Day-ahead vs. real-time markets
The two-stage structure of wholesale power: a forward market that schedules tomorrow, and a spot market that trues up what actually happens.
Wholesale electricity trades in two connected markets. The day-ahead market runs once a day: participants submit offers to sell and bids to buy for each hour of tomorrow, and the market clears a schedule and a price for every hour before the day begins. It is a financial commitment — a plan, made while there is still time to start up power plants deliberately.
The real-time market is the true-up. Tomorrow never unfolds exactly as scheduled — weather shifts, plants trip offline, demand surprises — so the grid operator continuously re-dispatches the system and prices the deviations. If you sold 100 MWh day-ahead but only produced 90, you buy the missing 10 back at the real-time price. If you produced 110, you sell the extra 10 at it.
The two prices usually track each other but can diverge sharply when something unexpected happens, and that divergence is where much of the risk and opportunity in power trading lives. Selling day-ahead locks in certainty; waiting for real-time is a bet that conditions will be tighter than the market expected. Batteries and traders constantly weigh one against the other.
For the technical reader
The day-ahead market is a financially binding, security-constrained unit commitment and economic dispatch over 24 hourly intervals, co-optimizing energy and (in most designs) ancillary services. Real-time is a rolling SCED — every five minutes in ERCOT — settling deviations from day-ahead positions at real-time prices, so every participant's net settlement is the day-ahead quantity at the day-ahead price plus the real-time deviation at the real-time price.
Because the two are settled independently, purely financial participants can arbitrage expected spreads (ERCOT DAM energy-only offers, virtual bids/offers elsewhere), which in theory drives day-ahead prices toward the expectation of real-time prices. The DART spread remains structurally nonzero, reflecting risk premia, forecast error, and scarcity dynamics that only materialize in real time.