ABOUT & METHODOLOGY

B1G+A Model 2.2 (current)

See the full version history and timeline →

The rating blend (from 1.1, still used in 2.2)

The lead input is the B1G+A Power Rating, fit from every 2026 FBS final and refit after each game day. Before any games, it starts from last season's final rating pulled back toward average (a carryover from 2025, not a poll). It is blended with ESPN FPI, and the FPI weight falls as the season goes on: 85% through Week 4, then 5 points less each week (70% in Week 7), with a 50% floor from Week 11. FPI is still shown on its own wherever it appears, as an ESPN comparison.

Being 100% transparent. B1G+A launched on Oct 9, 2026 (v1.0) and is data first, data only: no opinions, no fan bias, refit as new results come in. Right now it leans on public numbers it does not own (ESPN FPI and ESPN stats, scores and feeds; sportsbook lines for comparison only). B1G+A started building its own dataset in Week 6, after the Friday game (Iowa at Washington, Oct 9). Starting Week 7, B1G+A runs 100% live in-game analytics (live win chance updated every few seconds, every FBS game) and keeps that play-by-play as its own data to train the model, so its own share grows each week. Until the data is fully B1G+A's own, B1G+A uses only publicly available, shareable data from other providers, credited by name.

Who's B1G+A's dad? It doesn't have one; it's not human (if it did, he'd probably be in Lincoln wearing red, saying this is finally the year). It takes in data and puts out its own numbers, with no slant, no bias and no person picking winners.

Version history. v1.0 (Oct 9, 2026) was led by ESPN FPI. v1.1 (Oct 10, 2026, 11:29 PM CT) added the B1G+A Power Rating blend. It scored best for accuracy in 2024–2025 backtests (highest share of winners picked correctly).

* Numbers published before Oct 10, 2026, 11:29 PM CT came from v1.0 and keep their original values, so some differ a little from later versions (example: Nebraska pregame playoff odds 27.1% published vs 25.3% under v1.1). Model notes · Version history

What feeds what

1. Power rating: the B1G+A blended rating (B1G+A Power Rating + ESPN FPI, weights above).
2. Game win %: from the rating gap between the two teams plus home-field advantage.
3. Projected records: the average regular-season wins and losses across a 40,000-run season simulation that plays out each team's remaining ESPN schedule.
4. CFP odds: the same simulation adds conference title games and the 12-team format (5 highest-ranked champions + 7 at-large). The start-of-day playoff % blends the simulation with ESPN FPI's playoff %, using the same FPI weight. On game days, each live game's two possible outcomes are weighted by the live win probability.
5. Projected CFP bracket: the field and seeds 1–12 come from the "if the playoff were picked today" projection. Its estimated CFP rank averages each team's AP rank and ESPN Strength of Record rank. CFP odds are shown next to each team. Which quarterfinal bowl each top seed plays in is a B1G+A projection.
6. Bowl projections: CFP teams come straight from the bracket. Every other B1G team with at least a 50% chance of 6+ regular-season wins is slotted by projected wins into the 2026-27 tie-ins.
7. B1G+A Top 25: uses the blended rating as one of its inputs (see below).

B1G+A Top 25

A résumé-first B1G+A ranking, not an official poll. Points = 50% ESPN Strength of Record rank + 30% B1G+A blended rating rank + 20% loss rank, minus 2 for each win over a team in that week's AP Top 25. Lowest points ranks first.

No preseason polls are used (no preseason AP, Coaches or any other poll). Two notes: the "Top 25 win" bonus uses the current week's AP Top 25, and the blended rating still carries some preseason information early in the year (ESPN FPI's own preseason component and the 2025 carryover in the B1G+A Power Rating). That part fades as games are played.

Head-to-head

• Power Rankings (power.html and the B1G Power Rankings card) use head-to-head directly. When the teams are close, the winner goes ahead of the loser: the loser's blended rating is no more than 7.0 points higher and the winner's ranking score is no more than 4.0 lower.
• B1G+A Top 25 weighs head-to-head heavily when teams are close, but it isn't an absolute override. The winner moves directly ahead only when the loser's blended rating is at most 7.0 points higher and the winner's Top 25 points are no more than 8.0 worse. This applies from Week 8. Week 7 was published under the earlier rule (3.5-point rating gap).
• Both ignore split series (each team won once) and loops (A beat B, B beat C, C beat A). Those teams keep their rating and résumé order.

