Straight answers.

Three kinds of calls, each with its own public record, plus the tools around them:

• Halftime calls (NFL and college football). At the half, sportsbooks post a new 2nd-half line in minutes, mostly off what just happened. Our models call the spots where the book overreacts. These are BET calls we stand behind.
• Opening-line calls (NBA and college basketball). The moment DraftKings opens a line our model disagrees with, you get an alert to bet it before the market moves. Also BET calls.
• Basketball halftime leans. The strongest basketball halftime spots, labeled LEAN, never BET, and graded on their own record.

Around them: the live halftime board for four leagues with DraftKings' 2nd-half line filled in, a daily deep dive with one best bet every morning, a buy points calculator in the Tools tab, phone alerts, injury and weather checks, and a bet log that grades itself.

NFL, college football (FBS), NBA and men's college basketball. Calls are on 2nd-half totals at halftime and on spreads and totals at the opening line. The app also shows the model's spreads, totals, moneylines and NFL/NBA player props as unproven research plays, graded in public but not recommended.

It depends on the sport and the week, because we only call spots that held up in testing:

• College football halftime: about 1–3 calls a week. The model only bets its most confident 5% of halftime spots.
• NFL halftime: a handful a month (Unders when the book raised its number 7+ points).
• Opening-line basketball: about 15–25 a week once both seasons are going (NBA totals, college spreads and totals).
• Deep dive: one best bet every day, plus up to two more when they're strong.

Some nights there's nothing, and that's deliberate. A call is only worth something if it isn't made just to fill a quota.

No trial, but nearly everything that proves the product is free to check before you pay: the full record on the Results page with every win and loss, the method, the halftime calculator and the sport guides. Founding members pay $24.50 a month with code SHARP50, and you can cancel anytime.

Before kickoff the book sets a total for the whole game. At halftime it posts a 2nd-half line. Add the 1st-half points to that line and you have the book's new full-game guess; compare it with the pregame total and you see how far the book changed its mind after one half.

In college football, a model scores every halftime spot using the pregame total and spread, 1st-half points, the lead, how far the book moved and the line itself. It only calls its top 5%: mostly blowouts where the 2nd-half number stays too high because starters sit and the clock runs. Example: Wake Forest led Cal 34–0 at the half, the 2nd-half line was 26.5, the model called Under, and the 2nd half finished with 14 points.

DraftKings' 2nd-half line keeps moving for the first few minutes of halftime. If the model likes the very first line, you get an EARLY alert within about a minute, so you can bet before the number moves. About 4 minutes in, once the line settles, the official call is recorded. That's the one the record counts. Every BET shows the worst line it's still good at. If the line moves past that, the card changes to No bet. Early calls are graded on their own record.

Usually 10–15 minutes after the official call. Halftime runs about 13 minutes in the NFL, 20 in college football and 15 in basketball, and the app shows a countdown on every game at the half. 2nd-half lines come down when play resumes, which is why the EARLY alert exists.

Because in about 51,800 past games, nothing at the basketball half beat the 53.5% you need to break even at −115. The best spots won 51–52%. We show those as LEANs with their real rate and their own record, and we won't call them bets unless that record earns it. Basketball's real edge is at the opening line, so that's where the basketball calls are.

When a sportsbook first posts a line, it's often a little off, and the market spends the next day correcting it. Our model compares every new DraftKings opener with its own number. When they disagree by enough, you get an OPENER alert right away, and the call is recorded with its time and price. In testing, betting those openers won 55.7% in college basketball spreads, 57.5% in college totals and 58.5% in NBA totals. Most of that value was gone by game time.

Once both teams have played four games, so the ratings are built on real results and not just last season. For the NBA that's about two weeks into the season; for college basketball, late November. The model also skips a game when its edge is so large that it probably means news it doesn't have.

Every morning at 9:30 AM ET, Claude (Anthropic's AI) reads everything the app knows about the next day of games: our model's numbers, opening and current lines, which plays the market has already moved toward, injuries and weather. It then researches the strongest candidates on the web: starters, injury and practice reports, rest and travel, and why lines moved. It names one best bet every day, plus up to two more when they're strong, each with its reasoning, the facts behind it, the risks, and a buy points suggestion where it helps. You get an alert when it's live.

Good looks are in the Picks tab: 4 to 8 plays a day that we like but that don't clear the Deep Dive's bar. They include at least 2 team picks (a team to cover, a total or a moneyline) plus player props, in the NFL, NBA and NHL, at any price. Each one starts from recent numbers, the team's last five games or the player's game log against the line, then checks who's out, the matchup and game conditions. They're 1 unit each. The parlay of the day is one 2 to 4 leg long shot for 0.5 units, the "hey, maybe this hits" play. Neither has a proven edge, and books take a bigger cut on props and parlays. They have their own record and aren't in any other record. A prop is voided if the player doesn't play. A parlay leg that pushes or is voided drops out and the price is redone.

