Model the market
I'm Matt Jennings. I have spent about thirty years building and running production cost models of the North American power system — ERCOT, WECC, and the Eastern Interconnection, in Aurora, EnCompass, and Plexos. Utility resource planning first, then merchant trading risk, then fifteen years leading data teams whose job was to hand a number to someone who was going to act on it.
This site is where I write that work down. Not forecasts, and not opinions about where prices are going, but the machinery underneath: what a model is actually solving, which assumptions carry the result, and how to tell a careful answer from a confident one.
It is a teaching site. There is nothing to buy here. More about me.
Supply stacks up in merit order, demand crosses it, and the price is set where they meet. Most disagreements about a price are disagreements about the shape of one of those two lines.
A production cost model does not solve the year once. It solves the same fleet against a different load every hour, and the annual number is what is left when those hours are added up.
Underneath is an optimisation: the cheapest schedule that respects every physical limit. Knowing what it minimises, and what it is told to ignore, is most of knowing what to make of its answer.
Two competent teams model the same market and land eleven dollars apart. The gap is almost never in the engine, and knowing where it actually lives changes what you should do about it.
Forecasting · Price formation
The primer. What the solver is minimising, what the constraints represent physically, and why the answer is a schedule rather than a price.
Modelling · Fundamentals
One essay at a time, sent when it is finished rather than on a schedule. No news roundups, no market commentary, no sales.
Subscribing happens on Substack, which has its own privacy policy and sets its own cookies. Nothing about it touches this site. Privacy.