Secret Vote · The Arbitrage Window

kWh Cost

Batteries don’t make electricity — they move it in time. Charge when Austin’s grid price is cheap, discharge when it’s expensive, and pocket the difference. This page asks a simple question: on a typical day, how big is that difference, and how often is it worth catching?

NO RAW HOURLY PRICE FEED — the underlying LZ_AEN dataset behind this project has real daily revenue figures, not a literal hour-by-hour price series. The charts below use two kinds of data: a real, LZ_AEN-derived implied daily spread (revenue ÷ MWh discharged, the same numbers behind The Battery Ledger) for the “how big” and “how often” questions, and an explicitly-labeled illustrative shape for the “what does a day look like” question. See Methodology below before reading too much into any single number.
Median daily spread
$48.5/MWh
Jupiter Power, all 1,242 modeled days
Days worth arbitraging
89%
days with spread ≥ $20/MWh
2023 → 2026 yearly avg
$262 → $75.8
spread has compressed as more storage came online statewide
Peak single day
$5,473/MWh
Jun 20, 2023
Days modeled
1,242
Jan 2023 – May 2026, same dataset as The Battery Ledger
01

What a day looks like

Solar pushes midday prices down; demand stays high after sunset. The gap between the cheap trough and the evening peak is the window a battery arbitrages. This shape is illustrative — a relative index, not measured dollars — tracing the pattern ERCOT’s own market monitor describes for the real-time market.

Illustrative shape — not measured $/MWh025507510012am6am12pm6pm11pmcheapest, midday solarpriciest, evening ramp
02

The spread over time

Real numbers: each day’s simulated arbitrage revenue divided by the MWh discharged — an honest approximation of the average $/MWh price spread a perfectly-timed battery captured that day. Y-axis is log-scaled; a handful of extreme spike days (Winter Storm Uri’s successors, summer 2023’s heat) would otherwise flatten every ordinary day to a hairline.

$0$100$500$1,000$2,500$5,0002023202420252026
03

How often is it worth it

Every modeled day, sorted from the biggest spread to the smallest — a duration curve. The dashed lines mark what share of days cleared a few reference spreads. It doesn’t take a Winter Storm Uri for arbitrage to pay; it takes an ordinary handful of hours where solar drops out and demand hasn’t.

$0$100$500$1,000$2,500$5,000best dayworst day89% of days ≥ $2049% of days ≥ $5021% of days ≥ $10010% of days ≥ $200
04

ERCOT-wide context, 2025

Real, published market-wide figures — but not AE/LZ_AEN-specific. Austin Energy’s own zone tracks close to these system numbers most hours, diverging mainly during local congestion.

Average price by ERCOT zone, 2025
West
$41.6
South
$40.8
Houston
$40.1
North
$36.9
System average: $36.9/MWh
When price spikes happen
Evening ramp, 7–9pm30%
Morning ramp, 6–8am7.5%
Midday, 9am–2pm2.7%
Share of 2025 price-spike hours falling in each window.
Capacity offered at $0/MWh or below
35%
vs. 32% in 2024 — more negative-price hours to charge into.
Price-spike count, year over year
-40%
2025 vs. 2024 — fewer, not more, extreme-price events statewide.
05

Methodology & caveats

01
“Spread over time” and the duration curve are real, derived numbers. Each is a day’s simulated arbitrage revenue (from the same perfect-hindsight model as The Battery Ledger: charge in that day’s actual cheapest hours, discharge in its priciest) divided by MWh discharged. That’s an honest approximation of the average spread captured, not a literal hourly price series — a day with two enormous spike hours and one with many moderate ones can land on the same implied spread.
02
The hour-of-day shape is illustrative, not measured. No raw hourly LZ_AEN price series is available in this project or downloadable from its source (chrisgillett.org/ae-congestion, which exposes interactive charts but no data export). The shape shown traces the qualitative pattern ERCOT’s independent market monitor describes for real-time prices — a 0–100 relative index, deliberately not styled as dollars.
03
The ERCOT-wide context (section 04) is real but differently scoped. It covers the whole ERCOT market in 2025, not Austin Energy’s LZ_AEN zone specifically. Treat it as background on why the arbitrage window exists market-wide, not as AE’s own numbers.
04
Log-scaled axes. Daily and monthly spread charts use a symmetric-log scale so ordinary days stay readable next to rare four- and five-figure spike days. Read the gridline dollar labels, not the visual gap between points, when comparing magnitudes.
05
Same dataset, same caveats as The Battery Ledger. Perfect-hindsight dispatch, 85% round-trip efficiency, energy arbitrage only (no ancillary services, capacity payments, or resilience value), and Jupiter Power’s 2-hour duration is an estimate, not a disclosed figure. See that page for the full methodology on the underlying revenue model.