dacos

A Python library for Medium-Frequency Trading alpha research. Very early stage, experimental, and definitely not production-ready. Use at your own risk, and please be kind – we're learning here.

dacos (pronounced "dalam-kost") is a personal project born out of curiosity about statistical arbitrage, mean-reversion, and momentum strategies in crypto markets. It aims to provide a simple, modular, and safe toolkit for building and testing alpha signals without blowing up your laptop.

Why "dacos"?

It stands for Data Construction System, but honestly, and it’s available on PyPI—coming soon (as of this writing).

Features (so far)

• ETL Pipeline – Build "skinny" tables from raw Parquet files using Polars lazy evaluation. No RAM overload, I promise.
• Asynchronous Time Alignment – A pure-Polars metronome (upsample) that perfectly aligns different coin timestamps using forward-fill physics (No look-ahead bias).
• StatArb Engine & Physics Lab – Built-in OLS Rolling Z-Score calculation, Hurst Exponent, ADF Stationarity Test, and Ornstein-Uhlenbeck Half-life.
• Result Monad – Explicit error handling (Rust-style Ok/Err) so your code doesn't crash unexpectedly in the middle of a multi-million row calculation.
• Type Hints Everywhere – Because my ADHD brain forgets what a function returns after 5 minutes.

Installation

# From PyPI
pip install dacos

Status

This is a learning project. I'm not a quant, not a data engineer, just someone who enjoys messing around with data and trading ideas. The code works (for me), but it's full of imperfections, rough edges, and probably a few bugs.

If you're an experienced developer and spot something horrible, feel free to open an issue or PR – but please be gentle.

Roadmap [x] ETL pipeline (extract, filter, log transform) [x] Result monad for error handling [x] Data alignment (as-of join for multiple symbols) [x] Statistical tests (Hurst, ADF, half-life) [x] Pairs trading engine (Rolling Z-Score) [ ] Momentum signals [ ] Execution Tactics (Position Sizing & Entry/Exit Logic) [ ] Backtesting framework (maybe)

How to Use

dacos is designed to hide complex data engineering behind a simple API.

1. Extracting Data

from dacos.builder import create_skinny_builder

# Point to your raw Parquet folder and where you want the silver table
builder = create_skinny_builder("data/raw", "data/silver")
result = builder.execute_pipeline()

if result.is_ok():
    print("Success! Skinny table created.")
else:
    print(f"Oh no: {result.error()}")

2. Pairs Research

import dacos.api as dc

# 1. Run the entire pipeline (Ingestion -> Alignment -> Validation -> Math -> Z-Score)
res = dc.run_pairs_research(
    silver_path="data/silver",
    y_symbol="WIF_USDT",
    x_symbol="PEPE_USDT",
    frekuensi="1m",
    z_window=100
)

if res.is_ok():
    output = res.unwrap()
    df_zscore = output["data"]
    metrics = output["metrics"]

    print(f"Hurst Exponent: {metrics['hurst']}")
    print(f"Half-Life: {metrics['halflife']} minutes")
    print(df_zscore.tail(5)) # View the latest Z-Scores!

Point to your raw Parquet folder and where you want the silver table

builder = create_skinny_builder("data/raw", "data/silver") result = builder.execute_pipeline()

if result.is_ok(): print("Success! Skinny table created.") else: print(f"Oh no: {result.error()}")

Contributing

Contributions are welcome! But please keep in mind:

I'm still learning, so my code style might be weird.
Write tests if you add Features.
Be nice.

LICENSE

MIT – do whatever you want, but don't blame me if you lose money trading.