marketcalls/vectorbt-backtesting-skills
6 skills7.6k installs900 starsGitHub
Install
npx skills add https://github.com/marketcalls/vectorbt-backtesting-skillsSkills in this repo
1Backtestbacktest is a VectorBT-focused agent skill that writes a runnable Python backtest for a user-specified strategy and market symbol. Indie quant builders and solo traders use it when they want ema-crossover, RSI, Donchian, or similar templates turned into a single file with data load, signal logic, performance stats, and plots—not a one-liner in chat. Arguments default sensibly (for example SBIN on NSE daily) but the agent should ask when strategy is missing. It integrates with OpenAlgo for live history fetches or accepts a DuckDB path with Historify versus custom table detection per vectorbt-expert rules. Shell and file write permissions matter because the skill creates directories and scripts. Pair it with the broader vectorbt-expert skill for pattern consistency. The outcome is an executable artifact you can iterate in Ship and Operate phases as you add walk-forward tests or paper trading—without re-deriving boilerplate each time.1.8kinstalls2Vectorbt Expertvectorbt-expert packages procedural knowledge for running VectorBT backtests wired to OpenAlgo for Indian market data. Solo quant-curious builders can reproduce portfolio simulations that load dotenv credentials, pull daily bars for symbols like NIFTYBEES and GOLDBEES, apply allocation weights, and charge realistic delivery commissions plus fixed per-order fees. The skill threads through data fetch, signal or static entry, portfolio construction, benchmark comparison against index and fixed-income proxies, and Plotly visualization. It suits validating personal or indie research strategies—not regulated investment advice—before automating execution or embedding analytics in a SaaS. Expect intermediate Python, a running OpenAlgo instance, and comfort interpreting drawdown and CAGR outputs rather than one-click brokerage deployment.1.5kinstalls3OptimizeOptimize is an agent skill for indie traders and builders using the VectorBT backtesting stack who need repeatable parameter sweeps instead of one-off notebooks. You pass strategy name, symbol, exchange, and interval (for example ema-crossover, SBIN, NSE, D); the skill reads vectorbt-expert patterns, creates the backtesting directory on demand, and writes a Python script that loads `.env` from the project root, pulls history through OpenAlgo or a DuckDB file, computes indicators with TA-Lib and OpenAlgo specialty helpers, and tests parameter grids with heatmap-friendly output. It encodes discipline against VectorBT built-in indicators and mandates signal cleaning with exrem. Use it when you are validating whether a rule-based strategy tolerates parameter changes before live or paper deployment. Complexity is intermediate because you need Python, VectorBT, and either OpenAlgo credentials or a prepared DuckDB dataset.1.2kinstalls4Strategy CompareStrategy-compare is a generator skill for quant-curious solo builders using the vectorbt-backtesting stack. Given a symbol and optional strategy names, it produces a comparison script under backtesting/strategy_comparison/ that loads data once—via OpenAlgo or an optional DuckDB path—runs each strategy with consistent indicator rules, and emits a side-by-side statistics table. It enforces TA-Lib and OpenAlgo ta usage, cleans signals with exrem, and supports a long-vs-short lens when you need directionality contrast on the same rule set. Allowed agent tools include Bash and file writes, so review scripts before running against live keys. Use it in Validate when you are prototyping which rule set deserves deeper optimization, not when you need live execution or compliance-ready reporting.1.1kinstalls5Quick StatsQuick-stats is a VectorBT backtesting skill for solo quant builders who want immediate feedback on a symbol without scaffolding a script. Given optional symbol, exchange, and interval arguments, the agent emits a single code block to run in a notebook cell or terminal: pull OHLCV from OpenAlgo when available, else DuckDB if the user supplies a path, else yfinance; compute EMA 10 and EMA 20 with TA-Lib; clean overlapping signals with exrem; run a delivery-style fee backtest; and compare against NIFTY from OpenAlgo. The printed summary covers total return, Sharpe, Sortino, max drawdown, win rate, profit factor, and trade count over a stated period. It is intentionally opinionated for Indian equity delivery economics and NSE defaults (e.g. SBIN on NSE daily). Use when you need a fast sanity check on a crossover template before deeper strategy work in the vectorbt-backtesting-skills family.1.1kinstalls6Setupsetup is an agent skill from the VectorBT + OpenAlgo backtesting kit that runs a full local Python environment bootstrap for solo traders and indie quant builders. It detects the host OS, creates and activates a venv, installs system TA-Lib where required, pulls openalgo, vectorbt, plotly, and peers, and establishes the backtesting directory structure so later skills can load data and run simulations. Invoke it at the start of a trading or analytics repo when nothing is installed yet or when teammates need identical machines. The skill uses Bash, Read, Write, and Glob with explicit macOS Homebrew and Linux apt/wget paths, reducing the classic TA-Lib pip failure. For Prism’s finance journey it sits on the Build backend shelf as the gate before strategy coding, keeping procedural steps citable for long-tail “vectorbt environment setup” search intent.1kinstalls