Skip to content
Nothing But Bet · CFB intelligence

Research
desk

One place for what we forecast, what we observed, what the market captured, and what the data still cannot support.

Public players
4,835
1,145 weekly model players
PFF players
11,192
2025 normalized archive
Observed games
8
Pinned 2026 Week 1
Line games
98
Single Week 1 capture
Pinned 2026 results

Completed-game visual research

Eight Week 1 finals connect directly into existing matchup, team, and player dossiers. The upstream files and cfbfastR revision are hash-pinned in the checked-in contract.

Aug 30, 4:30 PM EDT
Pinned observed generation

Week 1 coverage

Completed games
8
Observed plays
1,233
Drives
110
Pass targets
443
Air-yard coverage
93%
xPass rows
911

Some completed games lack rows in the separate drive release. Each visualization fails closed at the game level.

What this layer is for

Use game flow and passing shape to explain how production happened. Use the frozen League range and an actual posted line to evaluate a prop. Never let a postgame chart masquerade as information available before kickoff.

Open CFB props
Observed player comparison

Week 1 passing and receiving leaders

Exact result rows link into player and game dossiers when identity coverage permits. Rankings describe this eight-game snapshot only.

21 passers · 134 receivers

Passing yards

PlayerTeamC/AYdsEPA/playGame
Jackson ArnoldUNLV25/38321-0.056View
Noah KimEMU26/36293+0.239View
Jayden MaiavaUSC25/29286+0.502View
Luke WeaverSJSU21/32234+0.302View
Billy Edwards Jr.UNC16/28232+0.141View
Marcus StokesMEM15/26223+0.065View
Davis WarrenSTAN12/22198+0.153View
Jaden CraigTCU20/32175-0.123View
Ashton DanielsFSU12/20162+0.279View
Nathan HayesNDSU14/21160+0.240View
Trey HeddenNMSU17/25159+0.231View
Beau PribulaUVA9/11155+0.463View
Every number has an owner

Local model ledger

Forecasts, matchup grades, historical evidence, and postgame models each keep their own clock and decision boundary.

5 active layers
Pregame forecastRefresh for Week 2

League weekly player distributions

cfb-fantasy-quantile-2026.1 · Aug 13, 1:02 AM EDT
Coverage
12,578 model rows across 761 matchups
Evidence
Causal development replay; p20/p50/p80 retained only for rows labeled model.
Public use
Player ranges, matchup prop monitor, fantasy decisions.
Boundary
39,666 prior/baseline context rows remain excluded from model surfaces. No game-side edge.
Season forecastRole refresh due

Custom 2026 season player board

custom-fantasy-2026-half_ppr_6pt_td_20pass-usage_aware-5b1c102afd16 · Aug 24, 7:01 PM EDT
Coverage
1,400 players; 138 teams
Evidence
Rate × games model with current-depth/usage allocation and custom scoring.
Public use
Season ranks, position context, draft value, floor/ceiling.
Boundary
Season and weekly projections are never averaged. Availability remains the noisiest component.
Historical + preseason contextStable preseason layer

Team and matchup intelligence

cfb-fantasy-season-2026.1 · Aug 13, 1:02 AM EDT
Coverage
136 structured team profiles; 892 games
Evidence
2025 efficiency, continuity, schedule, and authored team/matchup analysis.
Public use
Program fingerprints, unit ledgers, matchup scouting.
Boundary
Historical profile and betting record; never a 2026 side, total, or moneyline prediction.
Pregame matchup-fit gradeEarly-season context

CFB Scheme Score

Scheme Score Week 1 public contract · Aug 28, 12:13 PM EDT
Coverage
476 Week 1 WR/TE/RB receiving-market candidates
Evidence
Receiver alignment, coverage/funnel context, coordinator and team-change adjustments.
Public use
Rank matchup fit only for players with a posted prop line by default.
Boundary
Not a projection, hit probability, expected value estimate, or stake size.
Postgame descriptivePinned Week 1

cfbfastR observed game models

cfbfastR 3.0.0.9000 · Aug 30, 4:30 PM EDT
Coverage
8 completed games; 1,233 plays
Evidence
Observed play feed plus cfbfastR EPA, WP, CP/CPOE, and xPass models.
Public use
Game flow, drive, efficiency, air-yard, team, and player result views.
Boundary
Hindsight layer. It never updates the frozen pregame forecast or Scheme Score.
Freshness before features

Source health and next refresh

A current source can feed product. A stale or single-snapshot source stays labeled until a new immutable generation replaces it.

