跳到正文
TakoAPI
返回技能列表

backtest-expert-0-1-0

由 itsjustfred 提供

Coding Agents & IDEs
333下载量由 clawskills.sh 提供
0浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

Expert guidance for systematic backtesting of trading strategies.

描述

该技能的功能

Guides systematic backtesting of quantitative trading strategies using a methodology that prioritizes robustness over optimistic results. Covers parameter sensitivity testing, slippage modeling, bias prevention, and out-of-sample validation. Designed for rule-based strategies where every decision is codified in advance.

It enforces a stress-first framework that deliberately adds friction and punishes strategies, which filters out curve-fitted results that would fail in live trading.

何时使用

  • Testing whether a gap-up earnings strategy survives realistic slippage
  • Checking if stop loss parameters are fragile or stable across a range
  • Running walk-forward analysis on a mean-reversion system
  • Diagnosing why a backtest shows suspiciously high win rates
  • Evaluating year-by-year consistency before deploying a strategy live

示例工作流

输入

User asks: I have a gap-up earnings strategy with 87% win rate in backtesting — is it ready to trade?

智能体
  1. 1

    Ask user to state the edge hypothesis in one sentence and confirm all rules are fully codified with no discretion

  2. 2

    Check sample size — flag if fewer than 100 trades and warn about statistical confidence

  3. 3

    Walk through parameter sensitivity: vary stop loss and profit target at 50–150% of baseline values to identify plateaus vs narrow spikes

  4. 4

    Apply pessimistic slippage (1.5–2x estimates) and worst-case fill assumptions, then compare performance degradation

  5. 5

    Flag the 87% win rate as a red flag and audit for look-ahead bias or data alignment issues

输出

Assessment of whether the strategy survives stress tests, with specific failure patterns identified and a deploy/refine/abandon recommendation

安装

clawhub install itsjustfred/backtest-expert-0-1-0

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接