Glossary

A proposed successor to ERCOT's 4CP: allocate transmission costs over twelve monthly peaks instead of four summer ones.

12CP allocates transmission costs using the highest system-demand interval of every month, not just the four summer months. The PUCT is considering moving ERCOT from 4CP toward a 12CP-style methodology, with a final rule mandated by the end of 2026. The practical effect: peak-chasing spreads across the whole year, winter peaks start to matter, and the response strategy that worked under 4CP needs rethinking. WattMarkets ships a 12CP preview program so you can see the new exposure before the rule lands.

The four 15-minute intervals, one per summer month from June through September, when ERCOT system demand peaks. They set next year's transmission charges.

4CP refers to the single highest 15-minute settlement interval of ERCOT system demand in each of June, July, August, and September. A facility's average demand across those four intervals sets its transmission cost allocation for the following calendar year. Because the stakes recur annually and the intervals are unknowable until the month closes, an industry of prediction, curtailment, and second-guessing exists around catching them. WattMarkets publishes its 4CP calls and grades them against official settlements.

PJM's five highest daily peak hours each summer. Your load during those hours sets your capacity charges for a full year.

From June 1 through September 30, PJM tracks the highest hourly load of each business day across the whole RTO. The five biggest daily peaks are the coincident peaks, and a customer's average demand during those five hours becomes its Peak Load Contribution for the delivery year starting the following June. Weekends and NERC holidays are excluded by rule, and PJM publishes the official five in mid-October. Curtailing through the right five hours can save six figures per megawatt at current capacity prices.

WattMarkets' highest alert level: curtail now, the predicted peak interval is imminent.

ACTION is the third and highest escalation level (after WATCH and WARNING): an in-day, interval-specific call that the peak window is imminent, delivered by SMS with the highest-risk interval and a be-off-by time. ACTION alerts are deliberately rare, never more than two per day, because alerts that fire too often get ignored.

Running the current model against historical seasons to measure how it would have performed. Always labeled, never mixed with live calls.

A backtest scores the model against past summers using data available at the time. It is essential for calibration and honest exactly once: the moment backtest numbers are presented alongside (or as) live performance, they become marketing fiction. WattMarkets badges every scorecard row BACKTEST or LIVE CALL and never aggregates across the two.

Utility-scale batteries that charge and discharge against price and peak signals. Several GW now operate in ERCOT.

ERCOT has one of the world's fastest-growing grid-battery fleets. Around predicted peaks, batteries discharge to capture high prices, which shaves the observed peak, while others charge earlier in the day, adding load. Battery behavior is a first-order driver of whether a forecast peak actually materializes, which is why WattMarkets carries an explicit est. battery driver in every prediction.

PJM's capacity auction. Its clearing price is the dollar figure your PLC multiplies against.

The Base Residual Auction is where PJM procures capacity commitments for a future delivery year, and its clearing price sets what a megawatt of Peak Load Contribution costs. Prices cleared at the FERC-approved cap for the 2025/26, 2026/27, and 2027/28 delivery years, with 2027/28 at $333.44 per MW-day across the RTO, roughly $121,700 per MW-year before zonal scaling. Rising BRA prices are why 5CP management keeps getting more valuable.

Each logged call is hashed together with the previous one, and published terminal hashes make silent history-editing detectable.

Every call WattMarkets logs is written to an append-only table where each row includes a SHA-256 hash of the previous row plus its own content. The day's terminal hash is published at /v1/scorecard/attestations. Change any historical row and every hash after it breaks. Anyone can check for themselves that the history was never edited.

NYISO's capacity-market seasons: summer (May-October) and winter (November-April). The summer peak sets ICAP tags.

NYISO's capacity market runs in two capability periods each year: summer, May through October, and winter, November through April. The statewide coincident peak that sets ICAP tags lands in the summer period. In the four settled years from 2022 through 2025 it fell in June, July, or September, always on a business-day afternoon. Tags derived from one summer's peak apply to the following capability year.

The industry name for your PLC: the number your capacity charges key off.

Capacity tag is shorthand for the Peak Load Contribution. Energy managers talk about 'managing the tag': curtailing during likely coincident-peak hours so next year's tag, and with it a year of capacity charges, comes in lower. In PJM the tag is set by the five summer coincident peaks. The equivalent lever in ERCOT is the four 4CP intervals.

