skillfed

garch-volatility-toolkit

Build volatility forecasts using GARCH(1,1) and GJR(1,1) models with walk-forward fitting, then apply them to scale positions inversely to risk. The toolkit includes tested recipes for univariate fits, DCC correlation overlays, and position-sizing logic on BTC, ETH, SOL, and AVAX futures data. Results are cost-dependent: GJR vol-scaling works at institutional rates (2bps) but erodes at retail spreads (7bps); DCC de-weighting adds minimal economic value.

garch-volatility-toolkit forecasts asset volatility using GARCH and GJR models to dynamically scale portfolio positions.

AI-generated summary based on this skill's SKILL.md

58 8 MIT updated by terrylica

Install

terrylica/cc-skills/garch-vol-recipes · repository language: Shell

CLI (skillfed)coming soon
git clone https://github.com/terrylica/cc-skills
cp -r cc-skills/plugins/garch-volatility-toolkit/skills/garch-vol-recipes ~/.claude/skills/garch-vol-recipes

Frequently asked questions

AI-generated answers based on this skill's SKILL.md and metadata

What is garch-volatility-toolkit and what does it do?

garch-volatility-toolkit builds volatility forecasts using GARCH(1,1) and GJR(1,1) models with walk-forward fitting, then applies them to scale positions inversely to risk. The toolkit includes tested recipes for univariate fits, DCC correlation overlays, and position-sizing logic on BTC, ETH, SOL, and AVAX futures data. Results are cost-dependent: GJR vol-scaling works at institutional rates (2bps) but erodes at retail spreads (7bps); DCC de-weighting adds minimal economic value.

How do I use GARCH for trading with this toolkit?

garch-volatility-toolkit enables you to forecast asset volatility using GARCH(1,1) and GJR(1,1) models for dynamic position sizing. Fit models in walk-forward windows without lookahead bias, then scale position sizes inversely to predicted volatility. The toolkit provides cost-aware portfolio overlays with volatility targeting and risk control, letting you build overlays that account for trading costs at your execution venue.

What is the leverage effect and how does GJR model it differently than GARCH?

garch-volatility-toolkit lets you compare GARCH vs GJR to understand leverage effect in volatility modeling. GJR(1,1) captures asymmetry: negative shocks drive volatility clustering more than positive ones. Standard GARCH(1,1) treats all shocks equally. For crypto, where volatility spikes on downside moves, GJR often fits better. The toolkit shows which model suits your asset and regime.

What walk-forward volatility prediction methodology does the toolkit teach?

garch-volatility-toolkit teaches no-lookahead walk-forward backtesting methodology and leakage traps. Fit your GARCH or GJR model on historical data, forecast one step ahead, then roll forward by one period. Never fit on future data or use today's return to predict today's volatility. The toolkit demonstrates proper implementation on BTC, ETH, SOL, and AVAX to avoid overfitting and false performance claims.

How does inverse vol scaling affect position sizing and turnover costs?

garch-volatility-toolkit implements inverse vol scaling for position sizing: when volatility forecasts rise, reduce position size; when they fall, increase it. This keeps portfolio risk stable. However, turnover costs matter: GJR vol-scaling works at institutional rates (2bps) but erodes at retail spreads (7bps). The toolkit is cost-aware, showing you the breakeven spread for profitability at your venue.

What role does DCC correlation de-weighting play in portfolio construction?

garch-volatility-toolkit implements DCC correlation de-weighting for multi-asset portfolio construction. When correlations rise (stress regimes), the overlay reduces weights on correlated assets to control tail risk. Testing shows DCC de-weighting adds minimal economic value in most regimes, so the toolkit helps you decide whether the complexity justifies the cost for your strategy.

SKILL.md

rendered from the published skill — quoted content, verbatim

GARCH Volatility Toolkit

> Self-Evolving skill — if a recipe drifts from what actually reproduces (model spec, cost regime, campaign Sharpe numbers, library API), fix this SKILL.md in the same change; see the Post-Execution Reflection at the bottom.

Walk-forward GARCH(1,1) and GJR(1,1) recipes for volatility forecasting and portfolio construction. Tested on 20 seeds, 2025-26 test window, 2bps and 7bps cost regimes.

> Campaign Results (honest): A negative-to-marginal result set, not a deployable edge. GJR > inverse vol-sizing helps at LOW cost only (ensemble Sharpe +0.49→+1.00 @2bps) but is COST-FRAGILE — > its +35% turnover erases the benefit at retail 7bps (Δ+0.02, p=0.54, coin-flip). DCC de-weighting is > economically immaterial (+0.05–0.07). GARCH-as-features is a flat

(truncated - see the full file via the links below)

Read as markdown · JSON record · Browse the source repository

File tree — 4 files
plugins/garch-volatility-toolkit/skills/garch-vol-recipes/SKILL.md
plugins/garch-volatility-toolkit/skills/garch-vol-recipes/references/CAMPAIGN_VERDICT.md
plugins/garch-volatility-toolkit/skills/garch-vol-recipes/references/garch_angleB_verified.json
plugins/garch-volatility-toolkit/skills/garch-vol-recipes/references/garch_angleC_verified.json

Related skills

Tags

time-series-modeling risk-targeting backtesting-framework cost-aware-trading ensemble-methods leverage-asymmetry correlation-dynamics turnover-optimization crypto-futures walk-forward-validation