Marshall Thundering Herd Football · 2026 Season Preview

Marshall Thundering Herd

A data-backed preview built from the committed 2026 preseason SP+ snapshot, the CFBD schedule, and Bill Connelly's extracted conference preview tables. Every projection traces to a stored payload.

2026 SP+
#92
-6.4 overall · 4th of 14 in the Sun Belt
Projection
8.0
Expected wins across 12 games
Most Likely
8-4
29.7% of outcomes
Off / Def
54/124
National ranks · Connelly model: 7.8 avg wins

Executive read

Marshall Thundering Herd enters 2026 rated 4th in the Sun Belt (#92 nationally). The profile is a poor defense (#124) paired with an above average offense (#54). Schedule math puts the season at 8.0 expected wins, against Connelly's independent 7.8. 2 games project as clear underdog spots (Penn State, James Madison). The season turns on 2 close games: Old Dominion, Georgia Southern.

Page 2 · Team Quality

What the Ratings Say

Overall SP+
#92
-6.4 rating
Offense
#54
28.5 pts/gm · above average
Defense
#124
35.2 allowed · poor
Special teams
#48
+0.3 value

Unit ranks, 1–138 national

Offense#54
Defense#124
Special teams#48

Bars run left (worst) to right (best): a longer bar is a better national rank. Special-teams ratings are volatile year to year.

2025 resultRecordConfSP+ rank OffDefST
Where they finished5-7 3-5 #86 #52 #119 #48

Sideline

Tony Gibson — year 2, 5-7 career.

OC: Rod Smith
DC: Brad Lambert (NEW/reunion)

Lambert returns to Marshall DC role he held 2019-20 with massive improvement; Gibson hired for WV ties and defense

Returning production

48% overall, ranked #88 nationally.

Offense 61% · Defense 36%

112 returning starts, 20 redshirt freshmen.

Two models, one slate. This report's schedule math projects 7.98 wins; Connelly's published model projects 7.8 (5.1 in conference) with a strength-of-schedule rank of #102. His model gives 94.5% odds of six-plus wins and 1.9% of eleven-plus. Two independent methods landing close together is corroboration, not confirmation — they share SP+ as an input.
Page 3 · Schedule Math

Game-by-Game Projection

WeekDateOpponentOpp SP+ Proj marginWin %Tier
Wk 12026-09-05@ Penn State#17 / +15.7-24.64%Upset needed
Wk 22026-09-12vs Middle Tennessee#135 / -26.0+22.194%Should win
Wk 32026-09-19@ Missouri State#126 / -18.7+9.877%Lean win
Wk 42026-09-26vs Gardner-WebbFCS+21.193%Should win
Wk 52026-10-03@ James Madison#77 / -2.1-6.830%Upset needed
Wk 62026-10-10vs Coastal Carolina#119 / -13.8+9.978%Lean win
Wk 82026-10-20vs South Alabama#116 / -13.3+9.476%Lean win
Wk 92026-10-31@ Old Dominion#88 / -5.8-3.140%Lean loss
Wk 102026-11-07@ Georgia Southern#103 / -8.9+0.050%Swing
Wk 112026-11-14vs App State#111 / -12.1+8.274%Lean win
Wk 122026-11-21vs Georgia State#133 / -25.1+21.293%Should win
Wk 132026-11-28@ UL Monroe#132 / -24.3+15.487%Should win

Opening third

2.69

expected wins across 4 games
Penn State, Middle Tennessee, Missouri State, Gardner-Webb

Middle third

2.24

expected wins across 4 games
James Madison, Coastal Carolina, South Alabama, Old Dominion

Closing third

3.04

expected wins across 4 games
Georgia Southern, App State, Georgia State, UL Monroe

How to read the margin column. It is the SP+ rating gap plus home field, in points — a positive number means the model favors Marshall. Win percentage is that margin mapped through a logistic calibrated so a seven-point favorite wins about 70% of the time.
Page 4 · Outlook

Season Distribution

Expected wins
7.98
across 12 games
Bowl odds
97%
six or more wins
Total mass
100%
distribution sums to 100%
Conference
Sun Belt
14 teams rated

Win-total probability

4 wins0.5%
5 wins2.8%
6 wins9.7%
7 wins21.7%
8 wins29.7%
9 wins23.7%
10 wins10.0%
11 wins1.8%

Exact Poisson-binomial over the projected game probabilities. Bars are scaled to the most likely outcome, and the percentages are absolute.

Quarterback

Carlos Del Rio-Wilson

returning starter experienced

2043 pass yds 714 rush yds over 10 starts; former top-60 recruit; multiple career stops; good dual-threat

Program context

Win Drop · 10_to_5
Win total cut in half from 10 to 5 after Huff left for Southern Miss; dead last nationally in returning production

Late Close Losses · 0-2
Outgained App St and Ga Southern 6.6 to 5.1 ypp but lost by 7 combined points

DateMatchupWhy it matters
Oct. 3Marshall at JMUOther potential East favorite visits Harrisonburg; Napier won't have time to figure out roster
Oct. 31Marshall at ODUIf JMU isn't class of division probably one of these two; extremely East-heavy list
Page 5 · Conference & Method

Where They Sit in the Sun Belt

#TeamSP+ rankRatingOff DefRet prodRet rank
1James Madison#77-2.1#76#7530%#136
2Old Dominion#88-5.8#109#6340%#114
3Troy#90-6.0#90#9439%#119
4Marshall#92-6.4#54#12448%#88
5Louisiana Tech#100-8.3#98#9956%#53
6Arkansas State#101-8.5#88#11149%#82
7Georgia Southern#103-8.9#92#10943%#108
8Louisiana#104-9.1#72#11649%#81
9App State#111-12.1#102#11534%#127
10South Alabama#116-13.3#95#12544%#105
11Coastal Carolina#119-13.8#103#11842%#112
12Southern Miss#131-23.3#123#13322%#138
13UL Monroe#132-24.3#130#13051%#68
14Georgia State#133-25.1#120#13738%#121

Method & Provenance

Sp 2026. Committed preseason SP+ snapshot data/external/cfbd_2026/sp_plus_2026.csv via cfb_model_lab.preseason.load_sp_plus; CFBD /ratings/sp returns zero rows for 2026.

Schedule. data/external/cfbd_2026/schedule_2026.csv (CFBD /games snapshot, regular season only).

Connelly. Bill Connelly ESPN+ 2026 conference previews extracted to data/external/connelly_*_2026/.

Projection Method. SP+ differential plus 2.5-point home field, mapped to win probability by a logistic with scale 8.0 (illustrative scale; not validated for this SP+ snapshot). Win distribution is an exact Poisson-binomial assuming independent games. Not a betting model and not the repo's StackedEnsemble forecast.

Lower Division. FCS/lower-division opponents absent from the FBS SP+ table are assigned a -25.0 rating placeholder.

Not betting advice. No market lines were pulled for this report and no edge is claimed. These are schedule-strength projections from public preseason ratings. If a real market exists, price it independently.