Your facility's demand at the moment the whole grid peaks, not your own maximum. Transmission charges are allocated on it.

A coincident peak (CP) is your demand measured at the exact interval when the entire system hits its maximum. Your non-coincident peak, by contrast, is whenever you happen to use the most. In ERCOT, transmission costs are allocated to consumers based on their share of system load during four summer peaks (4CP). If your facility is idle during those intervals, you pay far less for transmission all next year. That is why predicting when the system peak will land is worth real money.

Deliberately reducing your facility's electric demand for a window of time. The action a peak alert exists to trigger.

Curtailment is the deliberate reduction of load (pausing production lines, pre-cooling and coasting, shifting batch processes) during a predicted peak window. Under 4CP, a few well-timed hours of curtailment per summer can cut next year's transmission charges by five or six figures for a large facility. The cost of a false alarm is real too: lost production hours. That trade-off is why false alarms matter as much as catches.

Day-ahead to real-time price difference. The market grading its own forecast, in dollars.

DART is trader shorthand for day-ahead versus real-time. When real time settles above the day-ahead price, buyers who waited paid more than the market predicted. When it settles under, day-ahead carried a premium. Congestion, weather misses, and plant outages drive the spread, and scarcity afternoons can blow it out by hundreds of dollars per megawatt-hour. WattMarkets grades ERCOT's spread daily, by hour and by load zone.

The market that prices every hour of tomorrow, cleared each afternoon. Real time then grades it.

Most wholesale power trades a day early. In ERCOT's voluntary day-ahead market, buyers and generators commit to prices for each hour of the next day, with results published around 12:30 CT. The real-time market then settles what each hour actually cost. The difference between the two prices is the market's forecast error in dollars, and the WattMarkets DART page tracks it hour by hour.

PJM's June-through-May billing year. Summer 2026 peaks set tags for the 2027/28 delivery year.

PJM plans and bills capacity on a delivery year running June 1 through May 31. The five coincident peaks of one summer set each customer's PLC for the delivery year that starts the following June, priced at that year's Base Residual Auction clearing price. The lag matters: the money saved by curtailing this summer shows up on bills starting next June and continues for twelve months.

Programs and behavior where loads reduce consumption in response to grid conditions or price signals.

Demand response (DR) is any arrangement where electricity consumers reduce or shift load in response to a signal: a price spike, a reliability event, or a predicted coincident peak. 4CP avoidance is a species of demand response driven purely by transmission-cost allocation. At scale it becomes self-referential: if several gigawatts curtail on a predicted peak day, the peak moves. WattMarkets models that feedback explicitly (the est. curtailment driver in every prediction).

The Electric Reliability Council of Texas: grid operator for ~90% of Texas load, electrically isolated from the rest of the US.

ERCOT operates the electric grid and wholesale markets for most of Texas, serving 27+ million customers. It is its own interconnection, largely isolated from the Eastern and Western US grids, which keeps it outside FERC jurisdiction and makes Texas policy (the PUCT and the legislature) unusually decisive for how its markets work. 4CP transmission cost allocation is one of those Texas-specific constructs.

Your demand during the single hour New York peaks each summer. It sets your capacity charges for the following capability year.

An ICAP (Installed Capacity) tag is a New York facility's demand measured during the one hour of the summer when statewide NYCA load hits its annual maximum. Suppliers must buy capacity for every megawatt of tag their customers carry, at auction prices that differ sharply by locality (New York City clears far above the rest of the state), and those costs ride the bill for twelve months. Unlike multi-peak programs, there is no partial credit: the season is one hour. WattMarkets publishes a daily call on whether today is likely to set that hour, and grades every call on a public track record.

Independent System Operator / Regional Transmission Organization: the entities that run wholesale grids and markets in the US.

ISOs and RTOs (the terms are near-interchangeable) operate transmission grids, run wholesale energy markets, and coordinate reliability across their footprints: ERCOT in Texas, PJM in the mid-Atlantic, MISO in the Midwest, CAISO in California, and others. Several use coincident-peak constructs similar to ERCOT's 4CP. PJM's 5CP peak-load contribution is the closest cousin.

