How WattPair reaches a verdict
WattPair does not look answers up in a table. Every verdict on this site is computed, at build time, from the load a machine actually presents and the limits its power station actually publishes: continuous output, start-up surge and how long that surge lasts, power factor, waveform, usable energy, inverter loss and the station's own consumption.
This page is the whole of it. Each threshold below is printed from the same file the engine reads, tagged with where the number came from, and the site refuses to build if a rule described here is not one the engine carries.
The four things a verdict can be
Every threshold, and where it came from
Four kinds of number go into a verdict, and they are not equally authoritative. The tag on each row says which kind it is.
Continuous output
The first question is whether the load fits inside the inverter at all. It is answered against the duty-weighted draw the load actually presents, not against the wattage printed on its box.
Above this share of continuous output the pairing stops being comfortable: the fans run flat out, the inverter is hot, and one extra load trips it. Strictly greater — a load sitting exactly at the threshold is still a clean yes.
Product rule compat.marginal_load_ratio
Past the rated continuous output the overload protection acts. There is no partial credit here and no boost mode is counted toward it.
Our judgement compat.max_load_ratio
Binary floating point cannot represent 0.82 exactly. Without a tolerance, a load computed at precisely the threshold would fail the comparison on some inputs and pass on others. Every boundary in the engine is tested at the boundary.
Our judgement numeric.comparison_epsilon
Start-up surge
Most pairings that fail in a kitchen at 2 a.m. fail here: the running watts fit and the start does not. The engine treats surge as three different phenomena that share one word, because the load library shows their durations differ by three orders of magnitude.
A manufacturer surge rating is a peak figure, and the data schema records that most brands publish no duration alongside it. Consuming almost all of an unqualified peak is not the same as a promise that it starts.
Our judgement compat.surge_marginal_ratio
Demand beyond the published surge ceiling is a hard no.
Product rule compat.max_surge_ratio
They are often read from different rows of the same manufacturer table. On rooftop air conditioners the locked-rotor figure and the published multiplier differ by hundreds of VA, so the engine takes whichever demands more.
Manufacturer document compat.surge_resolution
An inverter cannot surge below its own continuous rating, so this floor is always true and always pessimistic. Refusing to evaluate instead would flag a laptop as questionable on a machine that could run twenty of them.
Manufacturer document compat.surge_ceiling_fallback
Two soft-start vendors publish 70% and "up to 75%". The engine takes the conservative end, applies it only to load types that have a locked rotor to soften, and never lets the softened figure fall below the motor running watts.
Manufacturer document compat.soft_starter_surge_reduction_pct
Power factor and VA
A power station is rated in watts; the switching stage inside it is limited by current, which is volt-amps. For a motor with a poor power factor the two numbers are not close.
The check appears once the load record actually carries a power factor, or the station publishes a VA ceiling. It is never inferred into existence to make a page look thorough.
Our judgement compat.apparent_power_check
Portable stations almost never quote VA. Reading the watt rating as the VA ceiling is the conservative direction, and every verdict that leans on it says so on the page.
Our judgement compat.apparent_power_ceiling_basis
Each default is the lowest figure measured anywhere in our own load library for that class of machine, not a rule of thumb. A verdict that used one is labelled.
Manufacturer document compat.power_factor_defaults_by_load_type
Waveform
A handful of appliances are documented by their makers as requiring a pure sine wave. This is one of the two checks on the site that exists to protect hardware and health rather than to refine an estimate.
When the appliance maker demands a pure sine wave and the station maker never published its waveform, the pairing cannot earn a green verdict from silence. The same posture as the solar temperature-coefficient rule below.
Manufacturer document compat.on_unknown_waveform_for_pure_sine_load
Variable-speed machines
An inverter-driven compressor does not cycle on and off; it modulates. Its nameplate carries three points — minimum, rated and maximum — and reading the wrong one is how a heat pump that draws its maximum all winter gets called comfortable.
The verdict is taken at the maximum. One manufacturer note in the library says it outright: size a winter backup on the maximum heating figure, not the rated cooling one.
Manufacturer document compat.modulation_verdict_point
Hours are quoted at the rated point, which is the one a reader can look up, with the minimum and maximum shown as the ends of the interval.
Our judgement compat.modulation_runtime_point
For every other category, a low and a high figure describe a spread across different models, not one machine turning itself down. Treating a range of laptops as a modulation curve produces nonsense.
Manufacturer document compat.modulation_load_types
Runtime
Hours are the most-quoted and least-defensible number in this category. The engine computes them from usable energy and a duty-weighted denominator, or it declines to print them at all.
One manufacturer manual states the formula outright, with self-consumption in the denominator. Leaving that term out overstates the runtime of a small load by a factor, not by a rounding error.
Manufacturer document compat.include_station_idle_draw
Idle draw spans an order of magnitude between brands. Where it is unknown the term is dropped and the page says the figure is optimistic, because inventing it would put an unsourced number under every runtime on the site.
Our judgement compat.default_idle_draw_w
Portable inverters measure in the high eighties at mid load and worse at low load. The default is set low deliberately, and any figure that used it is marked as an estimate.
Our judgement compat.default_inverter_efficiency
If usable energy was already derived by multiplying an efficiency in, applying it again would silently subtract the same loss twice. The basis travels with the record.
