What we found after studying seasonal weather and ISO price patterns

As winter approaches, we studied weather data since January 2025 in relation to one of our favorite topics, TB4. We expected cold snaps and heat waves to explain most of a battery's high and low revenue days, but we realized weather is not as important for it as we thought.

Instead, power demand turned out to track temperature far more closely than battery spreads do, and the link between weather and spreads is not uniform across the US.

Some context on what we measured: TB4 is each day's gap between the average day-ahead price of the four most expensive hours and the four cheapest. It is a benchmark for what a four-hour battery could gross per MWh from one daily charge and discharge cycle, with perfect timing and before losses. We tracked it at 34 hubs across all seven US ISOs, from January 2025 to October 2026.

How closely electricity demand and battery spreads track temperature

We started with a simple question. Demand tracks temperature closely, but do battery spreads? For each hub, we fitted daily demand and daily TB4, on a log scale, against how far the day was above or below 65°F, plus the day of the week. Then we measured how much of the daily variation each fit explains.

The gap was wider than we expected. In ERCOT, temperature and weekday explain 86% of the daily variation in demand but only 12% of it in spreads. ISO New England was the exception, at 60% for spreads. So in Texas and California, temperature and weekday account for little of a day's spread. In New England, they account for much more.

ERCOT battery spreads in a hotter summer

Texas showed what we expected. Summer 2026 was hotter than summer 2025 at all four ERCOT hubs, but with its battery fleet still growing, we expected spreads to narrow.

They did. At Houston hub, the median summer day fell from $36 to $24, though the average day was 1.5°F hotter. Real-time spreads at the four ERCOT hubs fell too, by 14% to 24%. Over the same months, the median day-ahead spread rose at every MISO and PJM hub we studied.

We checked what else had changed. ERCOT's average summer demand rose 4.8 GW, or 7.5%. Wind and solar output rose 4.4 GW, so the load left for other plants grew by less than 1%. Gas cost about $3 per MMBtu in both summers. The most obvious explanation is the growing BESS fleet across Texas, which leads to lower bids and compressed spreads. This is visible on our ERCOT leaderboard too, where the fleet grew from about 13 GW in summer 2025 to about 20 GW in summer 2026.

Are battery spreads wider on hot days or cold days?

Next we wanted to know whether spreads run wider on hot days or on cold days. We fitted each hub's daily TB4 against degrees above and below 65°F, holding the calendar and the gas price fixed, and read off the change for a day 10°F hotter or colder.

At PJM's Eastern hub, the fit puts the spread on a day 10°F above 65°F at about 4.4 times that of a 65°F day. At ERCOT Houston, the fit puts it only 37% wider. Cold scores higher only at the four ERCOT hubs and CAISO's ZP26, and at all of them except ERCOT South the two ranges overlap. These counts compare our best estimates, so the order at any single hub is not settled.

One hypothesis is that heat is routine in Houston. In this window, 364 days there were at least 10°F above 65°F, against 128 at PJM Eastern and 65 at Boston. Markets that live with heat may have built for it, while markets that see it rarely pay for it. We have not tested it.

Cold days stand out more in the South. In our fit, a day 10°F below 65°F has a spread about 60% wider at MISO Louisiana and at MISO's Texas hub, which sits in southeast Texas outside ERCOT, against 10% at MISO Minnesota. When we left out the January 2026 storm weeks, the order of the hubs' cold estimates barely changed.

What comes with electricity price jumps when temperature holds steady

Temperature left most of the spread variation unexplained, so we ran a separate test on price jumps. This one looks at a hub's daily average day-ahead price, not its TB4. We took every day a hub's price jumped at least 50% and $10/MWh above the same weekday a week earlier while its temperature stayed within 5°F, and checked which other conditions held.

Each market had its own pattern. In ERCOT, wind was down on 83% of jump days, about three times the usual rate. In MISO, PJM and New York, a third to a half of jump days came with a local price gap, three to five times the usual rate. That means the hub's price sat unusually far from the typical price across its market that day, which can happen when transmission lines are congested. In New England, gas had risen on 43% of jump days. California had no strong pattern. Wind drops and gas rises each came at under twice their usual rate, and a third of jump days had none of these conditions.

What we saw at each power hub

Putting it together, here is what we saw in each market. The spread column comes from the temperature fits, and the jump column from daily average prices.

We will keep watching this through the winter on our weather, load and price map, which compares each hub's temperature and price with the week before. For how TB4 spreads are earned and traded, see our TB4 analysis.

Weather: Open-Meteo ECMWF IFS, CC BY 4.0. Demand and generation: EIA-930. Gas price: EIA Henry Hub. Prices: ERCOT, CAISO, PJM, MISO, NYISO, SPP and ISO-NE.