alpha_tools/backup_code/wqb-server2/untracked/skills/brain-how-to-pass-AlphaTest/SKILL.md

name	description
brain-how-to-pass-AlphaTest	Provides detailed requirements, thresholds, and improvement tips for WorldQuant BRAIN Alpha submission tests. Covers Fitness, Sharpe, Turnover, Weight, Sub-universe, and Self-Correlation tests. Use this when the user asks about alpha submission failures, how to improve alpha metrics, or test requirements.

BRAIN Alpha Submission Tests: Requirements and Improvement Tips

This skill provides key requirements and expert tips for passing alpha submission tests. For comprehensive details, thresholds, and community-sourced strategies, please read reference.md.

Overview

Alphas must pass a series of pre-submission checks to ensure they meet quality thresholds.

1. Fitness

Requirements

• At least "Average": Greater than 1.3 for Delay-0 or Greater than 1 for Delay-1.
• Fitness = Sharpe * sqrt(abs(Returns) / max(Turnover, 0.125)).

Tips to Improve

• Increase Sharpe/Returns and reduce Turnover.
• Use group operators (e.g., with pv13) to boost fitness.
• Check with check_submission tool.

2. Sharpe Ratio

Requirements

• Greater than 2 for Delay-0 or Greater than 1.25 for Delay-1.
• Sharpe = sqrt(252) * IR, where IR = mean(PnL) / stdev(PnL).

Tips to Improve

• Focus on consistent PnL with low volatility.
• Decay signals separately for liquid/non-liquid stocks.
• If Sharpe is negative (e.g., -1 to -2), try flipping the sign: -original_expression.

3. Turnover

Requirements

• 1% < Turnover < 70%.

Tips to Improve

• Use decay functions (ts_decay_linear) to smooth signals.

4. Weight Test

Requirements

• Max weight in any stock <10%.

Tips to Improve

• Use neutralization (e.g., neutralize(x, "MARKET")) to distribute weights.

5. Sub-universe Test

Requirements

• Sub-universe Sharpe >= 0.75 * sqrt(subuniverse_size / alpha_universe_size) * alpha_sharpe.

Tips to Improve

• Avoid size-related multipliers.
• Decay liquid/non-liquid parts separately.

6. Self-Correlation

Requirements

• <0.7 PnL correlation with own submitted alphas.

Tips to Improve

• Submit diverse ideas.
• Use check_correlation tool.
• Transform negatively correlated alphas.

General Guidance

• Start Simple: Use basic operators like ts_rank first.
• Optimize Settings: Choose universes like TOP3000 (USA, D1).
• ATOM Principle: Avoid mixing datasets to benefit from relaxed "ATOM" submission criteria (Last 2Y Sharpe).