2026 season · saved snapshot

Week 3: the model desk

Lines captured 2026-09-17 · lines may change

Current SP+ (captured 2026-09-17) basis · why?

In-season spread predictions require at least 30 qualifying prior games from Week 5 onward. This snapshot reports current sp+ (captured 2026-09-17) as its basis.

Model record · W–L–P

15–21–0

36 settled picks · 2026 · Weeks 1, 2

Model picks this week

Your Bet365 results →

2025 ATS: 50.5% vs 52.4% break-even at −110.

Gap measures disagreement, not confidence. Evidence · 511 graded games

Explore this week

Results ↓

Find a team

22 games · largest absolute gap first

Keyboard: ↑ and ↓ change matchup. Tab moves to its details.

Matchup 1 of 22

Georgia State at UCF

UCF -18.25

Model–market gap: +11.4 pts
Projected home margin: +29.7
Market-implied home margin: +18.25

Positive gap favors the home side; negative favors the away side.

Historical evidence for this gap

7-14 point bucket · 44% hit rate · 101 games · 2025 backtest

Below historical break-even at −110

Open full matchup →How to read these numbers

Settled model results

All model results & separate Bet365 table →

Model pick / final	Outcome	Details
Hawai'i -7.5 vs New Mexico State 29–19 final	win Covered by 2.5
Season 2026 · Week 2 · Model’s archived line. Your actual wager can differ; see the separate Bet365 table.
Fresno State -18 vs Sacramento State 49–3 final	win Covered by 28
Season 2026 · Week 2 · Model’s archived line. Your actual wager can differ; see the separate Bet365 table.
Arkansas +12.5 vs Utah 10–43 final	loss Missed by 20.5
Season 2026 · Week 2 · Model’s archived line. Your actual wager can differ; see the separate Bet365 table.
Air Force +2.5 vs North Dakota State 32–38 final	loss Missed by 3.5
Season 2026 · Week 2 · Model’s archived line. Your actual wager can differ; see the separate Bet365 table.

Method & evidence

ATS means against the spread. Model–market gap is projected home margin minus market-implied home margin. A positive home margin means the home team wins by that many points. Historical hit rates describe past results, not the probability that this week’s pick wins.

2025 backtest: edge buckets and sample sizes

0-3 point gap · 52% hit rate · 231 games
3-7 point gap · 51% hit rate · 170 games
7-14 point gap · 44% hit rate · 101 games
14+ point gap · 78% hit rate · 9 games

Model output, not betting advice. Historical hit rates are small-sample backtests; several edge buckets lose money at -110. Flat units only.

Last source check: 2026-09-17 11:00 UTC. 2026 results through CFBD Week 2. Rated teams: 137/138.

Check the source time above; freshness is evaluated when this page opens.

Source-check details