Why Big Spreads Hurt Preseason Favorites in Division Games

Overview

A common belief in NFL betting circles: divisional games are tighter than the spread, with rivalry matchups producing more upsets and tougher covers. We tested it on 11 seasons of joined game and market data and got a clearer picture than the conventional wisdom suggests. This note is descriptive research, not wagering advice.

Setup

This analysis pulls from the AGINT Intelligence Graph, Fanvatic’s multi-source NFL data product. The graph covers 13 NFL seasons (2013–2025) with game-level outcomes, multi-book line snapshots (FanDuel, DraftKings, BetOnline, Pinnacle, plus historical sources), preseason market win totals for all 32 teams, weather, advanced statistics, and intelligence assessments from a collection-agent layer. For this specific question, we used 11 seasons (2013–2024), joining three slices: divisional game outcomes, closing lines from FanDuel and nflverse historical sources, and preseason win totals via pro-football-reference. That intersection produced 3,271 games with full spread and outcome data, including 1,169 divisional matchups. The 2025 season was excluded because spread-anchored outcomes were not yet complete at the time of analysis. The breadth of the underlying graph is what made this join possible at scale.

Finding 1: Divisional home teams cover slightly less

Across 11 seasons, divisional home teams covered the spread 47.05% of the time compared to 48.29% for non-divisional home teams. On n=1,169 divisional games, the difference is marginally statistically significant. It is also a small enough effect on its own that it does not clear -110 juice (52.4% breakeven). But it is a real, persistent characterization of how the divisional market behaves: home teams in division games face slightly more pressure than the spread accounts for.

Two related observations from the same dataset:

• Divisional spreads average about half a point larger than non-divisional spreads (5.48 vs. 5.12), so the market is not pricing rivalry games tighter despite the conventional wisdom that says it should.
• Home teams in divisional games actually win by less on average (1.53 points vs. 2.29 for non-divisional), suggesting the home-field advantage is partially neutralized by divisional familiarity.

Finding 2: Preseason favorites cover less as spreads grow

This is the pattern that genuinely jumps out of the data. We isolated divisional games that were the first time the two teams played that season, where the preseason-favored team (higher market win total going into the season) was also the betting favorite. That produces 423 games across 2013–2024 meeting all criteria. Bucketing by closing spread magnitude shows a near-monotonic decline in the favorite’s cover rate as the spread grows:

The decline from smallest to largest bucket is 10.85 percentage points. The biggest spreads in divisional first matchups, games where one team was a heavy favorite based on both preseason expectations and current betting markets, saw the favorite cover only 40.7% of the time.

Possible mechanisms

No single explanation can be proven cleanly from this data, but several plausible mechanisms could contribute to the observed decline:

• Public bias on big-name teams may push the line beyond efficient pricing in lopsided matchups.
• Divisional opponents may play with more motivation than non-divisional opponents at the same spread, knowing they will see each other twice.
• Backdoor cover dynamics at large spreads, where favorites go conservative late in games and dogs cover during garbage time, can systematically reduce favorite cover rates.
• The market may not adjust enough for divisional familiarity in lopsided matchups, particularly early in the season before in-season information has been priced in.

Caveats and limitations

The huge-spread bucket (10+ points) has only n=54. The 40.7% cover rate in that bucket is suggestive but the sample is small enough that the pattern needs more data to be definitive. Across all four buckets the decline is monotonic enough to be worth flagging, but a single bucket’s magnitude should not be over-interpreted.

This analysis is descriptive rather than predictive. Whether the pattern persists forward requires validation that this work does not attempt. We will revisit this analysis when 2025 spread-anchored outcomes are fully joined into the graph, and publish what the expanded sample shows either way.

Methodology note

The AGINT Intelligence Graph is built to answer NFL market questions that span data sources, seasons, and time. Each game node carries its own context (rest days, travel, situational flags) and connects through relationships to outcomes, line snapshots from multiple books, weather, and assessments. Team-season nodes carry roll-up statistics and, as of this analysis, preseason market expectations. The graph preserves source provenance and temporal validity for every fact, which is what made it possible to join 11 seasons of preseason expectations to game-level outcomes without leakage or methodology drift. The architecture supports a range of analyses; this note covers one.

Takeaways

• The market does not price rivalry games tighter despite the conventional wisdom. Divisional spreads run about half a point larger than non-divisional spreads on average.
• Divisional home teams cover slightly less than non-divisional home teams (47% vs 48% across 1,169 games). Small, persistent, and not meaningful on its own.
• Divisional first matchups with large closing spreads are the most interesting bucket. The preseason favorite covered only 40.7% in the 10-point-plus group. Worth watching forward, and we will.