zerg estimate

/zerg:estimate

Full-lifecycle effort estimation with PERT confidence intervals, post-execution comparison, and historical calibration.

Synopsis

/zerg:estimate [<feature>] [--pre] [--post] [--calibrate]
                           [--workers N]
                           [--format text|json|md]
                           [--verbose] [--history] [--no-calibration]

Description

The estimate command provides effort estimation at every stage of the ZERG workflow. It operates in three modes depending on the current state of the feature or the flags provided.

Pre-Execution Mode

Before /zerg:rush has been run, estimate analyzes the task graph and produces time and cost projections:

1. Task complexity scoring — Reads task-graph.json and scores each task by file count, dependency depth, and description complexity.
2. PERT estimation — Calculates optimistic, most likely, and pessimistic durations per task using weighted averages.
3. Wall-clock projection — Given --workers N, simulates parallel execution across levels to estimate total wall-clock time.
4. API cost projection — Estimates token usage and Anthropic API cost based on task complexity and historical averages.

Post-Execution Mode

After tasks have completed, estimate --post compares actual execution data against the original estimates:

1. Reads worker logs and task completion timestamps.
2. Computes actual duration, token usage, and cost per task.
3. Generates a comparison table showing estimated vs actual values.
4. Calculates accuracy percentage and bias direction (over/under-estimate).

Calibration Mode

estimate --calibrate analyzes historical estimation accuracy across all features:

1. Reads all estimate files from .gsd/estimates/.
2. Computes per-task-type bias factors (e.g., "test tasks are consistently underestimated by 30%").
3. Stores calibration data for automatic bias correction in future estimates.

Options

Option	Default	Description
<feature>	auto-detect	Feature to estimate. Defaults to .gsd/.current-feature.
--pre	auto	Force pre-execution estimation mode.
--post	auto	Force post-execution comparison mode.
--calibrate	off	Show historical accuracy and compute bias factors.
--workers N	config default	Worker count for wall-clock projection.
--format	text	Output format: text, json, or md.
--verbose	off	Show per-task breakdown instead of level summaries.
--history	off	Show past estimates for this feature.
--no-calibration	off	Skip applying calibration bias to estimates.

Examples

Estimate effort before launching workers:

/zerg:estimate user-auth

Compare actual vs estimated after execution:

/zerg:estimate user-auth --post

View per-task breakdown:

/zerg:estimate --verbose

Calibrate based on all historical data:

/zerg:estimate --calibrate

Output as JSON for tooling:

/zerg:estimate --format json

Error Handling

• If no task graph exists, the command reports an error and suggests running /zerg:design first.
• If --post is used before any tasks have completed, the command reports that no execution data is available.
• If --calibrate finds no historical data, it reports that at least one completed feature is required.

Task Tracking

This command creates a Claude Code Task with the subject prefix [Estimate] on invocation, updates it to in_progress immediately, and marks it completed on success.

See Also

• zerg-design -- Generates the task graph that estimate analyzes
• zerg-rush -- Execute tasks that estimate projects
• zerg-status -- Real-time progress during execution