8 source systems
SourceRoleObservedCoverageStateNext action
ESPN schedule catalogSchedule / identityNot available239 linked teams · 892 gamescurrentRefresh status, scores, and kickoff changes on cadence.
CollegeFootballDataStats / roster / ratings / passingAug 29, 11:31 PM EDTCore model provider plus new air-yard/location/YAC endpointscurrentPromote postgame passing generations with availability counts.
RotoWire projectionsProjection / roleAug 24, 1:58 PM EDT1,803 rows · 99.3% CFBD identity resolutionrefreshNew capture before the next slate; preserve source observation time.
RotoWire depth and availabilityDepth / statusAug 24, 1:58 PM EDTFull FBS depth export with matched status flagsrefreshCapture changes as events, not only latest state.
PFF 2021–2025 archivesCharting / role / qualityAug 19, 8:00 PM EDT11,192 2025 players · 136 teams plus multi-year archivestableAdd 2026 measured samples when available; keep season labels explicit.
SportsDataverse / cfbfastRPBP / drives / model-derived resultsAug 30, 4:30 PM EDT8 Week 1 games · 443 pass targetscurrentRefresh after completed games and pin every asset hash.
The Odds APIMarket snapshotAug 11, 10:11 PM EDT98 Week 1 games with a spreadstaleReplace single snapshot with timestamped open/current/close captures.
Yahoo comparison snapshotIndependent raw-stat contextAug 13, 5:35 PM EDT599 matched playersstableRefresh only when an explicit comparison is useful; ordinary model weight stays zero.
Acquisition roadmap

What data is still missing

These are the next sources or capture systems that unlock real product—not generic requests for more data.

Ranked by product value
  1. 01

    Timestamped market history

    Unlocks: Truthful line-movement charts, CLV, forecast-to-quote joins

    Current: One dated Week 1 spread snapshot; no reliable quote sequence

    Acquire: Scheduled open/current/close capture with book, price, line, event/player IDs, and captured_at.

  2. 02

    Availability event history

    Unlocks: Role-change timelines, injury-adjusted projections, auditable stale-line warnings

    Current: RotoWire latest-state depth/status snapshot from Aug 24

    Acquire: Immutable refreshes plus effective_at, observed_at, status, source, and previous-state diff.

  3. 03

    Current-season routes and coverage

    Unlocks: 2026 Scheme Score components and real matchup-specific receiver fit

    Current: Early-season grade relies on 2025 alignment/funnel evidence

    Acquire: Refresh after sample gates; publish route/coverage counts and uncertainty.

  4. 04

    Player tracking / formation / route geometry

    Unlocks: Tracking playback, true formation viewer, and route tree

    Current: No XY frames, formation labels, or route coordinates in CFBD/cfbfastR

    Acquire: Licensed tracking source with stable game/play/player IDs and coordinate dictionary.

  5. 05

    Forward-test outcome ledger

    Unlocks: Live calibration, Scheme Score outcome tracking, honest promotion/demotion

    Current: Development evidence exists; 2026 results are not yet a public frozen ledger

    Acquire: Score frozen pregame rows after completion without retuning from the ledger.

  6. 06

    Weather and field conditions

    Unlocks: Kickoff-condition cards and late movement explanations

    Current: Venue is available; forecast/observation history is not normalized

    Acquire: Timestamped venue weather with forecast issue time and observed conditions.

Lab to public product

Visualization deployment map

Every gated demo now has a public destination or an explicit source requirement. Unsupported tracking concepts remain off until the data exists.

10 live patterns
VisualizationStatusPublic destinationData contract
Football field + current playLiveCompleted matchup pagesObserved start, line to gain, finish
Tracking playbackSource neededNot publicPlayer XY frames do not exist in current sources
Formation viewerSource neededNot publicNo verified formation labels
Route treeSource neededNot publicNo route geometry
Pass mapLiveMatchup, team, playerSchematic air yards + categorical direction
Passing-production profileLiveTeam pagesQB and target observed production
Field heatmapLiveTeam and player pagesShort/deep × left/middle/right target grid
Signed evidence barsLiveCompleted matchup pagesHome-minus-away observed efficiency
Win probabilityLiveCompleted matchup pagesPostgame cfbfastR series
Drive chartLive with fallbackCompleted matchup pagesFails closed when drive asset has no rows
Team comparisonLiveCompleted matchup pagesExact observed values beside signed bars
Player comparisonLiveResearch leaders + player pagesPasser and receiver result tables
Projection intervalLiveTeam, matchup, playerLeague pregame p20/p50/p80 only
Market-line historyCapture neededSingle-point status onlyNo fabricated time axis from one snapshot

Methodology and data boundary

The research desk is a routing and provenance layer, not a new blended model. League weekly/season forecasts, PFF historical measurements, Scheme Score matchup grades, captured market lines, CFBD passing results, and cfbfastR postgame models remain independently labeled through display. Missing or stale sources stay visible.

Research snapshot
Aug 30, 4:30 PM EDT
cfbfastR commit
413ce04d3c8a
Observed games
8
Model blending
Disabled across layers