← Back to HomeInternal

2026 Betting Playbook – Hybrid V2, Conflict Tags, and Fade V4

Last updated: prepping for the 2026 season.

This document defines how we actually use the Gridiron Edge models in 2026, not just how they're computed.

The core ideas:

  • Hybrid V2 is the only production spread model.
  • Conflict tags (hybrid_strong, hybrid_weak, hybrid_only) tell us when to trust Hybrid more or less.
  • V4 / Fade V4 remain Labs-only, but fading V4 has shown consistent positive ROI and helps define our "strong" vs "weak" buckets.

1. Models: Production vs Labs

Production:

  • Hybrid V2 (strategyTag = 'hybrid_v2')
    • This is the model used for My Picks and live recommendation logic.

Labs / Experimental:

  • V4 (Labs) (strategyTag = 'v4_labs')
    • SP+/FEI-inspired drive-based spread model.
    • Standalone performance has been unprofitable in backtests.
  • Fade V4 (Labs) (strategyTag = 'fade_v4_labs')
    • Takes the opposite side of every V4 bet.
    • Has shown positive ROI in both 2024 and 2025.
    • Still treated as experimental, not production.

All non-Hybrid strategies are informational overlays and do not drive My Picks directly.

2. Key Concepts

Edge

  • edge = model spread vs closing line (HMA format).
  • We use |edge| (absolute value) to group bets into tiers.

CLV (Closing Line Value)

  • How much the market moved in our favor between open and close.
  • Positive CLV = market agrees with us.
  • Strong Super Tier A performance comes with strongly positive CLV.

Conflict Types (hybridConflictType)

Stored on every Bet row as hybrid_conflict_type:

  • hybrid_strong
  • hybrid_weak
  • hybrid_only

These are diagnostic labels computed using 2025 behavior of:

  • Hybrid results
  • Fade V4 results
  • CLV behavior

Rough intuition:

  • hybrid_strong
    Games where Hybrid's signal has historically converted to high ROI with supportive CLV, and Fade V4 behavior is consistent with that strength.
  • hybrid_weak
    Games where Hybrid's edges look big on paper, but ROI and/or CLV have been much weaker or flat.
  • hybrid_only
    Games where only Hybrid has a bet (rare in 2025).

Exact formulas live in sync-hybrid-conflict-tags.ts. This doc is about how to use the labels.

3. 2025 Snapshot – Why Super Tier A Exists

Using 2025 Hybrid V2 spread bets:

3.1 Super Tier Candidate

Filter:

  • strategyTag = 'hybrid_v2'
  • hybridConflictType = 'hybrid_strong'
  • |edge| >= 4.0

Results (2025):

  • 274 bets
  • Record: 215–58–1 (78.8% win rate)
  • ROI: +50.16%
  • Avg edge: 18.49
  • Avg CLV: +15.96

This is the basis for the Super Tier A concept.

3.2 Hybrid Strong vs Hybrid Weak (All Edges)

Hybrid Strong (all edge sizes):

  • 357 bets
  • 73.2% win rate
  • +39.34% ROI
  • Positive CLV

Hybrid Weak (all edge sizes):

  • 351 bets
  • 52.8% win rate
  • +0.69% ROI
  • Slightly negative CLV

Conclusion:

  • Hybrid Strong is where the model is genuinely sharp.
  • Hybrid Weak is essentially a breakeven bucket.

3.3 V4 and Fade V4

Across 2024–2025:

  • V4 (Labs) standalone:
    • Unprofitable in both seasons (negative ROI, sub-40% win rate).
  • Fade V4 (Labs):
    • 2024: ~10% ROI
    • 2025: ~18% ROI

Fade V4's profitability helps separate strong vs weak Hybrid games, but V4 itself remains experimental and is not used as a standalone strategy.

4. 2026 Tiering Rules

4.1 Tier Definitions (Hybrid V2)

We tier Hybrid V2 bets by absolute edge:

  • Tier C: |edge| < 2.0
    • Default: ignore. Info only.
  • Tier B: 2.0 ≤ |edge| < 3.0
    • "Leans / action plays".
    • Requires other confirmation (e.g., Barnes, Crick, matchup notes).
  • Tier A: 3.0 ≤ |edge| < 4.0
    • Serious plays.
  • Super Tier A: |edge| ≥ 4.0 AND hybridConflictType = 'hybrid_strong'
    • Top-shelf, hammer-worthy bucket.
    • Origin: 2025 backtest (78.8% win rate, +50.16% ROI).

