The system is 36 panels rated 425 W each. That number describes a panel lying square to the sun in a laboratory, and it is the number on the contract, the permit and the incentive paperwork.
No roof delivers it. This one spreads its panels across four faces of a hip roof, which costs a fifth of the total before any electricity is generated — and that is the smallest of the three deductions between the rating and the meter. Over 657 days the array delivered 42% of what those 36 panels would make perfectly aimed under a permanently clear sky.
Start with the number on the contract — 36 panels at 425 W — and assume every one of them faces the sun squarely, all day, under a sky that is never cloudy. Nothing on any roof does that. Each step down is a different reason, and only the last of them is weather.
The first step is the roof itself, and it is settled by geometry alone: no measurement, no year of data, just where the four faces point. The second is everything that costs the same amount every single day — heat, wiring, the inverters, and the neighbor's tree over the west face. The third is the sky.
Each day is compared with what a clear sky would have given that day, at that sun angle, on these four faces. So the seasons are already divided out: a good December day and a good June day both land near the top, even though one made a third of the energy.
The top edge is flat, which is the useful part. No day ever beats about 80%, because the losses that produce that ceiling are there in all weather. Everything below the edge is cloud. The column of points on the floor in January is the week the panels were under snow.
The same days as a distribution. It has no useful average. The mass sits at both ends — clear days against the ceiling, overcast days near the floor — with little in between, so the mean of 51% describes very few days.
The utility offers a time-of-day rate whose overnight band its own guide calls intended for electric vehicle charging. It runs from 10pm to 6am, and it is priced at less than half the flat rate. On its face it suits a house that charges a car overnight.
Credits net within a band and not across bands, so energy exported at noon offsets consumption at noon only. An array generates between sunrise and sunset, so over the record it exported __SUPEREXPORT__ into the overnight band against __SUPERIMPORT__ imported from it.
Under that rate __STRANDED__ of credit would remain in bands with no consumption left to offset, while the overnight band was billed in cash. The flat rate has a single band, so every exported kilowatt-hour offsets any imported one. The installation is billed on the flat rate for this reason.
Because each day is already measured against its own ceiling, this is not a chart of how much sun there was. It is a chart of how much of the available sun was actually collected, which is a question about cloud rather than about the calendar.
The same 36 panels, modeled under the same clear sky, arranged three ways. A hip roof spreads them over four faces, and two of those face north-north-west and east-north-east.
Laying every panel flat would collect more than this roof does. A shallow tilt pointed away from south is worse than no tilt at all.