How accurate is Sun Hours, really?
We don’t just claim our forecasts are accurate — we publish the real numbers, straight from opted-in users comparing our forecast to what their panels actually made. Statistical outliers excluded automatically. We’re on v1, and this trend line is how we hold ourselves to it.
All-time average
85.7%
Across every forecast checked
Best month
86.8%
Jul 2026
This month
85.4%
Aug 2026 so far
The trend, month by month
From Jun 2026 — the earliest month with enough contributors to publish responsibly — to today.
| Month | Average accuracy | Contributors |
|---|---|---|
| Jun 2026 | 84.1% | 3 |
| Jul 2026 | 86.8% | 11 |
| Aug 2026 | 85.4% | 14 |
19
users are sharing their real generation data with us right now
Every one of them opted in, for free, so we — and everyone else using Sun Hours — can see exactly how well the forecast is working. Thank you.
Want to help? Turn on Settings → Forecast Accuracy in the app. Anonymous, opt-in, delete anytime.
How we calculate this
Each day, opted-in users’ apps report what their panels actually generated. We compare that to the forecast we gave for the same day, and turn the difference into an accuracy percentage — 100% minus the size of the miss.
We group every comparison by calendar month and publish the average. A small number of statistical outliers — readings so far outside the normal range that they’d distort the picture, like a misconfigured system or a one-off data error — are excluded automatically, using standard outlier-detection statistics. There’s no manual blocklist: the same maths would catch a new outlier tomorrow without us having to touch this page.
Any month with fewer than three contributing users isn’t published, so a quiet month can never be read as one or two individuals’ actual generation figures. Sharing is opt-in, anonymous, and uses only a coarse (roughly 11km) location — see our privacy policy for details.
Honestly
We’re still on v1
Sun Hours launched as a two-question, no-account solar forecast app. We’re a small, independent team improving it every month based on exactly this kind of real-world feedback loop. This page is how we hold ourselves accountable — the trend line is the truth, whichever direction it points.
See what we’re building next →Questions
How accurate are Sun Hours' solar forecasts?+
Averaging 85.7% accuracy across all opted-in users as of Aug 2026, calculated from real predicted-vs-actual daily generation data. Statistical outliers are excluded automatically. See the full monthly trend and methodology on this page.
How is Sun Hours' forecast accuracy measured?+
Opted-in users' apps report what their panels actually generated each day. We compare that to the forecast we gave for the same day, expressed as 100% minus the size of the miss, then average it by month. Outlier devices — readings so inconsistent with everyone else's that they'd distort the picture — are excluded automatically using standard statistical outlier detection, not manual judgement calls.
How many users share forecast accuracy data with Sun Hours?+
19 users have shared forecast accuracy data in the last 90 days. It's entirely opt-in, free, and anonymous — no account required.
Is my data private if I share my forecast accuracy?+
Yes. Sharing is opt-in and anonymous — there's no account and no personal identifier. Your location is only ever stored rounded to roughly 11km, never your exact address, and you can delete your shared data at any time from Settings.
Why don't all months appear on the accuracy chart?+
Any month with data from fewer than three contributing users is withheld from this page, so a quiet month can never be traced back to one or two individuals' actual generation figures.
How often is this accuracy data updated?+
Automatically, every time this page loads — there's no manual step on our end. As more users opt in and more days pass, the trend updates itself.
Get a forecast this accurate for your own panels
Free, two questions, no account. Works with plug-in, balcony, and rooftop solar across every major inverter brand.