Page 1 of 1

Long term forecasting 'cracked'

Posted: Wed Apr 02, 2014 10:30 am
by ourarka
No link to the article, as The Times charge for reading it online, but they ran a cover story today that the Met Office have 'cracked' long term forecasting .... particularly for the winter.

They said that previously, computer models fared only slightly better than chance, but that their new modelling software is 63% accurate (but that they hope it will rise to 80% over the coming years). If we assign chance as 50/50, isn't 63% still only slightly better than chance? I just wanted to throw it out there to the guys who knew about this sort of thing as it seemed very dodgy to me. Following that appalling 'research' on 7-a-day fruit/veg (which had more holes than a slice of swiss cheese .... but the papers decided was groundbreaking) is this genuine advances in meteorology, or more dodgy reporting (from a paper that we might expect better)?

Re: Long term forecasting 'cracked'

Posted: Wed Apr 02, 2014 10:37 am
by otorongo
If chance is 50/50, then 63/37 sounds much better than chance!
That's almost 2 out of 3 times right.

Re: Long term forecasting 'cracked'

Posted: Wed Apr 02, 2014 10:45 am
by Adam D
otorongo wrote:If chance is 50/50, then 63/37 sounds much better than chance!
That's almost 2 out of 3 times right.
otorongo is correct, they have essentially doubled their accuracy, which is pretty good going.

Re: Long term forecasting 'cracked'

Posted: Wed Apr 02, 2014 11:44 am
by GoggleboxUK
So they've admitted it's all guesswork but now think they've got software that will make them guess right 13% more effectively?

In essence, they are guessing that their guesses will be better guesses.

I'll stick to my tried and tested method of looking out of the window.

:lol:

Re: Long term forecasting 'cracked'

Posted: Wed Apr 02, 2014 11:56 am
by ourarka
GoggleboxUK wrote:will make them guess right 13% more
They will argue that they have improved by 26% on their previous accuracy.

Otorongo, you are right and if they can get to 80% then they'll really have something to shout about. However, at being right 2 in 3 times, I am not sure that is something that councils, retailers, power companies etc will want to make their plans based upon ........ however it's better than 50/50 .......... oh I give up!

Re: Long term forecasting 'cracked'

Posted: Wed Apr 02, 2014 7:50 pm
by flounder
It's amazing what you can do with seaweed and pine cones these days :wink:

Re: Long term forecasting 'cracked'

Posted: Wed Apr 02, 2014 9:01 pm
by Conifers
flounder wrote:It's amazing what you can do with seaweed and pine cones these days :wink:
Maybe they've got bigger and better pine cones? :lol:
Image

Long term forecasting 'cracked'

Posted: Thu Dec 20, 2018 10:04 am
by rentalcars1211
It's a lofty goal, the code to which hasn't really been cracked. So far, long-range forecasts have been modestly successful.