Chance of a plunge, next 7 days
How likely Octopus Agile is to dip below 0p — to pay you to use electricity — on each of the next seven days. Depth is probability, read against the scale; the time beneath each sounding is where in the day it would most likely land.
Price sounding
Published Agile half-hourly rates — below sea level (0p) you are paid
Plunge probability by slot
Half-hour risk across the model's latest 24h window
Conditions for heavy weather
The recipe for a plunge
Observations
What the data shows
Plunge odds vs wind
Share of days with a plunge, by decile of daily mean wind capacity factor
Wind vs deepest price
Each dot is one day: mean wind capacity factor vs lowest half-hourly price
When plunges strike
Plunge slots by hour of day — overnight wind and midday solar
Season by season
Plunge slots per calendar month
What the model weighs
Permutation importance — AUC lost when each feature is shuffled
Sea trials
How it would have done
Calibration
When the model says X%, how often does a plunge actually follow?
Alert thresholds
Tune to taste: catch more plunges or ring fewer false alarms
| Alert at | Plunge days caught | False alarms | Precision |
|---|
Reliability by lead
Skill measured against historical forecast vintages, 1–7 days out
| Lead | AUC | Caught @≥50% | Precision | Caught @≥20% |
|---|
Every real plunge day, hindcast
The probability the model gave each day that actually plunged
| Date | Model said | Slots | Deepest |
|---|
Ship's log
Track record
Every live forecast this instance has issued, scored against the prices Octopus actually published.
What we said, day by day
Latest published probability at each lead — read a row left-to-right as the day approached
| Target day | 7d out | 6d | 5d | 4d | 3d | 2d | 1d | Outcome |
|---|
| Issued | Window start | Forecast | Outcome | Plunge slots | Deepest price |
|---|