How we turn 20 years of weather data into a match score

Every match score on this site comes from about two decades of historical climate records, graded against the preferences you set. This guide explains exactly where the data comes from, how the four components are weighted, and why the result is a long-run average rather than a forecast for your dates.

Where the underlying data comes from

Every figure used to score a destination is drawn from the free Open-Meteo Historical Weather API, which serves the ECMWF ERA5 reanalysis. ERA5 is a respected scientific dataset that reconstructs past weather worldwide by combining recorded observations with a consistent physical model, producing a continuous gridded record rather than the patchy coverage of individual weather stations. For each destination the site reads daily values at that place's exact coordinates, so the numbers describe the location you actually plan to visit and not a distant airport or regional capital.

The window is deliberately long. When you check a destination or open the year calendar, the tool requests roughly twenty years of daily records: the span ends with the last full calendar year and reaches back nineteen further years before that, so it always works from complete years rather than a partial recent one. Using two decades smooths out freak hot summers and unusually wet springs, leaving a stable picture of what the climate is typically like rather than what one memorable year happened to do.

Which measurements we read each day

For every day inside the period you are interested in, the tool collects six daily variables from the archive: maximum, minimum and mean air temperature, total precipitation, sunshine duration, and maximum wind speed. Sunshine duration is reported in seconds by the archive and converted to hours before anything is scored. These six values are the raw material for everything the site shows, from the headline match number down to the temperature, rain, sunshine and wind breakdown beneath it.

The tool does not score every day of the year. It keeps only the calendar days that fall inside the date window you chose, for each of the roughly twenty years in the span, and then averages each variable across all those matching days. A check covering a single week therefore reflects that week across two decades, not the whole year, which is what makes the result specific to the timing of your trip rather than a generic annual summary.

How the four components become one score

Each averaged measurement is graded against the weather preferences you saved: your ideal temperature range, the most wet days per week you will tolerate, the least sunshine you want per day, and the most wind you will accept. A component earns full marks while the long-run average stays inside the limits you set, and it is reduced progressively the further the average drifts beyond them. This means a destination is never judged against an absolute idea of good weather, only against the conditions you personally said you wanted.

The four component scores are then combined into a single number from 0 to 100, and they are not weighted equally. Temperature carries the largest weight at 35%. Rain and sunshine sit at an equal middle weight of 25% each, with neither ranking above the other. Wind carries the least weight at 15%. In the code this is a single rounded calculation, temperature times 0.35 plus rain times 0.25 plus sunshine times 0.25 plus wind times 0.15, so a place can lose a great deal of ground on temperature while still scoring respectably if sun, rain and wind all suit you.

Quick scoring versus a live historical check

There are two paths through the same scoring formula, and the difference is only in where the averages come from. The Discover tool's default Quick mode does not call any weather service while you wait. It reads a precomputed file of monthly climate averages for each destination, holding the mean temperature, rain days, sunshine hours and wind speed for every month, and grades those stored figures against your preferences. This is what makes ranking every destination feel instant, and it is well suited to drawing up a shortlist.

The destination check, the year calendar and Discover's optional Precise mode take the slower, exact path instead. They fetch roughly twenty years of daily records live from the historical archive, filter to the days that match your window, and average them on the spot before scoring. The weights and the grading rules are identical to Quick mode, so the two agree closely; the live path simply works from the full daily record for your exact dates rather than a cached monthly summary. Use Quick mode to narrow the field, then run a precise check on the two or three places you actually care about.

Why this is an average, not a forecast

It matters more than anything else on this point: every score the site produces is a long-run climate average, not a prediction for your specific dates. It tells you what the weather is typically like for a place at that time of year across the past two decades, which is genuinely reliable when you are planning months ahead and no real forecast yet exists. It cannot tell you whether your particular week will be sunny, wet or windy, because no historical average can do that.

Read the score accordingly. A high number means the typical climate for your dates sits close to the conditions you asked for, and a low one means it usually sits far from them, not that a place is bad. The most useful response to a poor score is often to shift the dates rather than abandon the destination, because timing tends to move the result more than the location does. Treat the score as honest expectation-setting for the planning stage, and still check a normal short-range forecast in the final week before you travel.

Key takeaways

Related