4.2 Conflict Type Usage

For Hybrid V2 bets:

  • hybrid_strong
    • Forms the backbone of 2026 ATS strategy.
    • Used for:
      • Super Tier A: |edge| ≥ 4.0
      • Tier A (Strong): 3.0 ≤ |edge| < 4.0
      • Tier B (Strong): 2.0 ≤ |edge| < 3.0 (optional, needs external confirmation)
  • hybrid_weak
    • Not auto-bet in 2026, regardless of edge size.
    • May be used as "consider only with strong external confirmation".
  • hybrid_only
    • Rare; treat case-by-case.

5. Fade V4 in 2026

Fade V4 remains Labs-only, but:

  • Has shown consistent positive ROI as a standalone backtest.
  • Performs best in Hybrid Strong games.
  • Underperforms in Hybrid Weak games.

Usage guideline:

  • Only use Fade V4 as confirmation or secondary Labs overlay in games tagged hybrid_strong.
  • Ignore V4/Fade V4 in hybrid_weak games.

6. My Picks UI Semantics

To make the 2026 playbook usable from the couch:

For each Hybrid V2 pick shown on My Picks:

  • Show a Conflict badge:
    • Strong, Weak, or Only
    • Color-coded (e.g., green / yellow / neutral)
  • Show a Tier label based on |edge|:
    • Super Tier A (Strong + |edge| ≥ 4.0)
    • Tier A (Strong) (3.0–3.99, Strong)
    • Tier B (Strong) (2.0–2.99, Strong)
    • No label for weaker edges
  • Provide UI filters:
    • "Show only Super Tier A"
    • "Show only Hybrid Strong"

This keeps the highest-value plays front and center without hiding the rest of the model output.

7. Operational Notes

To keep this playbook valid during the season:

  1. Normal workflows (nightly ingest, grading, etc.) must stay green.
  2. Conflict tags must be kept up to date:
    • Run sync-hybrid-conflict-tags.ts after grading for each week/season as needed.
  3. Fade V4 remains Labs-only:
    • May be exposed in Labs screens and export tools.
    • Not used to drive My Picks directly.

8. Continuity Guardrails (Labs Only – Based on 2025 Backtest)

Continuity Score is a 0–100 roster stability metric built from CFBD returning production + portal churn. High = stable/veteran, Low = chaos/new pieces. It is not in the production Hybrid V2 model yet, but we use it as a tactical overlay.

What 2025 showed:

  • Hybrid V2 was profitable across all continuity bands (Low/Mid/High).
  • The big pattern was favorites vs dogs, not "high good, low bad."
  • Low-continuity dogs were consistently terrible.
  • Low-, mid-, and high-continuity favorites all crushed.

Soft rules for 2026 (subject to further testing):

  1. Low-continuity dogs (<0.60): yellow flag.
    • Avoid unless the game is Super Tier A and other context supports it.
    • These profiles were -14% to -21% ROI in 2025 backtests.
  2. Favor favorites.
    • Favorites performed extremely well across all bands, especially low-continuity favorites (~+50–60% ROI).
    • Continuity should not scare us off a strong favorite if Hybrid likes the number.
  3. Be picky on 14+ point spreads with mid/high continuity.
    • These were roughly breakeven in 2025.
    • Treat them as lower priority or require stronger edge / conflict alignment.

These are guardrails, not hard filters. They inform human review of the card and Labs experiments (e.g., re-running 2025 portfolios with low-continuity dogs removed). Any hard-coded model changes belong in a future Hybrid V5 cycle.

8.1 2025 Portfolio Experiment: Dropping Low-Continuity Dogs

We simulated the 2025 official card (official_flat_100) with and without low-continuity dogs:

Baseline (all bets): 937 bets, 59.5% win rate, +13.53% ROI, +$12,676.80 PnL

Removed subset (low-continuity dogs only): 316 bets, 40.8% win rate, -22.04% ROI, -$6,964.80 PnL

Filtered card (dropping low-continuity dogs): 621 bets, 69.2% win rate, +31.63% ROI, +$19,641.60 PnL

Impact: Removing 316 low-continuity dog bets would have improved PnL by +$6,964.80 and ROI by +18.10 percentage points (from +13.53% to +31.63%).

