See the numbers.
Know the sample.

A real, defined sample

The preview uses Shai Gilgeous-Alexander’s retained regular-season records from 2024–25 and 2025–26, collected from API-NBA on September 12, 2026. The question stays the same across all three views: did he score at least 30 points?

The source records were checked for game identity, participation, shooting consistency and agreement between player points and final team scores. Games that failed those checks were excluded. These figures describe the retained sample, not every game in either full season. No absent record is filled with zero.

The baseline

The hit rate is the number of qualifying results divided by settled games. An over 29.5 points line means 30 or more. Pushes are excluded where a line allows them. The chart shows the latest retained games, while the headline uses the full selected sample.

One condition per factor

Use the historical example, choose a recorded game or set custom conditions. Each factor has one active condition. Selecting home and one day off shows two independent comparisons, rather than restricting both cards to games meeting both conditions.

Uplift uses all checked history in this preview: the retained 2024–25 and 2025–26 regular-season appearances. Base filters and research weights do not narrow that history. Choosing a recorded game gives a retrospective comparison across this sample, not a reconstruction of what was known before that game.

Each card compares a stat average in one condition with a separate historical baseline from the same seasons and game type. The baseline follows the selected group’s season mix, so a season with more selected games contributes proportionally.

A positive difference is uplift; a negative difference is downlift. It describes an association in these records. It is not evidence that the condition caused the difference, and it is not added to a hit-rate percentage. The cards are separate comparisons, not adjustments to sum together.

Read the counts behind the difference

Condition and comparison counts show the appearances used. Expand a card’s audit to inspect contributing seasons, their counts and excluded or missing records. Groups without usable games on both sides are excluded.

Small-sample status means fewer than 30 games on either side of any contributing season and game-type group. It describes sample support, not predictive confidence. A large total can still contain a small seasonal group.

Where the context comes from

“Designated home” follows the source’s home/away designation. Some NBA games use neutral venues. Days off use the team schedule and Eastern calendar dates. Scheduled tipoff gaps and home/away transitions use the previous team game recorded by the same source. They describe the schedule, not the player’s actual travel or recovery.

The opponent comparison uses this player’s scoring history against the selected team. It does not measure what that opponent concedes to players at his position.

Geographic travel direction and distance, opponent concessions by position, head referee and officiating panel remain unavailable when their source records are unverified. Missing context is never treated as a zero or a known condition.

Weights you can inspect

Each category contributes its settled-game count multiplied by your chosen weight. The weighted historical hit rate is the sum of weighted hits divided by the sum of game contributions.

One game can appear in several categories. The number of contributions can therefore exceed the number of unique games. A 0% weight removes a category’s contribution; it does not remove those games from other categories.