OUR PICK STANDARDS
Know what you are paying for.
A selective game-market research board—not a promise of daily winners. Membership gives you access to analysis, not a guaranteed number of bets.
Qualified is a checklist, not a guarantee
Qualified selections must clear the applicable model checks, have at least three matching sportsbook quotes and at least +2.5% estimated expected value. Favorite prices cannot be shorter than -130. Plus-money selections must still clear the model screens. SharpSignal’s best-pick feature chooses only from qualified, upcoming selections; it returns no pick when none qualify.
A grade is not a win probability
Grades summarize estimated expected value: A+ at 6% or above; A at 4.5% to below 6%; A− at 2.5% to below 4.5%; B+ at 1% to below 2.5%; B at 0% to below 1%; C+ at -2.5% to below 0%; C below -2.5%. A high grade can still be a lean if other checks fail. Estimated probability is shown separately.
Model versus market
MLB, CBB, NBA, WNBA, NHL and UFC moneylines use statistical projections when sufficient inputs are available. CBB projections combine CollegeBasketballData.com team-season statistics and adjusted efficiency with hoopR/SportsDataverse opponent-adjusted ratings, pace, four factors and neutral-site checks. CBB selections require at least eight completed games per team, both independent sources, ratings agreement within eight points per 100 possessions and three matching books to qualify. NBA projections use licensed SportsDataIO team-season scoring, opponent defense, pace and four-factor efficiency, plus projected and confirmed starting-lineup data; estimates are blended conservatively with the no-vig sportsbook market. NBA selections require at least eight completed games per team, both five-player lineups confirmed and three matching books to qualify. WNBA projections use WeHoop possession-adjusted offense and defense, recent form, pace, effective shooting, turnovers, rebounding, free-throw rate, home/away splits, rest and player-impact availability; estimates are blended conservatively with the no-vig sportsbook market. WNBA selections require at least eight completed games per team, verified availability and three matching books to qualify. NHL projections use fastRhockey team box scores, blending current and prior-season scoring, shot generation and suppression, shooting and team save rates, recent form, special-teams goal margin, possession proxies and home/away performance. NHL selections require at least eight current-season games per team, three matching books and both starting goalies confirmed through the licensed SportsDataIO goalie feed to qualify. UFC moneyline projections use a regularized, symmetric model trained on pre-2024 TidyTuesday bouts and checked chronologically against separate 2024 and 2025–2026 samples. Inputs include record, experience, age, height, reach, significant-strike output and accuracy, takedown and submission rates, and finish profile. The statistical probability receives only a 10% weight against the multi-book no-vig market because it did not decisively outperform sportsbook prices by itself. Both fighters need at least five recorded fights, at least 72% paired-field coverage, a fighter snapshot no older than 180 days and three matching books for a moneyline to qualify. Qualified NFL selections use nflverse EPA per play, offensive and defensive efficiency, scoring and recent form, pace, turnover margin, rest and schedule-weather context after both teams have at least four completed games. Qualified CFB moneyline and spread selections use CollegeFootballData pregame win probability, SP+ and CORE when available; current Elo and completed-game scoring form provide the supported fallback after both teams have a meaningful sample. CBB, NBA, NFL, NHL, UFC and CFB estimates are blended conservatively with the no-vig sportsbook market. CFB totals and UFC round totals remain sportsbook-consensus research and cannot qualify. Additional matchup context does not change a probability unless explicitly stated.
Prices have a timestamp
The board uses shared feed snapshots, typically refreshed hourly, while your page checks for updates every five minutes. These are not streaming or guaranteed executable prices. Compare the exact selection, number and price with your sportsbook before betting. A changed line changes the value. Provider outages may leave only comparison markets or leans.
No forced picks
Leans are research only and are not promoted by the best-pick feature. AI recommendations stop five minutes before the listed start. UFC providers may supply a shared card start rather than individual bout times; once that cutoff passes, the AI does not assume an unconfirmed bout is still pre-fight.
Performance evidence
Members can view the new results history and download the complete record. It saves the first qualified featured pick per event and model version, starting at the ledger’s launch. One unit is risked per recorded pick. Pushes count in the ROI stake total; pending and review records do not. Earlier tracking is excluded. Recording happens on member-triggered board refresh, and final-score checks run when results are opened. UFC and ambiguous results remain unresolved rather than being assigned a win or loss.
A profitable live record for this version has not been established. Historical score-only research—including the UFC chronological holdout—is model validation, not a betting record, and does not establish returns at available odds. We will not present those results as member winnings. Demonstrating an edge requires recorded pre-event selections, actual prices, settled results and out-of-sample evaluation—including losses.
WNBA statistical and player-impact data is derived from WeHoop/SportsDataverse release datasets and modified into Sharplytics projections. WeHoop is licensed under MIT and does not sponsor or endorse Sharplytics.
NBA statistical and lineup inputs are requested from the licensed SportsDataIO NBA feed. If that NBA entitlement is unavailable or the data fails plausibility checks, NBA events may still show verified sportsbook comparison lines and clearly labeled leans, but they cannot become qualified NBA picks. Sharplytics does not ingest nba_api, nbastatR, the Kaggle NBA Database or the awesome-nba-data collection because those tools and datasets trace to NBA.com statistics, whose terms restrict gambling and commercial product use without permission.
CBB statistical data is provided by CollegeBasketballData.com under terms that permit commercial applications, and opponent-adjusted ratings are derived from the hoopR/SportsDataverse MBB releases under CC BY 4.0. Both are modified into Sharplytics projections and neither provider sponsors or endorses Sharplytics. The CC0 Kaggle College Basketball Dataset is retained as a historical research reference, not a live pick input, because its snapshot is not current enough for today's markets. Google BigQuery's NCAA basketball dataset is not ingested because its NCAA/Sportradar terms limit it to internal research and testing and prohibit commercial use without approval.
NFL statistical data is derived from nflverse data, licensed under CC BY 4.0, and modified into Sharplytics projections. nflverse does not sponsor or endorse Sharplytics.
UFC historical bouts and fighter profiles are derived from the R for Data Science TidyTuesday UFC release and modified into Sharplytics moneyline projections. The TidyTuesday repository is distributed under CC0 1.0. R for Data Science, TidyTuesday, fightr and UFC do not sponsor or endorse Sharplytics. Sharplytics does not run the unofficial Dave's Analytics UFC scraper against UFCStats.
NHL statistical data is derived from fastRhockey/SportsDataverse team-box-score release datasets and modified into Sharplytics projections. The SportsDataverse release repository is distributed under the MIT License and does not sponsor or endorse Sharplytics. Projected and confirmed NHL starters are checked through SportsDataIO. GoaliePost can be used as a manual secondary check, but its webpage is not scraped or treated as an automated Sharplytics data feed. MoneyPuck data is not ingested because its free-data permission is limited to non-commercial use; a separately approved commercial license would be required before it could be activated on this paid service.
Betting involves the risk of losing your stake. Do not chase losses or treat a grade as certainty.
View membership