LNG Trading Model
1st place trading model for LNG commodity markets
Python · ARIMA-GARCH · Monte Carlo · Quantitative Finance
Overview
Won 1st place at the Baringa Trading Competition (NTU x CEIT) with an LNG trading model that actually worked. The challenge was to build a strategy for trading liquefied natural gas — spot markets, forward curves, the whole thing.
What We Built
The model runs on two layers: forecasting and risk management.
For price prediction, we went with ARIMA-GARCH. ARIMA handles the directional trends; GARCH captures when volatility clusters (LNG prices get choppy around supply disruptions and seasonal swings). Not fancy, but it's robust and you can actually interpret what it's doing.
The risk layer uses Monte Carlo simulation to stress-test positions. We generate thousands of price paths, then calculate VaR and CVaR to size positions sensibly. No point having a sharp forecast if you blow up on tail risk.
For hedging, we designed a strategy using NYMEX natural gas futures. LNG doesn't have perfect liquidity, so you hedge with correlated instruments and manage the basis risk. We optimized hedge ratios based on rolling correlations between our target exposure and the futures.
Outcome
1st Place — Baringa Trading Competition 2025
The model beat 20+ teams. The judges liked that we had a clear view on when the model would work and when it wouldn't — we weren't overselling.
Gallery
