How it works
Quant strategy research in minutes
FiClaw takes a strategy idea through eight steps, from a written specification and generated code to a historical backtest report. You can review the assumptions and decide what to test next.
Describe the strategy idea
Write the logic in plain language, e.g. "a monthly rotation strategy on 20-day momentum, holding the 20 strongest names in the CSI 300".
10sAI generates the strategy spec
FiClaw turns the idea into a structured spec covering factor definitions, selection rules, rebalance frequency and risk constraints.
15-30sFactor evaluation and screening
Evaluate factor validity (IC, IR) and drop weak factors so the strategy rests on effective signals.
20-40sAutomatic code generation
Generate full Python from the spec, covering data preprocessing, signal computation, portfolio construction and risk logic.
15-30sReal-data backtesting
Run against real A-share history via QuantAPI and output Sharpe, annualized return, max drawdown and win rate.
30-60sDiagnosis and iteration
When results miss the target, review likely causes such as overfitting, factor decay or concentration, then decide whether to run another round.
as neededParameter optimization
Extract parameters, build a search space for grid optimization, and run walk-forward checks for robustness.
1-2 minGenerate the research report
Output a full report with performance metrics, risk analysis, parameter sensitivity and improvement suggestions.
10s