Edge is not a tout service. It is a set of models, each with a written rule for when it is allowed to make a pick, and a log that records every pick the moment the rule fires. This post is the overview: what each model measures, whether it is on the card or still proving out, and the record behind its trigger band. The live numbers are always on the Our Models page. This is the story behind them.
One rule for every model
Before a model makes its first live pick, its trigger is written down with a date. From that date, every play it makes is logged at the line posted at that moment and graded against that line. If the market moves toward us afterward, that shows up as closing line value, not as a win. If the model has a cold month, the cold month stays in the record.
That sounds obvious. It is also the thing most pick sellers do not do, because it means publishing the losses.
College football: three models, one card
Edge Power is the model that publishes. It rates every team with one overall strength number and prices each game from the gap between the two. A pick publishes when that price sits far enough from the market spread to clear the rule. Picks are tiered by the size of the gap: Good, Strong and Elite.
Why publish the tiers? Because this season the biggest disagreements have not been the best bets. That is useful to know, and it is only useful if it is visible.
Edge Snap CFB rates the plays instead of the team. It is built from every snap: how efficiently a team moves the ball and stops it, adjusted for who it played. It runs beside Edge Power under its own rule, and its plays are logged exactly like published picks. Its trigger is the smallest gap that held up across seasons at lines a bettor could actually get.
Edge Scout is a full scouting report, dozens of team traits weighed against the number the market posted on Sunday. Its rule is the most specific of the three: it fires only inside a band of disagreement, not above it. The biggest disagreements looked great at one line time and fell apart at another, so they are left alone. The rule was registered before a single live result existed. The live record is the test.
College basketball: Edge Tempo
Edge Tempo goes live on November 3. It is a possession-based model: points per possession on both ends, shooting, turnovers, rebounding and pace, adjusted for schedule. Every morning at 10 AM Eastern it prices the day's games against the line on the board at that moment, and a pick publishes when the gap clears the rule.
The four seasons on the records page, 2022-23 through 2025-26, were priced the same way: every pick graded at the line that was actually posted that morning, with the closing line checked afterward. The edge survived the close, which is the test that separates a real pricing edge from a timing trick. The trigger was raised before the first live pick, because the smaller gaps were a coin flip across all four seasons and the larger ones were not. Bigger gaps won more, and the model's underdog plays won more than its favorite plays, so both splits are published.
NFL: one on the card, two in the Lab
Edge Power publishes NFL picks only when its number sits a long way from the market. The NFL market is the sharpest there is, so the bar is high on purpose. Fewer picks, but none of them are coin flips dressed up as edges.
Edge Snap NFL is the NFL version of the per-play model, with the starting quarterback accounted for. It runs every Monday morning against the first full market line of the week. In testing, with no same-day information allowed, it only leaned slightly ahead of the market. So it is in the Lab to learn, not on the card.
Edge Drift does something different. It does not predict who covers. It has studied years of past games and how their lines moved from Monday to kickoff, and from that it predicts where this week's line will go. It fires when it expects a meaningful move toward one side, and the play is to take the current number before it shortens. Beating the closing line is the most reliable sign of a real edge, and this model is a direct attempt to do it on purpose. Both NFL Lab models are graded in the open, and their plays appear in the subscriber channel from time to time as previews.
Baseball, basketball and golf
The MLB model prices run-based team ratings, with the starting pitcher priced in, against the moneyline, run line and total, and every pick is graded at the real price that was available. The NBA model prices pace and efficiency against the total, one pick per game at most. The golf model compares tournament probabilities for every player to every book's price across outrights, placings, matchups and three-balls, with the card built Tuesday and Wednesday and settled Monday.
How to read any of it
- ROI first. Win rate only means something next to the price. At -110, break-even is 52.4%.
- Graded at the trigger line, flat one unit, pushes count for nothing.
- Seasons before a model went live were priced at the line posted that morning, never at the close. Where that was not possible, the page says so.
- Every settled pick is public on the sport's performance page. Open picks are for subscribers.
The Our Models page keeps the live numbers. The Records page has every season. If you want the picks as they publish, they go to subscribers by email, text and Telegram the moment a model fires.