> ## Documentation Index
> Fetch the complete documentation index at: https://partners.centaur.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Trading Metrics

> How Centaur computes time-based performance, win rate, bias, and related trader metrics.

Centaur separates the sample you are evaluating from the window used to measure performance.

## Trade Sample Window

The Trade Sample Window controls which position opens are included in the current sample.

Common product options include:

| Sample     | Includes                                         |
| ---------- | ------------------------------------------------ |
| **30d**    | Positions opened in the last 30 days.            |
| **all**    | Accessible full history.                         |
| **custom** | Positions opened inside a selected custom range. |

API stats reads use explicit `startTime` and `endTime` bounds for the same idea.

## Time-Based Performance Window

The Time-Based Performance Window controls when return is measured after entry.

| Window  | Measures                                |
| ------- | --------------------------------------- |
| **1D**  | Return one day after position open.     |
| **7D**  | Return seven days after position open.  |
| **30D** | Return thirty days after position open. |

Example: sample = `30d`, performance window = `7D` means "positions opened in the sample, measured seven days after each open."

## Core metrics

| Metric             | Meaning                                                                              |
| ------------------ | ------------------------------------------------------------------------------------ |
| **Win Rate**       | Share of evaluated positions with positive return at the selected time-based window. |
| **Average Return** | Mean time-based return across evaluated positions.                                   |
| **Median Return**  | Median time-based return across evaluated positions.                                 |
| **Sharpe Ratio**   | Risk-adjusted return signal when enough evaluated returns are available.             |
| **Total Trades**   | Count of positions in the selected sample.                                           |
| **Asset Focus**    | Assets most represented in the selected sample.                                      |
| **Direction Bias** | Long/short/neutral balance in the selected context.                                  |
| **Holding Period** | Typical time between position open and close where lifecycle data is available.      |

Metrics can be null when there are no accessible rows, not enough evaluated positions, or no usable fixed-window return for the requested window.

## Open-position return

Open-position return is different from time-based performance. It uses the current mark price when available:

**Long:** `(current price - entry price) / entry price`

**Short:** `(entry price - current price) / entry price`

Open-position reads are the right source for current exposure and open-position skew. They do not prove how the trader will close the position.

## What not to do

* Do not treat generated narrative summaries as evidence for exact performance.
* Do not compute win rate by manually scanning event rows.
* Do not mix current open-position return with `1D`, `7D`, or `30D` time-based performance.
* Do not assume a missing metric means zero performance; it usually means unavailable or inaccessible data.