AI Trading Agent Performance Review

Paper-first boundary

Trading Boy does not execute live trades, hold funds, or provide financial advice. This page is for simulated paper-trading review, prompt design, journal structure, and human review. It is not a signal feed, broker instruction, or promise of live trading results.

When to use this page

Use it after a sample window when the reviewer needs to understand whether the agent process improved, not whether one trade won.

The search intent behind how to evaluate an AI trading agent performance review template is usually practical: the user wants a reusable asset, an example, and a checklist. This page treats those variants as one owner page rather than splitting guide, example, checklist, and template into thin sibling pages.

Use the asset only after the paper-trading workflow, prompt version, permission boundary, and journal output format are clear. If the user cannot identify the setup, invalidation, risk field, or review question, the next action should be a skip, caveat, or human review rather than a confident simulated entry.

Performance review worksheet

Review field	Pass condition	Paper-mode boundary
Sample size	Counts reviewed paper trades, skips, exits, and exclusions.	Tiny samples cannot justify broad conclusions.
Process quality	Separates rule fit, output completeness, and review readiness.	Paper PnL is not the only score.
Risk behavior	Reviews drawdown, sizing, exposure, and pause compliance.	Risk exceptions are named.
Prompt changes	Lists every prompt or rule version in the window.	Untracked changes caveat the comparison.
Next action	Chooses collect, tighten, pause, or revert.	The review avoids large uncontrolled changes.

Reusable prompt or worksheet text

Role: You are helping organize a simulated paper-trading review. You may summarize context, apply written rules, identify missing fields, and prepare journal evidence. You may not route live trades, request secrets, or provide financial advice.

Required output: Decision type, setup name, rule version, thesis, invalidation, paper risk, data caveat, behavior tag, and one human review question.

Skip rule: If setup, invalidation, paper risk, or data quality is missing, produce a skip or excluded row. Do not invent the missing field to make the record look complete.

Review handoff: Send the output to the paper journal, evaluation checklist, or human review checklist. Choose one next paper-mode action: collect more samples, tighten one field, pause, or revert a prompt change.

Example paper workflow

Scenario: A two-week sample has positive paper PnL but only eight decisions and one prompt rewrite. The performance review says collect more samples under the same version before changing rules.

Good output: The review explains what evidence is strong, what is caveated, and what action is allowed next.

Weak output: The review says the agent is good because recent paper results were green.

Decision: The reviewer keeps the evidence in paper mode, checks the output with the AI paper trading agent evaluation checklist, and records any prompt change in AI trading agent prompt versioning.

Use it with these controls

Prompt version: every sample should name the version that produced the output.
Output format: the agent should use consistent fields for entries, exits, skips, and missed setups.
Risk gate: size, exposure, drawdown, stop distance, and pause rules should be checked before judging the idea.
Human review: a person should approve prompt, risk, or workflow changes before the next sample.
Data privacy: private exchange credentials, account identifiers, and secret values should stay out of prompts and screenshots.

How it supports ranking

This owner page consolidates the guide, example, template, checklist, and for-paper-trading variants into one durable page. That gives searchers a complete answer without flooding the sitemap with near duplicates.

It also links into the surrounding Trading Boy system: paper-trading hub, AI paper agent, prompt template, output format, risk controls, permission boundaries, and paper-trading limitations.

Related AI paper-agent pages

Use these links to move between setup, output, risk, journal, and review pages without leaving the paper-first cluster.

AI Trading Agent Performance Review FAQ

What should an AI trading agent performance review include?

It should include sample size, rule fit, risk behavior, output quality, prompt versions, and one next paper action.

How often should performance be reviewed?

Review on a fixed cadence such as weekly or after a defined sample count, not after one emotional result.

Does performance review predict live returns?

No. It reviews simulated paper evidence and does not prove future performance.