How PickGauge works
Last updated: September 2026
PickGauge turns several college-football data sources into a simpler decision layer: a projected spread, the gap between that projection and the line you are picking against, and supporting context that helps you understand the matchup. The app deliberately separates model inputs from context so extra data does not silently change a recommendation.
PickGauge Model #
PickGauge Model # is PickGauge's default projected spread for a game. It uses a fixed proprietary blend of five selected prediction models plus a smaller live-market component. The exact internal weights are intentionally not exposed in the app.
The branded number requires a live market line and at least 3 of its 5 prediction-model inputs. If one or two model feeds are unavailable, PickGauge redistributes only the missing model weight across the models that are available while keeping the market's intended share fixed. If fewer than three model feeds are available, the PickGauge Model # is incomplete rather than pretending a one- or two-model estimate is the full branded number.
In an ATS pool, the PickGauge Model # remains a projected spread informed by the current market. The pool's locked contest line is then used as the number the model is compared against for the pool-specific Edge and pick decision.
Custom Blend
Custom Blend is optional and user-controlled. You can keep PickGauge Model # as the anchor and add other enabled models, Brad Powers, the Computer Line, or the Market with your own relative weights. When a Custom Blend is active, it becomes the projection that drives Edge and Cover %. If you turn PickGauge Model # off, the enabled weighted inputs form a fully custom Model # instead.
The standalone PickGauge Model # itself does not change when you create a Custom Blend. It stays visible as the branded reference number, which makes it possible to see whether your custom weighting moved the projection and by how much.
Edge and Cover %
Edge is the point difference between the active model projection and the line you are actually evaluating. In No Pool / Market view, that reference is the current market line. In a pool, it is the pool's locked contest line once one exists.
Cover % is an estimated probability that the recommended side covers that spread. PickGauge derives it from historical FBS-vs-FBS cover-margin behavior rather than treating the same raw point edge as equally meaningful at every spread size.
PickGauge also flags proximity to common football scoring margins such as 3 and 7. These key-number indicators are context for how meaningful a point of line value may be; they do not independently create a pick.
CLV and market movement
In pool views, CLV compares the contest's locked line with the live market after that line was locked. Before you make a pick, it is simply market movement. After you choose a side, PickGauge expresses it from your pick's perspective: positive means the current market moved in a direction that makes the number you already hold more favorable.
For historical model-performance grading, PickGauge uses the last pre-kick market snapshot the account actually observed. It does not reconstruct a supposedly official closing line after the fact.
Data sources
PickGauge can display and compare projections from a curated set of independent computer-rating and prediction systems. Many of these projections are obtained through a third-party aggregation source. PickGauge is not affiliated with or endorsed by that source or by the individual model publishers. If you operate one of these systems and have questions about its use here, contact PickGauge.
Live and historical sportsbook spreads used by PickGauge come from The Odds API, a commercial odds-data provider.
PickGauge uses CollegeFootballData.com for schedules, scores, ratings, advanced team statistics, postgame data, and other matchup context. Most CFBD data is context only and does not change Model #, Edge, or Cover %. Specific rating-derived projections such as SP+ or CORE can be enabled explicitly as model inputs; when that happens, the app identifies them in Models & weights like any other enabled model.
The Powers PDF importer can populate the Brad Powers and Computer Line projections. My Numbers are projections entered or imported by the user. My Numbers stay separate from PickGauge Model # so simply entering a personal projection never changes the branded model.
What is context versus a model input?
Matchup Intelligence, team efficiency statistics, descriptive postgame separators, schedules, and most CFBD ratings are shown to help explain a matchup. They do not silently alter the PickGauge Model #. A data source only affects the active projection when it is part of the fixed PickGauge Model # recipe or the user explicitly enables it in a Custom Blend.
Matchup Intelligence uses current-season offense/defense data, separates FBS and FCS comparison groups where appropriate, and is hidden after a game is final so the game's own result is not fed back into what is presented as pregame context.
How model performance is tracked
The Results page separates model performance from your pick performance. When PickGauge has a market line and model projections before kickoff, it can freeze a compact snapshot for an available game even if you never select that game. That allows prediction models to be graded as hypothetical ATS leans without limiting the sample to the user's shortlist.
A projection equal to the comparison line is recorded as no lean. Pushes can appear in W-L-P records but are excluded from the ATS win-rate denominator. PickGauge does not backfill historical model records with projections obtained after kickoff.
What PickGauge computes
- PickGauge Model # — the branded blended projected spread described above.
- Custom Blend — a user-controlled weighted projection when enabled.
- Edge, Cover %, and key-number context — derived from the active projection versus the relevant line and historical FBS spread/margin behavior.
- CLV / line movement — movement from a pool's locked number to the live or retained pre-kick market.
- Survivor strategy metrics — win probability, future-value/scarcity context, exact season-path optimization, and multi-entry diversification logic where available.