Hacker News Reader

Comment by petodo | Hacker News Reader

This use infamous extremely unreliable Open Weather data source, so while it's nice there is API and you built website, it's not much of use if data it provides is useless. Form over function.

For instance this was accuracy result for Prague in data collected over 6 weeks in forecasts where sources provided different forecast (how many out of 6 correct): AerisWeather 5.5, Foreca 4, MET Norway 4, AccuWeather 2, Dark Sky 1.5, WeatherBit 1.5, Open Weather 1, World Weather Online 1

And I still keep collecting these data and it's not changing.

View on HN

Hi, as mentioned in the post and on the website, it is using pirateweather.net, not Open Weather (are you referring to openweathermap.org?). If you are curious, the author is transparent about the methodology to achieve the API and what are the shortcomings.

It would be interesting if you could evaluate this properly and submit feedback again. What is your criteria for correctness? Do you share your samples collected? If I understand correctly you also concluded dark sky was unreliable (1.5). This does not match my experience (Canada). In which way?

When I open https://pirateweather.net it says "An Open Weather Forcast API", I think it's pretty clear if you are using "Pirate Weather" you are really using just Open Weather data source.

As for accuracy, it's mostly about false positives/negatives regarding rain, I don't care whether temperature is right to the degree Celsius, but I care if source says it's gonna rain and then it doesn't and vice versa.

Of course your experience may vary, though not sure why you mention Dark Sky accuracy when we talk here Open Weather used by your site.

Beyond the subtitle "Open Weather", it doesn't have anything related to what I believe you are referring to: Open Weather Map (OpenWeather Ltd) https://openweathermap.org/

In the case of pirateweather.net, "Open" is referencing to the fact that the processing and transforms are openly divulged. Read this http://docs.pirateweather.net/en/latest/ it's interesting

I mentioned darksky because you listed it as part of your evaluation and it was not graded good. This means you were not satisfied already with the approach and results in your region at least.

I also care the most about daily and hourly precipitations and how intense they get

So from your linked document it seems they are using these sources:

> Starting from the beginning, three NOAA models are used for the raw forecast data: HRRR, GFS, and the GEFS.

Though I don't understand why they use very very confusing term "An Open Weather Forcast API" including typo (quite unprofessional), which will anyone familiar with weather data sources take as Open Weather (Map).

Hi,

I'm the dev behind the Pirate Weather API. I really appreciate your feedback here, since I missed that typo, but it's fixed now.

I went through a ton of iterations of the subtitle, since it's a tricky thing to get right. I think the "Open" part is important, but agree that I don't want it to be confused with "Open Weather Map". I've retitled it to "A Free, Open, and Documented Forecast API", which is a little wordy, but might be better

Reply on news.ycombinator.com