Our judgement compat.efficiency_application_by_basis
There is no defensible typical duty cycle across appliance categories. Without one the page shows no runtime, which is the honest output.
Our judgement compat.default_duty_cycle
Refrigerators have two defensible duty cycles: the constant inside a regulator test formula, and the run time the manufacturer says the compressor actually keeps. They are not interchangeable, so the runtime is an interval and the headline is the field figure.
Manufacturer document compat.duty_cycle_point_basis
Solar input
Panel matching is a legality question before it is a yield question. A string that is legal at noon in July can exceed the charge controller on a cold clear morning, because panel voltage rises as temperature falls.
Substituting the room-temperature open-circuit voltage would understate the risk on the one check that can destroy a charge controller. The engine refuses to answer instead.
Our judgement solar.on_missing_voc_temp_coefficient
Crossing into the margin is called questionable, not illegal — it is a design cushion, not a device limit.
Schema default solar.voc_design_margin_pct
The right number is a property of your location, not of the hardware, so a result computed with the default says so and asks for your local record low.
Schema default solar.default_design_min_ambient_c
Below the window the tracker never starts. That is a real failure, not a warning.
Our judgement solar.vmp_below_window_is
Soiling, angle and cell temperature all reduce yield, and we have no measured curve of our own to quantify them. Nothing multiplies them in, and nothing pretends to.
Our judgement solar.power_derate_factor
Surge, by how long it lasts
The durations below are read from manufacturer documents stored in the load library. They span three orders of magnitude, which is why a single surge rule would be wrong for most of them.
A power supply charging its input capacitor. Gone within three mains cycles; a station rides it out.
Noted, not enforced
Locked rotor until the shaft turns. This is what a surge rating actually describes.
Checked against the surge ceiling
Longer than any motor start in the library, but still inside a plausible surge window.
Checked against the surge ceiling
A glow plug, not a surge: it can run for minutes. Compared against continuous output instead.
Judged as a continuous load
When nobody published an inrush figure
Absence means different things for different machines. A resistive element has no locked rotor to publish; an induction motor does, and a missing figure there can hide a load that never starts. The load library blocks page generation for that case rather than guessing at it.
Assumed power factors
Used only when a load record carries no power factor of its own, and flagged on the verdict whenever they are.
How hours are computed
Usable energy, adjusted for inverter loss exactly once, divided by everything drawing from the battery — including the station itself.
hours = usable Wh × inverter efficiency ÷ (duty-weighted load W + station idle draw W)
The denominator is the part most sizing guides drop. A station that consumes ten watts to keep its own inverter awake is drawing more than a router, and on a small load that term is the difference between a plausible number and a flattering one.
Loss already inside the figure
Loss already inside the figure
Inverter loss still to subtract
Inverter loss still to subtract
Inverter loss still to subtract
Beyond the arithmetic there is a gate. A station may print hours only when its record carries a usable-energy figure that survives scrutiny; today 6 of 29 station records do, so 23 of them show no runtime at all rather than dividing a nameplate capacity by a load. On the load side, 3 profiles are barred from being multiplied out, because their running watts and their duty cycle come from sources that do not describe the same machine.
Where the data comes from
Every stored figure carries a tier, the document it was read from, the date it was read, and a confidence. The library today holds 29 station records (one per model and region, never one sheet for both), 30 load profiles and 5 panel records.
Transcribed from a published sheet, manual or product page, with the URL and the date it was read. Government and standards documents are held to the same tier. Every figure on the site today is this one.
The library holds none, so the site claims none. There is no test bench behind these pages, and nothing here is written as though there were. If that ever changes, the figure arrives dated and labelled, and it says what it disagreed with.
Submissions are not open yet. When they are, a figure waits in review until three independent logs of the same pairing agree, and it stays marked as a reader log afterwards.
Records are also gated before they become pages. A record with a hole where a verdict-bearing number should be does not get a page: 3 station records and 3 load profiles are held back on those grounds right now.
When two sources disagree
Manufacturers contradict themselves more often than a specification sheet suggests: a product page and a manual quoting different maximum inputs, a store page listing an outlet type its own photograph does not show. Cross-checking finds these, and when it does, the number is not quietly averaged and not quietly picked.
An open conflict is shown on every page that touches the field, above anything derived from it. Where it blocks a verdict, no verdict is issued. Displaying the disagreement is the point: hiding it would mean printing one of two numbers and knowing that the other one exists.
Prices
Every price carries the date it was checked. Past 14 days it is visibly degraded rather than silently kept, and the age is printed next to it. The comparison runs on whole calendar days in UTC, so the same page tells a reader in Berlin and a reader in Denver the same thing. Where a merchant's own page is not machine-readable, the price is checked by hand or the figure is simply absent.
What this site will not do
How the money works
Retailer links on this site may earn a commission, at no extra cost to you. They are marked as paid links in the markup, open in a new tab, and go straight to the product page with no interstitial. The engine has no access to which link pays more, and no verdict, ranking or threshold on this page has a commercial input. When the honest answer is that a machine cannot run your load, the page says so and links to one that can — which is worth more to a reader, and, over time, to us.