Breaking

Miles Deutscher Open-Sources GARCH Method Trading Framework for Claude Code and TradingView

September 16, 2026 at 22:46 EDT

AI Agents
Finance
Open Source

Miles Deutscher has published GARCH Method, an MIT-licensed trading framework that packages walk-forward GARCH(1,1) volatility forecasting into a Claude Code skill, Python scripts, and a TradingView Pine Script indicator. The release is notable because it turns a classic quant-risk technique into an agent-callable workflow focused on position sizing rather than price prediction.

July 8, 2026 | GARCH Method

Volatility sizing becomes agent-ready

An open MIT project wraps walk-forward GARCH(1,1), position sizing and a TradingView gauge into a Claude Code workflow. Its core promise is narrower than prediction: decide how much exposure to take when volatility changes.

Forecast

1 day

Ahead volatility estimate

Refit

21 days

Parameter refresh cadence

History

510

Daily prices minimum

Demo

9/21

EMA crossover strategy

Sizing cap, drawn to scale

The maximum multiplier is eight times the minimum allowed exposure.

0.25x

Minimum cap

2.0x

Maximum cap

Workflow

1. Price data

Ticker or CSV

2. GARCH forecast

Walk-forward

3. Regime and size

Calm to storm

4. Compare

Fixed vs targeted

Why traders noticed

No API keys, CSV support, a chart overlay and a plain-English agent entry point make a professional risk idea easier to test.

What it does not do

It does not call market direction or replace a strategy; users still need signals, validation and drawdown review.

The honest label

How much, not which way. The package is a risk-sizing layer: the strategy chooses entries, while the GARCH workflow scales exposure as forecast volatility rises or falls.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.