Research
The opening-line edge.
Our model often disagrees with a new line, and the market usually moves its way. The catch: by game time, most of that value is gone. That's why HalfSharp's basketball calls go out the moment a line opens.
Precision Sports Mathematics · Nov 14, 2026
The short version
- At the opener, four markets passed every test: college basketball spreads (55.7%) and totals (57.5%), NBA totals (58.5%) and college football totals (64.9%).
- At the close, most of it is gone. The market moves toward the model on most plays and finishes the job before tip-off.
- Eight markets didn't pass, including NBA spreads and every moneyline.
- So timing is the product. An opening-line call is only worth it if you bet it soon after the alert.
These are backtest results at −110: hypothetical, not results of actual bets. The live record is on the Results page.
Why we looked at the opener
Two things didn't add up. A moneyline the model liked went from +180 when it was posted to +124 by game time. And a college football Under rule won 55.6% against the opening line but only 45.6% at the close. The model seemed to be right about where lines were going, just too late to bet them. So we asked directly: does the same, unchanged model beat the opening line, on seasons it was never tuned on?
How it was tested
- Written down first, on Oct 24, 2026, before any result was computed. Same model, same thresholds as the app.
- Two unseen periods: the archive seasons (roughly 2014–2022) and 2023 onward from a second data source.
- A look-ahead guard. An opener can be stale only because it was posted before news the model already knew. So every play was re-run with the model frozen 8 days (football) or 2 days (basketball) before the game.
- Strict pass mark, adjusted for testing 12 markets at once, so one lucky result can't sneak through.
What passed
1,785 plays, about +6% return. Beat the close 61% of the time. The sturdiest result: it wins even when the line barely moves.
1,657 plays, about +10%. Weaker from 2024 on (53.5% on 312), where only the big-move plays won.
557 plays, about +12%. 54.9% with the lagged model. Nearly all the profit came from totals that later moved 6+ points.
211 plays, about +24%. Rare, about one a week, and only 30 plays from 2024 on.
With the model frozen days early, all four stayed above break-even (55.6%, 57.9%, 54.9% and 60.7%). So the edge isn't hindsight.
What didn't pass
- NBA spreads: 50.6% on 1,987 plays. Nothing.
- NFL totals: 52.4% on 168. The line moves our way, but the games don't follow.
- College football spreads: 55.0% on 471. Above break-even, but not by enough to be sure.
- Moneylines in college football, the NBA and college basketball: about break-even. Prices move our way, but not enough to cover the book's cut.
- NFL spreads: 58.2% on 256, but only one clean period of data. Promising, not proven. NFL moneylines couldn't be tested at all.
That's why HalfSharp makes opening-line calls only in college basketball spreads and totals and NBA totals, and none in the NFL.
Why the big moves carry the profit
For totals, the money comes from openers that were far off. In NBA totals, plays where the total later moved 3 points or less won only about 52–55%. Plays where it moved 6 or more won 72–78%. The model knows where the number should be, and some openers aren't there yet. College basketball spreads are the exception: they win even when the line barely moves, which looks like plain model skill.
One honest limit: openers carry low betting limits, some move within minutes, and old data can't tell a real, bettable opener from a recording error. That's why every live call is recorded with its time and price, and graded against the closing line as well as the score.
What this means for members
- Bet soon after the alert. The value is at the number in the alert, not at tip-off.
- Expect the line to move. If it already moved past the call by the time you look, the edge may be gone. That's the system working, not failing.
- Judge it by the closing line first. If our number keeps beating the close, the edge is real long before wins and losses settle.
Related research
- The college football halftime model
- What we tested and threw out
- Month one: the live record so far
- All research
See every live call on the Results page, or how the whole system works on How it works.
Backtest figures are hypothetical. They are not results of actual bets, and past results don't predict future results. No call is guaranteed. HalfSharp is an analytics and information service for adults 21+; we don't accept wagers. Gambling problem? Call 1-800-GAMBLER.