Frequently Asked Questions
About Our Predictions
What is SharpBetz?
SharpBetz is a free sports prediction site that uses machine learning to generate data-driven picks for college basketball (NCAAB) and NBA games. Every prediction is produced by our XGBoost statistical model, not human opinion. We publish every pick transparently, wins and losses alike, on our results page.
How are predictions generated?
Our predictions are powered by an XGBoost gradient-boosted decision tree model trained on thousands of historical games. The model evaluates 27+ features including scoring differentials, efficiency metrics, situational factors, strength of schedule, and market signals to project spreads, totals, and win probabilities. For a full breakdown, visit our How It Works page.
What sports do you cover?
We currently cover NCAA Division I men's college basketball (NCAAB) and the NBA. We plan to expand to additional sports in the future. You can browse all available predictions from the NCAAB and NBA sport pages.
Are your picks free?
Yes, all predictions, analysis, and historical results on SharpBetz are completely free. We monetize the site through display advertising, so you never have to pay for access to our picks or track record.
How often are predictions published?
Predictions are published daily during the active season for each sport. We typically generate picks in the morning for that day's games. The number of predictions depends on the schedule — during a busy college basketball Saturday, we may publish 50+ predictions.
Understanding Our Picks
What do the confidence percentages mean?
The confidence percentage reflects how strongly our model's projection disagrees with the market line. A higher confidence means the model has identified a larger edge — a greater discrepancy between what our model expects and what the sportsbooks are offering. Historically, our higher-confidence picks have outperformed lower-confidence ones.
What are unit sizes (1u, 2u, 3u, etc.)?
Unit sizes represent the relative strength of a pick. A "unit" is a standardized bet size — for example, if your unit is $10, a 3-unit pick would suggest a $30 wager. Higher unit picks (3u, 4u) indicate larger edges detected by the model, while 1u picks represent smaller leans. See our glossary for more betting terminology.
What does "play to" mean?
"Play to" indicates the maximum line at which our model still sees value. For example, if we recommend "Duke -5.5, play to -7," that means the pick is worth taking at any spread up to Duke -7, but not beyond. If the line has moved past the "play to" number, the edge may no longer exist.
What's the difference between the "Actual" and "Projection" columns?
The "Actual" column shows the current market odds and lines from sportsbooks. The "Projection" column shows what our XGBoost model predicts the line should be. The gap between these two is the edge — when our model's projection differs significantly from the market, that's where we find value picks.
Why do some picks say "Pass"?
A "Pass" means our model did not find a meaningful edge on that game. The model's projection is close enough to the market line that there is no statistically significant value on either side. We believe it's better to pass on a game than to force a pick without conviction.
Track Record & Results
How can I verify your track record?
Visit our results page to see every prediction we have ever published, along with the actual outcome. We display win/loss records broken down by sport, bet type, confidence tier, and time period. Every pick is timestamped and recorded before game time — nothing is added or modified after the fact.
What is your overall win rate?
Our win rate varies by sport, bet type, and confidence tier. We aim for a 55-60% win rate against the spread (ATS), which is considered elite in sports betting. Check our results page for current, up-to-date performance figures across all categories.
How is ROI calculated?
ROI (Return on Investment) measures profitability as a percentage of total units wagered. For example, if you wagered 100 total units and profited 5 units, your ROI would be 5%. We calculate ROI using standard -110 juice on spread and total bets. A positive ROI means the picks are profitable over time; even a 3-5% ROI is strong for sports betting.
Do you delete losing picks?
Absolutely not. Every prediction is permanently recorded in our system the moment it is published. We never delete, hide, or retroactively edit losing picks. Transparency and accountability are core to our mission. You can verify this by checking any historical date on the results page.
Using the Site
How do I find today's picks?
The easiest way is to visit the homepage, which displays today's predictions front and center. You can also navigate to a specific sport page like NCAAB or NBA to see today's games for that sport. Use the date picker at the top of any sport page to browse predictions for other dates.
Can I filter by sport or confidence level?
Yes. On the sport pages, you can filter predictions by confidence tier to focus on our strongest picks. You can also use the sport navigation in the header to switch between NCAAB and NBA. The results page offers additional filters for analyzing historical performance by sport, bet type, and date range.
What timezone are game times shown in?
Game times are displayed in your local timezone as detected by your browser. The underlying data is sourced from ESPN, which provides times in Eastern Time (ET). Your browser automatically converts these to your local time, so you always see the correct start time for your location.
How do I switch between American and decimal odds?
Use the odds format toggle found at the top of prediction pages. Click the toggle to switch between American odds (e.g., -110, +150) and decimal odds (e.g., 1.91, 2.50). Your preference is saved in your browser so it persists across visits. If you're unfamiliar with odds formats, check our glossary for explanations of each format.