It has its own record, graded automatically and checked against the closing line, but it isn't added to the halftime or opening-line records. It's new and unproven, so you can judge it on its own numbers.

Most books let you move a spread or total in your favor for a higher price: Under 44.5 at −115 might become Under 47.5 at −160. You win more often, but each win pays less. Buying points is a good bet only when the book charges less than the extra points are worth. In football, moving through 3 or 7 is worth the most, because so many games end on those margins.

It's in the Tools tab, and on pregame calls. It lists the bought lines, how often each wins, and the most each is worth paying. Tap the line you want, type what your book charges, and it answers Worth it or Not worth it, and compares that with betting the line as is. For pregame games it also shows the market's view next to our model's, because the market is usually right by game day.

Yes. Every halftime call is also graded as if 3 points were bought, and that record sits next to the main one on the homepage. When members type real prices into the calculator, we save them, so over time we learn what books actually charge for 2nd-half points.

Every call is timestamped when it's made: at the half for halftime calls, at the opener for opening-line calls, before kickoff for pregame plays. Every call grades itself automatically from the final score at the price posted. Losses stay on the board, and nothing is edited or removed. The full ledger is on the Results page.

So one product can never prop up another. Halftime calls, opening-line calls and basketball leans each have their own record, and "All bets" combines only the two kinds of bets. Leans, early calls, the deep dive and research plays are tracked separately and never counted as bets.

We updated the halftime method on October 19, 2026. The record opens with the three calls the updated method would have made on October 15–17, replayed after those games with the model's settings frozen, and tagged "Replay" everywhere they appear. Everything from October 19 on was called live. Calls made under the old rules stay public on the Results page under Retired rules.

Live tracking started in October 2026. We'd rather show a short record that's real than a long one we can't prove. Judge any betting record over hundreds of plays, not a good or bad weekend. A faster signal is the closing line: if our number keeps beating the market's final number, the edge is real long before wins and losses settle. The app tracks that for every rule.

Halftime calls at −115, the extra juice books usually charge at the half (break-even 53.5%). Pregame calls at the price posted when they were made. Each call counts at its unit size: halftime BETs are 5 units, the deep dive's best bet of the day 3 units, and every other call (halftime leans, early calls, opening-line calls) 1 unit. Pushes are no action.

On iPhone, add HalfSharp to your Home Screen first (tap Share, then Add to Home Screen), open it from there, and tap Turn on alerts. Apple only delivers web alerts to Home Screen apps. On Android and computers, just tap Turn on alerts. You can choose your leagues and whether you want every game at the half or only BET calls. If alerts are quiet, check Settings › Notifications › HalfSharp, turn on sounds and allow it in Focus modes.

Any legal sportsbook in your state. The app fills in DraftKings' lines automatically. If your book's number is different, type it into the game card and everything recalculates, including the buy points calculator.

Yes. Sign in with your username and PIN on each device, and turn on alerts on each one you want buzzing.

The ratings model itself doesn't, so the app checks. Before every pregame play it reads injury reports, starting quarterbacks, each team's key players, inactive lists and weather. A play is pulled automatically when a starting quarterback, a basketball team's best player, or two key football players are out. The deep dive researches the news on top of that.

$49 a month. Founding members who use code SHARP50 at checkout pay $24.50 a month, and that rate stays as long as the membership stays active.

Open Settings in the app, tap Manage billing, and cancel on Whop. No phone call, no email. You keep access through the period you've already paid for.

Whop processes every payment. Your card details go to Whop, not to us; we never see or store them. Questions about a charge? Contact us.

No. No call, method or record is a guarantee, and past results don't predict future results. Even a real edge loses often: a 58% bet loses 42% of the time. Keep a unit small (around 1% of your bankroll, so even a 5-unit call is a modest bet) and judge results over months, not days.

No. HalfSharp is an analytics and information service run by Precision Sports Mathematics. We don't accept wagers and aren't affiliated with any sportsbook.

Adults 21 and older, where sports betting is legal. If gambling stops being fun, call or text 1-800-GAMBLER. More on our responsible gambling page.

We use your email for your account and, if you opt in, our updates. Analytics cookies stay off unless you allow them, and we use no advertising cookies. We also record anonymous clicks and scrolling, without cookies or anything you type, to improve the site. Details are in the privacy policy.