You've probably read the rule: tilt your panels at your latitude. It's close in the far south of the US, and it runs up to 14° too steep as you go north. The curve below is fitted to PVGIS satellite measurements for more than 35,000 cities and tested on 35,000 more, and the corrected numbers are: 25° at latitude 25, 35° at latitude 40, 41° at latitude 50, always facing true south in the northern hemisphere.
Fixed mount: °. If you adjust twice a year: ° in winter, ° in summer.
Don't know your latitude? Search your city, or use your zip code or UK postcode: every city page has the exact angles plus a monthly table.
Only up to about latitude 35. Beyond that it runs steep: by latitude 50 it is about 8° too steep, and by 60 about 14°.
Why the correction? Winter sun sits low and a growing share of light arrives as diffuse sky light rather than direct beam, and both effects flatten the true optimum as you go north. The latitude rule ignores them.
| Latitude | Old rule says | Calibrated optimum | Rule's error | Example city |
|---|---|---|---|---|
| 20° | 20° | 20.5° | +0.5° | Honolulu, HI |
| 25° | 25° | 24.6° | -0.4° | Miami, FL |
| 30° | 30° | 28.4° | -1.6° | Houston, TX |
| 35° | 35° | 31.9° | -3.1° | Albuquerque, NM |
| 40° | 40° | 35.2° | -4.8° | Denver, CO |
| 45° | 45° | 38.3° | -6.7° | Minneapolis, MN |
| 50° | 50° | 41.1° | -8.9° | London, UK |
| 55° | 55° | 43.6° | -11.4° | Glasgow, UK |
| 60° | 60° | 45.8° | -14.2° | Anchorage, AK |
South of latitude 35 the old rule is fine. By London's latitude it's 8° too steep; by Anchorage's, 14°. If someone tells you to tilt Glasgow panels at 56°, they're quoting the rule, not the sun.
Fixed rooftop panels sit at whatever pitch the roof has. Here's what common pitches cost against the optimum for a mid-US roof at latitude 40 (optimum 35.2°):
| Roof pitch | In degrees | Loss vs optimum at latitude 40 |
|---|---|---|
| 3/12 | 14° | about 6.8% |
| 4/12 | 18.4° | about 4.3% |
| 5/12 | 22.6° | about 2.4% |
| 6/12 | 26.6° | about 1.1% |
| 8/12 | 33.7° | none |
| 10/12 | 39.8° | about 0.3% |
| 12/12 | 45° | about 1.5% |
A 6/12 roof gives up about 1.7%. Tilt racks on a pitched roof rarely earn back their cost and wind risk for that; spend the attention on facing the panels south instead.
| Tilt error | Direct-beam loss |
|---|---|
| 5° | about 0.4% |
| 10° | about 1.5% |
| 15° | about 3.4% |
| 20° | about 6% |
| 30° | about 13.4% |
These are beam-geometry figures (1 minus the cosine of the error); diffuse light makes real annual losses smaller still. 5° is noise. 15° is a percent or three. Only being wildly off, flat in Scotland, vertical in Texas, costs real money.
A fixed mount at the calibrated optimum collects the most for zero effort. If you're willing to adjust twice a year, go steeper in winter (latitude × 0.875 + 19.2°, about 54° in Denver) and shallower in summer (latitude × 0.93 − 21°, nearly flat in the southern US). Monthly adjustment squeezes out the last few percent; every city page has the twelve-row table.
True south (true north below the equator), which differs from compass south by your local magnetic declination: up to 15° in parts of the US. Check a declination map or enable "true north" in your phone's compass settings before you drill.
The curve behind this page is a quadratic fitted to PVGIS v5.2 optimal-slope measurements for 35,342 cities, and tested on 35,332 others it never saw: on those it misses a typical place by 2.4° with no lean either way. The curve it replaced was fitted to 52 large cities and ran 1.9° too steep on average. What's left is climate and terrain, which latitude can't see: Paris and Harbin sit within 3° of latitude of each other and want 38° and 47° respectively, because one is cloudy and one is not. That is why each city page shows the angle PVGIS measured at that city rather than this curve. The curve itself is CC0; cite it freely.
The rules change when your roof moves. See when tilting RV panels beats lying flat, the caravan angles for British latitudes, how many watts a campervan needs, and how to aim portable panels at camp.
A boat swings on its mooring, which changes the answer completely: see why flat wins afloat. A fixed installation has the opposite problem and should be set for December rather than the annual average. For a weekend under canvas, the panel you need is smaller than you would guess.