
Monte Carlo Backtest Validator
io.github.tylerscomic-labv1.0.0更新於 Oct 2, 2026
Bootstrap Monte Carlo backtest validation and prop-firm challenge pass-probability simulation.
概覽
使用自助式蒙地卡羅方法對交易回測進行統計驗證,並模擬自營交易挑戰的通過機率。
- 功能
- 提供將自助重取樣與重排蒙地卡羅方法套用於回測交易結果的工具。它計算每筆交易期望值的 90% 信賴區間,並在區間包含零時提出標記;推導回撤路徑百分位數;根據勝率、平均獲利與平均虧損計算每筆交易期望值;並根據勝率、風險報酬比、目標與回撤限制輸入模擬自營交易挑戰的通過機率。比較工具可依模擬通過率對多種勝率/風險報酬比組合進行排序。
- 適用情境
- 適合演算法交易者與量化開發者用來判斷回測優勢是否可能真實存在,而不是單一條幸運的淨值曲線;也適合準備自營交易挑戰、希望比較不同風險結構通過機率的人。
- 執行需求
- 以採用 streamable HTTP 的遠端託管 MCP 端點執行;未宣告驗證、環境變數或標頭。README 也說明了透過 npm install 與 node server.js 自行託管的選項,這需要 Node.js。託管層免費,付費 Pro 層提供更高額度。
安裝
在 SourceWeft 中
- 開啟 儀表板中的 Monte Carlo Backtest Validator,將其新增到工作區。
- 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": {
"montecarlo-validator-mcp": {
"type": "http",
"url": "https://montecarlo-validator-mcp.mcpize.run/mcp"
}
}
}README
montecarlo-validator-mcp
[License: MIT] [Live on MCPize]
An MCP server that statistically validates whether a backtest's edge is real, using bootstrap-resampling and reshuffling Monte Carlo methodology, plus prop-firm-specific pass-probability simulation.
The problem this solves
A single backtest equity curve tells you what happened on one path through history — it doesn't tell you how likely that result was to happen by chance, or what the range of plausible outcomes looks like on the next set of trades. This wraps the actual statistical validation (bootstrap confidence intervals, drawdown-path percentiles, challenge pass-probability simulation) instead of eyeballing one curve.
Tools
monte_carlo_validate
Bootstrap 90% confidence interval on per-trade expected value (flags when the interval includes zero), plus drawdown-path percentiles via reshuffling.
expected_value_calculator
Per-trade EV from win rate, average win, and average loss.
prop_firm_pass_probability
Simulates challenge pass probability from win-rate/risk-reward/target/drawdown-limit inputs.
risk_geometry_comparator
Ranks multiple win-rate/risk-reward geometries by simulated pass rate — surfaces that tight, high-win-rate setups often out-pass high-RR/low-win-rate setups on a fixed-target challenge, independent of raw expected value.
Use it
Hosted (recommended): MCPize — free tier, paid Pro tier for higher limits.
Self-host:
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, overfitting-audit-mcp, walkforward-validator-mcp, pinescript-audit-mcp, backtest-cost-sensitivity-mcp, pinescript-mcp.
License
MIT
來源:README.md,提交 3a0e98c
工具
0版本歷史
1- v1.0.0最新Oct 2, 2026