January 2018 C.E.T. forecasts (with optional precip contest)

[NorthYorksWeather]

Finished 4.1c here, 0.6c above the 1981-2010 average.

Rainfall was 56.9mm which is just below average. (90%)

[Summer Sun]

Jan finished 1.5c above average

https://www.metoffice.gov.uk/hadobs/hadcet/cet_info_mean.html

[Stargazer]

First time I've been spot on for a long time . Many thanks to the organisers for all the hard work.

Reading University, our nearest official station, ended up at 6.1 for the month, 1.3°C above the 1981-2010 mean. Rainfall at 57mm was 94% of average - it'll be interesting to see how the national average turns out.

[J10]

January results

Monthly -> As mentioned above 2 entrants got the 5.3c figure spot on.

Stargazer and Polar Gael

Overall / Seasonal -> There is still no difference between the Overall and Seasonal competitions. Steve B remains in the lead after being 0.1c this month, and being spot on last month.

Stargazer up to 2nd (from 12th ) after getting it spot on.
pegg24          up to 3rd (from 13th) with
Polar Gael up to 7th (from 33^rd) after getting it spot on.

I know there are issues with these files, but it you download them and rename to XLSX and PDF respectively they should work.

Jan 2018 CET.xlsx

Jan 2018 CET.pdf

 

Edited February 3, 2018 by J10

[Roger J Smith]

Here's an update on how the robots are doing in the contest ...

"Forecaster" __ Dec Error ___ Dec Points (range) ___ rank (range)

1987-2016 ____ +0.2 ________ 89.4 - 93.4 _________ 6 to 9

1981-2010 ____ --0.2 ________ 89.4 - 93.4 _________ 6 to 9

Consensus ____--0.6 ________ 71.0 - 74.9 ________ 20 to 23

___________________________________________________ Contest year averages to date _____

"Forecaster" _Jan Error_Jan Points (range)_rank (range) // mean error_mean points_mean rank

1988-2017 ___ --0.6 _____ 66.7 - 74.1 ____15 to 19 _____ 0.4 ____ 78.1 - 83.8 __ 11 to 14

1981-2010 ___ --0.9 _____ 50.0 - 55.6 ____ 25 to 28 ____ 0.55 ____69.7 - 74.5 __ 16 to 19

Consensus ___--0.9 _____ 50.0 - 55.6 ____ 25 to 28 ____ 0.75 ____60.5 - 65.3 __ 23 to 26

Edited February 3, 2018 by Roger J Smith

[Roger J Smith]

I would love to be able to report the results of the January EWP contest but as of a few minutes ago, no updates posted on the website. I have the tables more or less ready based on 125 mm, if that's close I won't need to do much editing because only a handful of guesses were higher so the order would remain similar. So I will continue to look in on the website every few hours and post these results very soon after discovering the actual value, bearing in mind that my sleep period is a large part of the U.K. daylight hours when people are most likely to be on, so I may be the last to know. 

The January winner would be Jeff C at any value 121.6 to 134.4 mm. Above that range, Virtualsphere at 148.0 mm will be hoping my estimate is low as there are 13 lower forecasts closer to my estimate than that. If the actual value comes in lower, there are forecasts perched every 1-2 mm below Jeff C waiting to take the top scores. I also have an annual update to offer and in that, ChrisBell-not-the-weatherman and Reef look to be the most likely leaders. 

Let's hope we don't have to wait too long to get these results. 

[Roger J Smith]

The daily EWP data have been posted but the monthly totals are not confirmed. I added up the dailies to 103.6 mm which was a bit lower than I had estimated (Ireland generally speaking came in around 150% of normal, this would be more like 110%). Anyway, if that total is confirmed, Syed2878 and Born from the Void (105 mm) would be closest and Polar Gael (102 mm) just barely further out; both BFTV and PG have late penalties so the top score of 10.0 would go to Syed2878. Will see if they post a monthly total soon, surely they can add it up faster than me!

This changes my estimates on annual leaders, the top two now would appear to be Dr(S)No and vizzy2004. 

Tables will be posted as soon as I can confirm the monthly total. 

[Roger J Smith]

I checked again in the last few minutes, still no updates on EWP monthly totals.

The daily amounts actually add up to 104.6 mm but I'm not sure if they add up the 2-digit values and round off, or add up the rounded-off values, which could change the total, anything 104.5 or less would change the rankings slightly, 104.7 to about 105.1 probably won't (or perhaps they will just come in with something entirely different after doing a quality control check). But for now, I have posted the scoring based on 104.6 mm partly because I already lost this table once due to a power failure here (the previous reply content was sadly the item before and not this set of tables, monthly and annual scoring). 

If I have to amend this, so be it ... but for now the results are posted over in the contest thread. 

https://www.netweather.tv/forum/topic/75427-cet-competition-the-scores-ongoing/?page=8

 

[Stargazer]

Looks we have a final figure of 103.9mm.

https://www.metoffice.gov.uk/pub/data/weather/uk/climate/datasets/Rainfall/date/England_and_Wales.txt

Edited February 10, 2018 by Stargazer

[Roger J Smith]

Okay, I wasn't aware this data set existed. It is not quite the same (for Dec 2017 either) as the set I had found which is linked below:

http://hadobs.metoffice.com/hadukp/data/monthly/HadEWP_monthly_qc.txt

Note that the UK met office value for Dec 2017 is 108.8 mm and the value in the other source (Hadley EWP) is 114.2 mm. Meanwhile I also notice that the value of 114.2 has changed since I scored December, from 113.7 mm. Apparently there is ongoing data adjustment after the fact.

I guess we need a policy decision on how to score this contest if they are going to have two different publications with two different values, either one of which could change later. For the time being, my policy will be to take the first published value in the source I used last month, and that source has not yet confirmed the totals although they have published daily data that add up to 104.6 mm (still do as of today). If I add up the rounded off amounts it comes to 104.4.

I'll take this under advisement as they say, if people want to go with the direct met office stats and toss the Hadley stats, then I will rescore December and January -- probably will go this route, but wondering if anyone more familiar with the situation can tell me if these met office stats are likely to change in coming days or once published are they set? In other words, was this December value posted as such first time out, or has it been changed? 

Edited February 10, 2018 by Roger J Smith

[Roger J Smith]

Also, since we use that source (the one I linked) for our CET contest, it seems more consistent to use it for this precip contest, on the other hand, if they are often this slow to update and they change their numbers later, I might be more inclined to go with the other source. Will make a decision by Monday based partly on what others have to say about it.

I thought this was going to be more straight-forward since I've been using the CEP values for ongoing research and they seem to publish those around four or five days into most months. But here it is the tenth and no monthly totals (on the Hadley site). 

If I changed the source, then I would also change the scoring procedure comparing ranks, as the data set goes back only to 1910 and a narrower filter would be required, so perhaps just 10.0 minus 0.1 per 5 ranks different for top score, or forget about the precision factor and go with 10.0 no matter how far off the best forecast happens to be. 

[Roger J Smith]

So I have revised the scoring tables to reflect the values in the link provided by Stargazer, and that's what we will use for this contest from now on ... but I will continue to use the Hadley link for extreme values in the announcement material.

This has led to slight revisions of both December and January scoring for the EWP contest, the new scoring can be seen over in the contest scoring thread. It has made only slight changes for the annual scoring tables.

Sorry about this confusion, but I will now be able to score these contests faster, I think, as the Met Office link seems to update quite soon after the end of each month.