# PropAverages

> Windowed values of the analyzed stat: the full in-scope window (`season`), the two trailing windows (`last_5`/`last_10`, most-recent-N games), and the home/away split (by whether the person's team was the contest's home or away side; a game whose side can't be determined is excluded from both splits, not from `season`).

Windowed values of the analyzed stat: the full in-scope window (`season`),
the two trailing windows (`last_5`/`last_10`, most-recent-N games), and
the home/away split (by whether the person's team was the contest's home
or away side; a game whose side can't be determined is excluded from both
splits, not from `season`). Each value is `0.0` over an empty window.

**Aggregation depends on the stat's kind.** A counting stat (`h`, `pts`,
`total_bases`, `ip`) is the **mean of per-game values**, over the games
that carried the stat — `game_log_detail` is chronological, so the
trailing windows are its tail. A derived **rate** stat (`era`, `whip`,
`avg`, `obp`, `ops`, `save_pct`, …) is the **formula over the window's
summed bases** (season ERA = `27·Σer/Σouts`) — never a mean of per-game
rates — computed over every game the person played the stat's phase,
including games whose per-game rate is undefined (a 0-out outing has no
game ERA, but its earned runs count).

**Kind:** object

| Field | Type | Required | Description |
|---|---|---|---|
| `away` | `number<double>` | yes |  |
| `counts` | `PropWindowCounts` | yes | The honest sample size behind each windowed value — how many games each window actually contained (a `last_10` over a 3-game log is a 3-game value). |
| `home` | `number<double>` | yes |  |
| `last_10` | `number<double>` | yes |  |
| `last_5` | `number<double>` | yes |  |
| `season` | `number<double>` | yes |  |

**Used by:** [PlayerPropCard](/docs/reference/data-model/analysis/PlayerPropCard.md)