Game-by-game CLV picks fire when Pinnacle posts a spread and a retail book offers meaningfully better odds on the same side. Minimum threshold: +3.0% EV. Pinnacle anchors all calculations — no Pinnacle line, no pick.

Kraken Pure — team strength on a neutral field. Five signals, all normalized before blending so raw scales do not distort weights. Market signals (Pinnacle, Kalshi) have zero influence here. QB adjustment is applied at the spread prediction layer only.

All 24 teams where both Pinnacle and at least one retail book are actively pricing. Sorted by Pinnacle Fair % (sharpest read on true championship odds). Positive EV = retail is more generous than Pinnacle implies. Fair probabilities use proportional normalization · 30.0% vig removed.

How Lockness CFB works — three connected but separate layers. Market prices do not influence power ratings. They are comparison benchmarks used after the model generates its own independent ratings.

Pinnacle is the sharpest sportsbook in the world. Their closing line is the best available predictor of true game probability. When retail books offer meaningfully different odds on the same outcome, that gap is exploitable expected value. Lockness does not pick winners. It finds the gap.

Kraken Pure measures expected point margin against an average FBS team on a neutral field. It uses only football signals — no market prices, no championship odds, no public perception. Pinnacle and Kalshi are not inputs here. They appear later as external benchmarks to compare against.

Championship probability is not the same as team strength. A futures price incorporates schedule difficulty, conference path, playoff seeding dynamics, brand value, and market liquidity — not just how good a team is. Using Pinnacle or Kalshi as inputs to the power rating creates a circular problem: market prices would influence Kraken, and Kraken would then be compared against those same prices to identify value. The comparison would be testing the market against itself.

Instead, Kraken Pure produces an independent rating. That rating feeds into a schedule simulation (Layer 2) to generate an independent championship probability. That probability is then compared against Pinnacle and Kalshi as external benchmarks (Layer 3). The layers are connected but separate.

Takes Kraken Pure ratings and simulates the full 2026 season across 10,000 iterations: schedule, home-field advantage, conference title race, tiebreakers, CFP seeding, and playoff bracket. Produces an independent championship probability per team. This is compared against Pinnacle and Kalshi as benchmarks — never as inputs.

Compares Pinnacle's no-vig fair probability against retail books. Fair probabilities use proportional normalization across all 24 Pinnacle championship selections (30.0% vig removed). Strong Flag: EV ≥15% AND probability gap ≥0.75pp. Watchlist: EV ≥5% AND gap ≥0.25pp.