Data freshness and sample size
Check the period covered, games included, and update time when provided. Data can be delayed. Missing injury or opponent information should lower how much you rely on that part of the analysis.
UNDERSTAND YOUR NBA RESEARCH
Read the numbers.
Know what they mean.
Start with a player prop or a game matchup. Follow the examples below, then explore the details that matter to your research.
3
Steps: choose a matchup, enter your line, and review the estimate. All figures on this page are teaching examples.
01
PLAYER PROP EXAMPLE
Compare a player’s projection
with your prop line.
A player prop is a bet on a player’s performance. In this sample, we are researching points scored in one game. These numbers are not a current prediction or betting recommendation.
- 01
Choose the player and opponent
Select the player’s team, opponent, and player. Confirm the matchup you want to research.
- 02
Choose points and enter 29.5
29.5 is the prop line: over means 30 or more points, and under means 29 or fewer. Enter the line you are considering; it can change before tipoff.
- 03
Read the projection and supporting data
Compare the estimated points with your line, then review recent results, opponent history, and any missing information.
WORKED EXAMPLE · POINTS · NOT LIVE DATA
- Your prop line
- 29.5 points
- Model projection
- 31.8 points
- Projection minus line
- +2.3 points
31.8 − 29.5 = +2.3 points. The estimate is above the line. That does not tell you how likely the over is to win or whether the offered odds are good value.
The decimal in a projection represents an estimated average. An actual points total is a whole number. Even when the projection is above the line, the final result can finish below it.
02
GAME PREDICTION EXAMPLE
Read the winner,
score, and margin separately.
A projected winner is the team the model favors. A projected score estimates how many points each team might score. Neither is a final result.
WORKED EXAMPLE · FICTIONAL TEAMS · NOT LIVE DATA
- Team A projection
- 112 points
- Team B projection
- 108 points
- Projected margin
- 4 points for Team A
Team A is favored in this example because 112 is greater than 108. The projected combined total is 220 points (112 + 108).
Winning and covering differ
A moneyline bet concerns who wins. A spread bet uses a points handicap. At Team A −5.5, Team A would need to win by at least 6 points to cover, even though a smaller win would still win the game.
A projection is not a sportsbook line
The model’s 4-point margin and 220-point total are estimates. Compare them with the actual spread or total you are researching.
Probability needs a definition
An estimated 60% win probability would mean about 60 wins in 100 comparable situations if the model is well calibrated. It is not a measured 60% historical success rate or a guarantee for this game.
03
BETTING TERMS, EXPLAINED
Common terms.
Plain meanings.
Open a term for a short explanation. You can skip ahead to the analysis checklist if you already know the basics.
Player prop, prop line, and over/under
A player prop is a bet on a player’s result, such as points or rebounds. The line is the number you compare that result with. Over 29.5 points requires 30 or more; under requires 29 or fewer. A whole-number line can tie the final result, so check the sportsbook’s settlement rules.
Projection, difference, and model lean
A projection is an estimated result. Here, difference means projection minus the line, in the same units as the statistic. A positive difference is above the line; a negative one is below. A lean describes the side the analysis favors, without promising it will win.
Odds and implied probability
Odds describe the potential payout. For American odds, −110 means risking $110 for $100 profit; +150 means risking $100 for $150 profit. A winning cash bet also returns the stake. Implied probability converts that price into a break-even rate before adjustments for the sportsbook’s margin. At −110, it is about 52.4%.
Moneyline, spread, and total
Moneyline concerns which team wins. A spread adds or subtracts points for bet settlement. A total concerns the teams’ combined score. Winning the game does not necessarily mean covering the spread.
Bet slip, parlay, leg, and same-game parlay
A bet slip lists your selections. A parlay combines selections into one bet; each selection is a leg. A 3-, 4-, 5-, or 6-leg parlay combines that many selections. A same-game parlay combines selections from one game. Generally, every leg must win, subject to the sportsbook’s void and push rules. More legs add more ways for the bet to lose.
Points, boards, assists, and threes
Boards means rebounds. Threes means made three-pointers. Common labels include PTS (points), REB (rebounds), AST (assists), and 3PM (made three-pointers). Check whether a market counts attempts or makes.
Recent form, sample size, and head-to-head
Recent form means recent performance. Sample size is how many games or observations the summary includes. Head-to-head refers to previous meetings with the opponent. A small sample can give a misleading picture of what is typical.
Confidence score versus win probability
A confidence score summarizes model support or data quality according to that model’s method. A win probability estimates the chance of a defined outcome. Do not treat the two as interchangeable. A confidence score of 82 does not, by itself, mean an 82% chance of winning your bet.
04
FOR DEEPER ANALYSIS
Check the evidence
behind the estimate.
Use this checklist for a single pick or to research each leg of a parlay. Individual player projections do not provide a combined parlay win probability.
Minutes and opportunity
Consider expected playing time, shot opportunities, lineup changes, and injuries. A change in a player’s role can make older averages less representative.
Price and uncertainty
A projection-line difference is not expected profit. Assess the offered odds and the chance of clearing the line. If a range is shown, check what it represents before treating it as a probability interval.
Related parlay legs
Selections in the same game can affect one another. A player’s scoring and the team’s total, for example, may be related. Do not multiply individual probabilities as though related outcomes were independent.
How to evaluate model performance
Ask which predictions the record includes, how many there are, the date range, and how outcomes are scored. Compare predictions made before games with results from games the model did not train on. A few successful picks are not enough to establish reliability.
Calibration checks whether estimated probabilities match observed results across many comparable predictions. For example, predictions labeled 60% should succeed about 60% of the time in a well-calibrated model. This is an explanation of the test, not a claim about Hoopsline’s measured performance.
Accuracy and return differ. Accuracy measures correct predictions. Betting return also depends on prices and stakes. This site does not present verified profitability or a live performance record.
05
HOW THE ANALYSIS WORKS
From basketball data
to an estimate.
- 01
Gather available information
Team and player statistics, recent results, previous meetings, and available injury information provide context for the matchup.
- 02
Calculate the projection
Models look for patterns in the data to estimate a game result or player statistic. Game predictions and player props answer different questions and can use different methods.
- 03
Review the result in context
Compare the estimate with the line you entered and read the supporting information. Check current odds and team news separately before making a decision.
Hoopsline provides basketball research. You cannot place bets on this site, and predictions do not guarantee outcomes.
RESEARCH YOUR NBA PICKS BEFORE TIPOFF
Bring more context
to your next pick.
Join the early access waitlist for launch updates and access details.
Join early access →