longbridge-strategy-optimizer

Response language: match the user's input language — Simplified Chinese / Traditional Chinese / English.

When to use

• Building or optimising a quant strategy — "帮我优化这个均线策略", "strategy optimization", "参数网格搜索"
• Walk-forward or overfitting checks — "走前验证", "过拟合检测", "walk-forward validation"
• Strategy combination — "多策略组合", "strategy combination", "correlation diversification"
• Sharpe / Calmar ratio targets — "最大化夏普比率", "Calmar ratio optimisation"

Note: computation-intensive calculations should be executed locally. This skill generates the Python code framework; the user runs it on their own machine.

Workflow

1. Clarify the strategy type (momentum / mean-reversion / breakout / factor) and target universe.
2. Fetch historical daily OHLCV data for the symbol(s).
3. Define the parameter search space.
4. Generate a Python framework that:
   • Implements the strategy logic
   • Runs

     GridSearchCV

     -style parameter sweep
   • Splits data into in-sample (IS) and out-of-sample (OOS) windows
   • Applies walk-forward validation (rolling IS/OOS windows)
   • Computes Sharpe, Calmar, max drawdown, and IS/OOS degradation ratio
   • Flags overfitting if OOS Sharpe < 0.5 × IS Sharpe
5. For multi-strategy combination: compute correlation matrix and suggest weights.
6. Present results as a parameter heatmap description and key metrics table.

If unsure of exact flag names, run

longbridge <subcommand> --help

before proceeding.

CLI

# Daily OHLCV — backtest data (up to 500 trading days)
longbridge kline <SYMBOL> --period day --count 500 --format json

Output structure

STRATEGY OPTIMISATION REPORT — <SYMBOL>  <Date>

STRATEGY: <Name / Type>
Universe:  <SYMBOL>
Period:    <start> – <end>  (xxx days)

PARAMETER SEARCH
Parameter       Range         Step    Best Value
fast_ma         5–50          5       xx
slow_ma         20–200        10      xxx
stop_loss       0.5%–5%       0.5%    x.x%

BEST RESULT (IN-SAMPLE)
Sharpe:  x.xx   Calmar:  x.xx   Max DD:  -xx.x%
CAGR:    xx.x%  Win Rate: xx.x%  Trades: xxx

OUT-OF-SAMPLE VALIDATION
Sharpe:  x.xx   Calmar:  x.xx   Max DD:  -xx.x%
IS/OOS Degradation: xx%  → [Acceptable | Possible Overfit | Overfit]

WALK-FORWARD SUMMARY
Window 1: IS Sharpe x.xx → OOS Sharpe x.xx
Window 2: IS Sharpe x.xx → OOS Sharpe x.xx
...

PYTHON CODE FRAMEWORK
<generated Python code>

Error handling

command not found: longbridge

not logged in

longbridge auth login

longbridge auth login

longbridge auth login

MCP fallback

longbridge

mcp__longbridge__kline

Related skills

• longbridge-kline

  — candlestick and OHLCV history

• longbridge-execution-model

  — realistic execution cost assumptions

• longbridge-risk-return

  — portfolio efficiency and risk-adjusted return

• longbridge-capital-flow

  — intraday volume and order flow signals

File layout

skills/longbridge-strategy-optimizer/
└── SKILL.md