I think I’ve stumbled on a DIY way to “nowcast” quarterly margins for NSE: BALAMINES (Balaji Amines) using only free public data, and my rough backtests look shockingly decent. Before I get carried away, can you poke holes in this and suggest improvements?
What I’m doing:
- Map core feedstocks to finished products: methylamines ≈ methanol + ammonia; ethylamines ≈ ethanol + ammonia; add power/freight proxies.
- Pull weekly/daily spot prices for those inputs from public commodity portals, convert with USD/INR, and build a rolling “input basket” cost.
- Layer on shipping/freight proxies and energy costs (grid power and natural gas proxies).
- Check India import/export price trends for relevant amine categories (average unit value as a crude proxy for realizations).
- Adjust for product mix by tracking management commentary/capacity changes and industry news on downstream pharma/agro demand.
- Nowcast gross margin each month, aggregate to the quarter, and compare to reported numbers after results.
Early observations:
- Margin inflections seem to line up when methanol/ammonia spreads move sharply for 6-10 weeks.
- Export unit values and INR moves appear to matter more than I expected.
- A simple 3-factor model (feedstock basket, INR, freight) already explains a surprising chunk of margin variance in my small sample.
What I’m unsure about and need help on:
- Data quality: Any better free sources for consistent methanol/ammonia/ethanol spot series and India trade unit values? Pitfalls in relying on average unit values?
- Product mix drift: How do you sensibly proxy the evolving share of methyl vs ethyl amines and derivatives without overfitting?
- Regulatory shocks: Anti-dumping duties/sunset reviews can swing realizations overnight. Where do you track these early, and how do you reflect them in a model?
- Capacity/maintenance: Any public breadcrumbs to detect outages or ramp-ups (environmental clearances, power consumption proxies, port data) before results?
- Risk management: If the signal is “good but not great,” is this better used to avoid bad quarters rather than to time entries? How would you size such a cyclical exposure?
- Expression: Would a pairs setup (e.g., BALAMINES vs a close peer) reduce idiosyncratic risk if both share similar feedstocks but different mix? Any pitfalls you’ve seen?
- Liquidity/surveillance: Experience with ASM/GSM moves or sudden lot-size changes impacting execution in small/mid-cap chemicals right when the signal flashes?
- Taxes and practicality: For a retail investor, what trade frequency is the sweet spot so this doesn’t devolve into high-churn, tax-inefficient tinkering?
If there’s interest, I can share a cleaned-up template that:
- Ingests weekly input prices and USD/INR
- Computes a rolling feedstock basket
- Overlays a simple realization proxy
- Spits out a quarterly margin “range” with confidence bands
What would you add or change to make this robust enough for real-money decisions?