ERCOT's published prediction of system demand, updated hourly. The dashed line on our charts.

ERCOT publishes rolling forecasts of system load by hour, driven primarily by weather models. Forecast vintages matter: a 2 PM forecast for the 5 PM hour is far more accurate than yesterday's. WattMarkets stores every vintage, displays the forecast as the slate-blue dashed line, and grades its own probability calls against what actually happened.

ERCOT settlement region for energy prices. Not the same thing as a weather zone.

A load zone is a settlement region used for pricing energy in ERCOT's markets (LZ_HOUSTON, LZ_NORTH, LZ_SOUTH, LZ_WEST, plus NOIE zones). Prices settle per load zone, while weather and load reporting use the eight weather zones. The two get confused constantly: load zones are for pricing, weather zones are for weather data.

Non-Opt-In Entity: a Texas municipal utility or co-op that hasn't joined retail competition. Many actively chase 4CP.

NOIEs (municipally owned utilities and electric cooperatives that did not opt into retail competition) serve millions of Texans. They pay transmission costs on their own 4CP shares, and many run aggressive peak-shaving programs, dispatching generation or curtailment on predicted peak days, because a successful shave benefits every member. NOIE behavior is a meaningful chunk of the reflexive curtailment the model estimates.

Your demand at your transmission zone's annual peak. It sets network transmission charges the way PLC sets capacity charges.

Network Service Peak Load is the transmission-side sibling of the capacity tag. It is set by a customer's demand at the moment its own transmission zone hits its annual peak, and network transmission charges are allocated on it for a year. Zone peaks usually land on or near 5CP days, so curtailing the right summer afternoons often reduces both tags at once.

The whole New York grid as NYISO operates it: 11 load zones, lettered A through K, from Buffalo to Long Island.

The New York Control Area is the footprint NYISO balances: all of New York State, divided into 11 load zones lettered A (West, around Buffalo) through K (Long Island). Statewide load, the NYCA total, is the sum of the 11 zones, and the hour it peaks each summer sets ICAP tags. Zone prices differ because transmission between upstate generation and downstate load congests. Zone J (New York City) routinely clears at multiples of the western zones.

p_raw is the peak probability before accounting for curtailment and batteries. p_final is after. Both publish with every call.

p_raw answers: if nobody reacted, how likely is today to set the month's peak? p_final answers: given the curtailment and battery response we estimate, how likely is it really? The gap between them is the market reacting to itself. Most services hide this adjustment. WattMarkets publishes both numbers in every API response and every driver panel.

The highest rate of electricity consumption over a period, measured in MW or GW.

Peak demand is the maximum instantaneous (in practice, interval-averaged) load over some window: an hour, a day, a season. ERCOT's all-time system peak stands above 85 GW, set in the summer heat. Grid infrastructure is sized for the peak, not the average, which is why cost allocation keys off peak intervals. The peak is what the wires were built for.

Your average demand during PJM's five coincident peaks. Capacity charges multiply against it for a year.

The Peak Load Contribution is a customer's average metered demand across PJM's five coincident peak hours of the prior summer. It is recalculated once a year and applied for the delivery year that starts June 1. Suppliers pass capacity costs through as PLC times the capacity rate, so a lower PLC means a lower bill every month for twelve months. The only way to move it is to use less power during the five hours that end up counting.

The named inputs behind every WattMarkets probability: forecast max, month max, temp anomaly, curtailment and battery estimates.

Every probability WattMarkets publishes ships with its inputs: the forecast maximum, the month's standing maximum, the margin between them, temperature anomaly, estimated curtailment response, estimated battery activity, and day-of-week factor, plus p_raw (before demand-response adjustment) and p_final (after). The call is never naked. You can inspect, adjust, and argue with every input in the Sensitivity Lab.

WattMarkets' published record of every call: hits, misses, and false alarms, graded against official settled peaks.

The Scorecard lists every prediction with its outcome: precision (of flagged days, how many held the true peak), recall (of true peaks, how many were flagged), average lead time, and false-alarm hours, next to a naive baseline for context. Misses are explained, not buried, and each links to a Replay. An append-only hash chain (published attestations) makes retroactive editing detectable by anyone.