Freeze rule

Once a weekly ranking or report is published, it isn't changed. Rule and model changes apply going forward only. The published B1G+A Top 25 and the B1G Power Rankings card are locked for their week. The all-FBS table on Power Rankings is a live rating that updates after every game day. It is not a frozen weekly report.

Live win probability

How the live win chance works (live and box pages). This method covers Week 7 games. Each snap is scored from the score, time, field position, down and distance, timeouts and the pregame B1G+A rating, and B1G+A projects each team’s win chance from there.

• Earlier in the game: a logistic model is used.
• Final 6:00 of regulation: a drive-by-drive model takes over. It considers whether the offense can score a TD or FG with the snaps and timeouts left, whether a leader can run out the clock, and what the other team can do with the time left after. Its drive model is fit on 2014–2025 college play-by-play; its field-goal make-rate curve is fit on 2014–2026 kicks (21,579 attempts of 35+ yards, all divisions).
• Overtime: a separate overtime model is used.

Known limits.

• Between 6:00 and 5:55 left, B1G+A win probability can briefly move up to ~3 points the wrong way with timeouts while blending into the late-game model.
• In the final minute, when a team trailing by 4 to 8 is inside the opponent’s 10, this model can understate its chances by roughly 10–20 points (2024–25 test: 29% projected vs 41% actual, 102 plays).
• Overtime win chances come from a separate overtime model and are less reliable than regulation numbers, especially on the first possession of each overtime period, when the other team’s answer is still to come.
• Field-goal odds beyond 60 yards rest on thin data (17 or fewer attempts per yard, plus an assumed leg limit), and may run a bit high: at 58–62 yards the model averages about 31% vs 27.9% actual (111 kicks).
• Fourth-down call changes before 6:00 (win probability unchanged): the new field-goal curve flips two earlier recommendations on the published site, both toss-ups where the options are nearly tied: BUFF@PSU Q2 3:36 FG → punt; IOWA@WASH Q4 7:12 FG → go. Win probability is identical to the previous engine on every play before 6:00 of the 4th quarter.
• Fix: the previous engine treated an untimed down at 0:00 as final (e.g., WAKE@PUR’s last-play FG showed Purdue at 0%); it now accounts for the remaining play and overtime.

Held-out 2024–25, final 2:00 of regulation: Brier 0.0671 vs 0.0727 for the previous engine (lower is better); 3:00–6:00: 0.0698 vs 0.0725. Accuracy against ESPN’s live win probability is on Model Record.

Receipts

Every pick is graded and shown, hits and misses alike. A hit means the projected winner won outright. Margin error is the gap between projected and actual margins. Spread and total leans are graded against the DraftKings line used at posting time. See Model Record.

Weekly drops

Receipts on Monday, the B1G+A Top 25 on Sunday at 9:05 AM CT, CFP projections at 9:15 AM CT, bowl projections at 11:05 AM CT, then Power Rankings that afternoon. Full times are on the Weekly Drop Schedule.

GOAT Playoff

The GOAT Playoff is a fun event for fans to kick off the launch of B1G+A. It is fan-voted: results combine votes on this website and X polls. It is not a B1G+A model output.

Data sources and stamps

Scores, records, AP/Coaches/CFP polls, ESPN FPI and Strength of Record come from ESPN's public feeds. Lines are DraftKings via VegasInsider. Every third-party number is labeled as such, and figures carry the date of the data they came from.

• ESPN public scoreboard, play-by-play and box score data: used for live game state, plays, scores, records, schedules, standings and team stats.
• ESPN FPI, Strength of Record and AP/Coaches/CFP polls (via ESPN): shown as labeled comparisons and used as inputs where noted (the FPI share of the B1G+A rating blend; résumé inputs for the Top 25 and CFP projection).
• Sportsbook lines (DraftKings via ESPN; DraftKings, FanDuel and Caesars lines via VegasInsider): shown for reference and used to grade spread and total leans. Lines are never our picks.
• Coach salaries (USA Today), roster-budget reporting (The Athletic, On3, SI) and the FBS coaches list (Wikipedia): used for the CFB Payrolls page.
• Breaking news: linked to the original report, credited to the named reporter or outlet.
• The B1G+A Model, win chances, 4th-down calls and B1G+A ratings are our own work. All Big Analytics is independent and not affiliated with ESPN or any other source listed here, and it is not a betting site.

Independence

All Big Analytics is fan-made. It is not affiliated with any conference, school, network, sportsbook, bowl or the College Football Playoff. B1G+A is the name of the All Big Analytics in-house model.