This is Labs-only evidence and not yet a hard rule, but it strongly supports the guardrail: low-continuity dogs were a significant drag on the 2025 official card. The experiment suggests that avoiding these bets in 2026 card construction could meaningfully improve performance.

8.2 Hybrid V2 Portfolio Experiment: Dropping Low-Continuity Dogs (Labs)

We simulated the 2025 Hybrid V2 portfolio (hybrid_v2) with and without low-continuity dogs:

Baseline (all bets): 710 bets, 63.2% win rate, +20.38% ROI, +$14,468.70 PnL

Removed subset (low-continuity dogs only): 239 bets, 44.7% win rate, -14.50% ROI, -$3,464.60 PnL

Filtered portfolio (dropping low-continuity dogs): 471 bets, 72.6% win rate, +38.07% ROI, +$17,933.30 PnL

Impact: Removing 239 low-continuity dog bets would have improved PnL by +$3,464.60 and ROI by +17.70 percentage points (from +20.38% to +38.07%).

Observations: The pattern matches the official card: low-continuity dogs were a significant drag on Hybrid V2 performance. Hybrid V2's baseline ROI (+20.38%) was already strong, but filtering low-continuity dogs would have pushed it to +38.07% ROI. The filtered portfolio would have achieved a 72.6% win rate (vs 63.2% baseline).

This is Labs-only evidence and not yet a hard production rule, but it strongly supports the guardrail for both the official card and Hybrid V2 strategies. Avoiding low-continuity dogs in 2026 could meaningfully improve performance across both portfolios.

9. 2026 Guardrail Policy – Low-Continuity Dogs

The official card treats "Low-Continuity Dog" as a hard guardrail by default.

Definition

A Low-Continuity Dog is a spread bet where:

  • The bet team's continuityScore < 0.60 (Low continuity band)
  • AND the bet team is a dog (getting points, based on closing spread)

Historical Evidence

These plays have been a large negative ROI segment in both 2024 and 2025 simulations:

  • 2025 Official Card: 316 low-continuity dogs had -22.04% ROI (vs +13.53% baseline)
  • 2025 Hybrid V2: 239 low-continuity dogs had -14.50% ROI (vs +20.38% baseline)

Policy for 2026

Default behavior:

  • The official card auto-excludes low-continuity dogs by default.
  • This is a risk management rule, not a model feature.
  • The model (Hybrid V2) can still like these games; the guardrail is on portfolio construction, not on the rating engine.

Manual override process:

  • If a low-continuity dog is included in the official card:
    • It must be manually whitelisted with a written handicap (injuries, matchup context, etc.).
    • It should be treated as a Labs-only or reduced-size play.
    • The rationale should be documented for review.

Implementation: Low-continuity dogs are flagged with a red "Low-Continuity Dog" pill on the /picks page. The Portfolio What-Ifs panel (/labs/portfolio) shows the impact of dropping these bets. This guardrail applies to the official card (official_flat_100); Labs strategies may experiment with different filters.

10. Future Model Enhancements (V5+)

We're planning to add Portal & NIL Meta Indices as Labs overlays, with potential integration into a future V5 Hybrid model:

  • Continuity Score: Measures roster stability (returning production + transfer portal activity)
  • Positional Shock Index: Flags teams with extreme turnover at key positions (QB, OL, DL)
  • Mercenary Index: Identifies teams heavily reliant on short-term transfers
  • Portal Aggressor Flag: Flags teams that aggressively use the transfer portal (net talent gain)

These will initially live as Labs overlays to test their predictive value. If they prove stable and additive in backtests, they may be folded into the core Hybrid model in a future V5 release.

See Data Inventory for current data structures, and Bowl & Postseason Ops for the operations side.

11. Disclaimer

All numbers above are based on historical backtests (2024–2025).

Performance can and will regress. The rules here are designed to:

  • Lean into the strongest, most stable buckets we've observed so far.
  • Avoid overreacting to noisy or weak buckets.
  • Keep the UI honest about what is "hammer-worthy" vs "just interesting."