Backtest Cost Sensitivity

io.github.tylerscomic-labv1.0.0更新於 Oct 2, 2026

Commission/slippage sensitivity sweep, breakeven cost finder, and lookahead-bias checklist.

已驗證Streamable HTTP可網頁執行FinanceData & Analytics

概覽

AI 產生的概覽

透過掃描手續費與滑點假設來分析回測結果,找出策略的損益兩平成本並稽核前視偏差。

功能
它提供三個用來評估執行成本如何影響交易策略回測的工具。cost_sensitivity_sweep 針對回測結果掃描一系列手續費與滑點假設,並回報各項績效指標如何隨成本變化而下降。breakeven_cost_finder 找出策略優勢歸零的確切成本水準。lookahead_bias_checklist 針對跨 K 線邊界的資料洩漏、重繪指標和讀取未來 K 線等前視偏差模式提供結構化稽核。
適用情境
適用於在實盤前驗證演算法交易策略,尤其是當回測只依賴單一固定手續費或滑點假設時。它有助於判斷策略優勢能否在真實執行成本下存活,以及回測是否包含前視偏差。
執行需求
以遠端 streamable HTTP 端點方式執行;未宣告任何套件、環境變數、標頭或身分驗證。自架方式為透過 npm 安裝並執行 node server.js,需要 Node.js。託管方式據描述提供免費方案與付費 Pro 方案。
安裝前請注意
託管方式據描述提供付費 Pro 方案以取得更高額度,因此超出免費方案後可能產生費用。這些工具分析回測輸入並回傳分析結果;未描述寫入、傳送或刪除動作。未宣告任何憑證或密鑰。

安裝

在 SourceWeft 中

  1. 開啟 儀表板中的 Backtest Cost Sensitivity,將其新增到工作區。
  2. 為需要使用其工具的對話啟用該服務。

Web executable,透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。

其他 MCP 客戶端

把它新增到你客戶端的 mcpServers 設定中。

{
  "mcpServers": {
    "backtest-cost-sensitivity-mcp": {
      "type": "http",
      "url": "https://backtest-cost-sensitivity-mcp.mcpize.run/mcp"
    }
  }
}

README

backtest-cost-sensitivity-mcp

[License: MIT] [Live on MCPize]

An MCP server that sweeps commission and slippage assumptions across a backtest to find the real breakeven cost per trade, plus a structured look-ahead-bias audit.

The problem this solves

Most backtests run with one fixed commission/slippage assumption, which hides how fragile a strategy's edge actually is to real-world execution costs. A strategy that's profitable at 0.5 ticks of slippage and dead at 1.5 ticks needs to know that before going live, not after.

Tools

cost_sensitivity_sweep

Sweeps a range of commission and slippage assumptions against backtest results and reports how performance metrics degrade across the range.

breakeven_cost_finder

Finds the exact commission/slippage level at which a strategy's edge goes to zero — the real cost ceiling, not a guess.

lookahead_bias_checklist

A structured audit for the specific look-ahead-bias patterns that inflate backtest performance (data leakage across bar boundaries, repainting indicators, future-bar reads).

Use it

Hosted (recommended): MCPize — free tier, paid Pro tier for higher limits.

Self-host:

bash
npm installnode server.js

Part of the AlgoForge suite

Prop-firm and quant-validation tools for algo traders: prop-rules-mcp, trade-journal-mcp, payout-calc-mcp, econ-calendar-mcp, montecarlo-validator-mcp, overfitting-audit-mcp, walkforward-validator-mcp, pinescript-audit-mcp, pinescript-mcp.

License

MIT

來源:README.md,提交 e25023c

工具

0
工具後設資料尚未被收錄。

版本歷史

1
  1. v1.0.0最新Oct 2, 2026