Key insights
- An analyst presents a bullish case for corn using a gradient boosted trees model, identifying urea prices as a key predictor. The analysis uses a 30-year dataset and considers various macro and commodity variables. A simulator models the performance of a leveraged ETP based on a moderate corn price escalation. While interesting, the direct impact on broad US equities is limited, hence a low influence score.
Watching all the chaos unfold and the difficulty in finding value investment, I carried out some analyses to support a bull case for corn by the third quarter this year. I'd welcome everyone's critical feedback and discussion.
***Disclosure I conducted this analysis iteratively on Claude, from which the simulator is published as an Artefact***
I began by first constructing a 30-year monthly dataset (March 1996 to March 2026, 361 observations) across six commodity prices (corn, urea, Brent oil, natural gas, live cattle, gold) and three macro variables (US CPI inflation, Fed funds rate, S&P 500 level and monthly growth). Each series was built from approximately 50–80 known anchor points sourced from my training data, with linear interpolation between them to fill monthly gaps.
Applying a Bayesian MCMC linear regression model failed to reliably predict trends from 2020-2025, so I moved to Non-linear Gradient Boosted Trees. Between the chosen variables, there were 69 features between the raw series, lagged values (3, 6, 12 months), momentum features, rolling statistics, cross asset ratios and macro derivative features. The best model (500 trees, depth 4, learning rate 0.03, 80% subsample, min 10 samples per leaf) was then fitted seven times at different quantiles (5th, 10th, 25th, 50th, 75th, 90th, 95th) to produce a full predictive distribution, not just a point estimate. In the end, urea (level and momentum) was the dominant predictor of future corn price.
Final step was feeding feature vectors through fitted models (7 quantile GBTs + Random Forest + Extra Trees) to produce point predictions and uncertainty bands at each time step. Taking a leverage ETP as the trading instrument, the interactive simulator being shared takes a conservative moderate escalation corn price path and models the actual mechanics of a 2× daily-reset leveraged ETP held over 12 months (3,000 Monte Carlo paths are run per parameter set, and percentile bands (5th through 95th) are computed at each month to produce the fan chart.)
Qualitatively, the literature out right now is mostly bearish on U.S. corn fundamentals in isolation, but the contrarian bull case is well described and centres on three converging forces: tight global ex-China/ex-U.S. stocks, fertilizer driven acreage reduction, and the Hormuz supply shock transmitting through nitrogen prices into planted area.
Near term events that could validate the thesis is the USDA Prospective Planting report and June Acreage.
https://claude.ai/public/artifacts/fe5e4155-6f4f-4c82-aa05-49e98da64857