6.89 yards per rushing play, the highest average in this processed FBS feed.
2025 college football · all FBS
What kind of
offense was each team?
An investigation into offensive style: how every team gained yards, how it scored, what survived stronger defenses, and eight groups that blend all three signals.
136FBS teams
5,622offensive touchdowns
1,837touchdowns of 20+ yards
8offensive styles
Three clues
Efficiency and scoring shape tell different stories.
9.6 yards per passing play, including sacks, leads the FBS sample.
2.23 touchdowns of 20+ yards per game, the highest rate in this sample.
01 · How they gain yards
Does the offense move through the run or the pass?
Rushing yards per play runs left to right; passing yards per play runs bottom to top. The diagonal separates run-heavier efficiency profiles from pass-heavier ones.
Run vs. pass yards per play · all defenses
Global FBS percentile on both axes · 136 offenses with data
Passing yards per play · global percentile
Rushing yards per play · global percentile
02 · How they score
Does scoring come from short finishes or long strikes?
Under-20 touchdowns per game run left to right; touchdowns of 20+ yards run bottom to top. Both include rushing and passing scores.
Short scores vs. long scores
Global FBS percentile on both axes · 136 offenses with data
20+ touchdowns per game · global percentile
Under-20 touchdowns per game · global percentile
03 · How competition affects them
What survived better defenses?
Final DFEI ranks all 136 defenses globally—not within each team's schedule. Top 50% compares with the complementary bottom 50%; top 25% compares with the complementary bottom 75%.
All touchdowns
Global FBS percentile on both axes · 135 offenses with data
Vs. top-50% DFEI defenses · global percentile
Vs. bottom-50% DFEI defenses · global percentile
Oregon
Top-50% defenses11 games
35 all touchdowns
Bottom-50% defenses3 games
21 all touchdowns
Strong-defense rate change
55% lower7.00 → 3.18 per game
Per-game rates reduce the schedule-count distortion, while the raw scale exposes the actual production behind each percentile. Teams with no game in the selected FEI cut are omitted. FCS opponents are excluded, and this remains descriptive rather than causal.
04 · The eight groups
Eight efficiency styles, not eight grades.
The model uses overall efficiency, pass–run efficiency balance, and sensitivity to defensive strength. Select a cluster to compare every member with that cluster's median.
Cluster 1 · 10 teams
Ground-leaning, pressure-sensitive
Median Overall efficiency4.87 YPPMedian Pass–run balance0.42 YPPMedian Defense sensitivity-0.86 YPP
TeamOverall efficiencyPass–run balanceDefense sensitivity
Western MichiganMid-American5.31+0.44 vs. median0.37-0.05 vs. median-0.82+0.04 vs. median
UCLABig Ten5.27+0.40 vs. median-0.09-0.51 vs. median-0.35+0.51 vs. median
WyomingMountain West5.20+0.33 vs. median0.64+0.22 vs. median-0.92-0.06 vs. median
WisconsinBig Ten4.93+0.06 vs. median1.00+0.58 vs. median-1.19-0.33 vs. median
Oklahoma StateBig 124.90+0.03 vs. median1.22+0.80 vs. median-0.91-0.05 vs. median
TroySun Belt4.83-0.04 vs. median1.54+1.12 vs. median-1.30-0.44 vs. median
RiceAmerican Athletic4.79-0.08 vs. median0.48+0.06 vs. median-0.59+0.27 vs. median
Northern IllinoisMid-American4.79-0.08 vs. median-0.51-0.93 vs. median-0.91-0.05 vs. median
Ball StateMid-American4.50-0.37 vs. median0.02-0.40 vs. median-0.81+0.05 vs. median
MassachusettsMid-American4.26-0.61 vs. median-0.26-0.68 vs. median-0.37+0.49 vs. medianTeam index · All 136 teams
The complete style board.
Each team has four vertical touchdown-length bars. Red segments came against top-half DFEI defenses; blue segments came against the bottom half. Scores against unrated opponents remain in the all-TD total.
Top-50% defense Bottom-50% defense
Bar height shows touchdowns in each labeled distance bucket136 teams shown
TeamTDs by labeled distance bucketRatedAll TDsCluster
Method and limits
One consistent cut of an imperfect feed.
The yardage views use non-nullified 2025 FBS rush and pass plays with a plausible gain; passing yards per play includes sacks. The scoring views count rushing and passing touchdowns, then split them at 20 yards. The style board groups those scores into four labeled distance buckets.
The k-means model uses three equally weighted traits: overall yards per play, the passing-minus-rushing efficiency gap, and the stabilized change in overall yards per play against top-half versus bottom-half defenses. The eight names describe styles rather than grades.
The defense-strength view ranks all 136 FBS defenses by final 2025 DFEI once globally. The top-half view compares ranks 1–68 with 69–136; the top-quarter view compares ranks 1–34 with 35–136. It classifies 4,925 of the 5,622 offensive touchdowns in this feed; scoring against FCS and other unrated opponents remains in overall totals but not the comparison.
Exact model and data notes
Style-board bins: 1–5, 6–19, 20–39, 40+ yards. Feature: Three standardized scalar traits: overall yards per play, passing-minus-rushing yards-per-play balance, and the change in overall yards per play against top-half versus bottom-half DFEI defenses. Algorithm: k-means++ initialization, Euclidean distance, 50 starts, fixed 2025 seed. Strength feature: Defense sensitivity is top-half overall yards per play minus bottom-half overall yards per play, with each side stabilized toward the team's full-season efficiency using a 150-play prior. Negative values indicate a larger decline against stronger defenses.. Exclusions: Defensive, kick-return, punt-return and other non-rush/non-pass touchdowns; nullified plays; plays without a valid gain. Yardage filter: Rush and pass plays with a valid gain from -30 to 100 yards; nullified plays and duplicate play IDs excluded. Passing yards per play includes sacks.. Defense split: Strong-defense comparisons use either the global top 68 or global top 34 FBS defenses by final 2025 DFEI. Their x-axis baselines are the complementary global bottom 68 or bottom 102. Per-game comparisons use only games against the selected rated opponents..
Processed public play-by-play contains known omissions and duplicated or miscoded plays. Figures are approximate and are shown as published after one consistent scoring-play filter.