The Public Utility Commission of Texas: ERCOT's regulator, currently weighing the 4CP-to-12CP transition.

The Public Utility Commission of Texas regulates ERCOT, sets transmission cost allocation methodology, and approves TCOS rates. Its open rulemaking on transmission cost allocation (the 4CP successor question) is required to conclude with a final rule by December 31, 2026. No other regulatory event matters more for how this product's users will operate next season.

Re-running a real historical day through the live prediction pipeline, hour by hour.

Replay takes a captured real day, the load curve as it unfolded, and runs the exact production model over it, showing what WattMarkets would have said at each hour. The archive includes the famous June 30, 2024 Sunday peak that day-of-week models missed. Golden replay days are frozen in CI: a model change that flips one from HIT to MISS cannot ship without an explicit version bump.

The WattMarkets page where you adjust the model's inputs (curtailment, battery, temperature, forecast bias) and watch the probability respond instantly.

The Lab exposes the model's final layer in your browser: drag the curtailment estimate up, watch the probability fall, and share the resulting scenario as a link. It exists because 'your model is wrong about curtailment depth' should be a testable claim, not a comment-section fight. Replay mode runs any archived day through the exact production pipeline, hour by hour.

The 15-minute block ERCOT uses to measure and bill demand. Each 4CP peak is one settlement interval.

ERCOT settles demand in 15-minute intervals. The official 4CP determination uses the highest 15-minute interval of system demand per summer month, published after settlement quality checks. Live telemetry is faster but provisional, which is why WattMarkets displays provisional peaks in-season and reconciles against official settled intervals when ERCOT publishes them.

The all-clear signal: today's peak risk has passed and curtailed loads can resume normal operation.

A stand-down is the explicit end-of-risk signal, as important as the alert itself. Without it, a facility that curtailed at 3 PM is left guessing whether 6 PM is safe. WattMarkets issues STAND DOWN when the day's residual peak risk drops below threshold, with the day's observed maximum and its time, so operators can resume with confidence and quantify what the day cost.

The regulated revenue transmission owners collect. 4CP shares decide who pays what portion of it.

Transmission Cost of Service is the total regulated cost of building and operating the transmission network, approved by the PUCT for each transmission owner. In ERCOT, that cost pool is divided among load-serving entities in proportion to their 4CP shares. Your transmission rate is a share of that fixed pool, and the 4CP intervals decide the share.

How far today's temperature sits above or below the seasonal normal. The strongest single peak driver.

Load in Texas summers is air conditioning, and air conditioning is temperature. The anomaly, degrees above or below the seasonal normal for that month and zone, carries more predictive signal than the raw temperature. WattMarkets computes a load-weighted anomaly across the eight weather zones, displays it per zone on the map, and feeds it to the model as a first-class driver.

PJM's official peak measure: observed load plus an estimate of the demand response that curtailed away.

When PJM determines the official five coincident peaks in October, it uses unrestricted peaks: the load that would have occurred had demand-response programs not curtailed. On a day with heavy curtailment, the unrestricted peak can rank higher than the observed load suggests. Live standings built from observed load are honest about this: they can differ from the official determination on DR-heavy days, and the October publication settles it.

WattMarkets' escalation levels. WATCH means elevated day-ahead risk, WARNING means high day-of risk.

The alert ladder borrows deliberately from severe-weather practice. WATCH (day-ahead, probability above the watch threshold) means plan for a possible curtailment tomorrow. WARNING (day-of) means today is likely a peak day, so confirm readiness. ACTION means execute. Every level carries its own glyph and color, so no single channel (color, word, or symbol) is ever the only signal.

One of eight ERCOT regions with similar climate, defined by ZIP code, used for load reporting and forecasting.

ERCOT divides its footprint into eight weather zones: Coast, East, Far West, North, North Central, South Central, Southern, and West. Each is a geographically contiguous area with similar climate, formally defined by five-digit ZIP codes in the Load Profiling Guide. Actual load and temperature forecasts are published per zone. The WattMarkets Texas map draws these exact boundaries, built from ERCOT's official ZIP-to-zone table.