Reading a prediction
What you see on a match page
Each match page shows a structured breakdown of the model's output. Here is what each element represents.
- The Pick
- The model's recommended outcome — which team wins, covers the spread, or whether the total goes over or under. This is the primary output of the prediction engine.
- Win Probability
- The estimated probability that each team wins the match, expressed as a percentage. The two values do not sum to 100 because the model accounts for the bookmaker's built-in margin.
- Projected Score
- The model's point-total estimate for each team at the end of regulation. Used alongside Win Probability to give a concrete expected outcome.
- Confidence Rating (1–5)
- A 1-to-5 scale indicating how strongly the model favors its pick. A rating of 5 signals the model sees a clear edge; a rating of 1 or 2 means the outcome is genuinely uncertain.
- Model Uncertainty
- A qualitative flag — Low, Medium, or High — reflecting the variance in the model's input signals. High uncertainty means the inputs are noisy or conflicting; treat the pick as more speculative.
Glossary
Key metrics defined
Efficiency metrics
These rate how efficiently each team scores and defends, normalized per 100 possessions to remove the distortion of game pace.
- Offensive Rating (ORtg)
- Points scored per 100 possessions. A higher number means a more efficient offense.
- Defensive Rating (DRtg)
- Points allowed per 100 possessions. A lower number means a more efficient defense.
- Net Rating
- Offensive Rating minus Defensive Rating. The single best summary of a team's overall efficiency.
- Pace
- Estimated number of possessions per 48 minutes. High-pace teams play faster and inflate raw point totals.
- True Shooting % (TS%)
- Shooting efficiency that accounts for two-point field goals, three-point field goals, and free throws in a single number.
- Effective Field Goal % (eFG%)
- Field-goal percentage adjusted to reflect that a made three-pointer is worth 50% more than a made two-pointer.
- Turnover Rate (TOV%)
- Turnovers committed per 100 possessions. Lower is better.
- Offensive Rebound Rate (ORB%)
- Percentage of available offensive rebounds a team secures. Directly affects second-chance scoring opportunities.
- Free Throw Rate (FTr)
- Ratio of free throw attempts to field goal attempts, measuring how often a team gets to the line.
Four Factors
Statistician Dean Oliver identified four factors that explain the majority of winning and losing in basketball. Courtdeep uses this framework to compare teams head-to-head.
- Effective FG% (eFG%)
- The most important factor. Shooting efficiency, weighted for three-pointers.
- Turnover Rate (TOV%)
- Second most important. Turnovers are direct gifts of possessions to the opponent.
- Offensive Rebound Rate (ORB%)
- Third. Second-chance points are high-value and low-cost for the offense.
- Free Throw Rate (FTr)
- Fourth. Getting to the line is a reliable, defense-proof scoring method.
Prediction quality signals
These indicators help you assess the model's conviction and the betting context around a match.
- Star Rating (0–5)
- An overall quality score for the prediction. More stars indicate a larger calculated edge relative to the available market odds.
- Edge %
- The percentage-point gap between the model's estimated win probability and the fair implied probability derived from the odds. Positive edge means the model sees value.
- Fair Implied Probability
- The model's estimated true probability of the pick, with the bookmaker's overround removed.
- Overround
- The margin a bookmaker builds into the odds — the reason the two implied probabilities sum to more than 100%.
- Expected ROI
- The model's projected return per unit wagered, given the current odds and the estimated edge. Not a guarantee of profit.
- Kelly Fraction
- A staking formula output suggesting an optimal fraction of a bankroll to allocate to this bet, based on edge and odds. The quarter-Kelly variant is shown for conservative risk management.
- Confidence Interval
- The score range within which the model expects the true result to fall with a given level of statistical confidence.
Contextual indicators
These signals add qualitative context on top of the core efficiency numbers.
- Regression Signals
- Indicators of whether a team is currently performing meaningfully above or below their statistical baseline. Hot, Cold, or Neutral — used to adjust projections for recent trend momentum.
- If-Then Scenarios
- Conditional game states the model identifies as high-leverage — for example, what typically happens when a team plays at an unusual pace, or when a key contributor is unavailable.
- Consensus Check
- A comparison of the model's output against public betting sentiment. Divergence between model and public money is sometimes a signal of value.
Step by step
How to read a match page
Model Pick panel
Start at the top-right card. The pick and confidence rating give you the model's primary conclusion in two numbers.
Win Probability
The horizontal meters show each team's estimated win probability. Use this alongside the pick — a 58/42 split is very different from 72/28.
Projected Score
The estimated point totals for each team at full time. Compared against the bookmaker's over/under line, this tells you whether the model leans high or low on total points.
Four Factors comparison
The radar chart and edge bars show which team has the structural advantage in each of the four key winning factors.
Efficiency table
A full head-to-head of advanced efficiency numbers — ORtg, DRtg, Net Rating, pace, shooting splits, and rebounding.
Market context
The Star Rating, Edge %, and betting math section show whether the model sees quantifiable value at the current odds.
Regression signals and scenarios
The bottom sections surface recent trend deviations and conditional scenarios the model flagged as decision-relevant.
Responsible use