Crew · your people, graded together

No crew yet — crews arrive with sign-in.

Loading your Playbook board…