October 2018 C.E.T. forecasts and optional EWP contest

[Midlands Ice Age]

23 hours ago, Roger J Smith said:

Relativistic also had 10.6 C together with Midlands Ice Age (by order of entry, MIA is first and Relativistic is second). 

Six shared an error of 0.1, so it depends on order of entry as to who finishes third to eighth. 

J10 will confirm the actual results, this is how it appears in the table of entries:

 

10.7 __ 3rd -- Summer Blizzard (1), 6th -- Summer Sun (34), 8th -- metaltron (50) __  average 1981-2010

10.6 __ 1st -- Midlands Ice Age (17), 2nd -- Relativistic (22)

10.5 __ 4th -- Dr(S)No (25), 5th -- Dancerwithwings (29), 7th --The Pit (36)

I expect the EWP to finalize around 70 mm. These are the forecasts closest to 70 mm ...

 

76.0 __ vizzy2004

 75.0 __ Midlands Ice Age, DAVID SNOW (+1d late)

 70.0 __ Pegg24, Feb1991blizzard, The Pit, Mulzy

 69.6 __ Kirkcaldy Weather

 68.0 __ TJM14, Don

 65.0 __ weather-history, jonboy (+1d late)

 64.0 __ SteveB

 63.0 __ Radiating Dendrite

Guys...

I think I have found the answer to this competition!!!

I would like to share It with you all..

 

I have only been anywhere near on 2 occasions -

1) February. I entered my 'guess' about a week early as I was on holiday for the last week of the month in Marrakech.

My guesses for both rainfall and  CET were almost spot on (1st and 2nd (beaten only on Hadley rainfall)).

2) October. Again I was on holiday  for the last week and took a complete guess (Italian Lakes), before I left..

Result - 1st in the CET, and in with a chance on the rainfall 

I normally take a long look on the last day of the month and try to apply some rationality to my choices (bearing in mind my cold bent!).         Results - Horrible.

 

There must be a name for  going with 'the first figure you can think of'!

'Jammy devil' might be a good name for it! 

MIA

PS I made the fatal mistake (again) when I entered my guess for this month.

.. 

Edited November 2, 2018 by Midlands Ice Age

[feb1991blizzard]

Anyone know the results of the PPN contest yet?  -  also, i never knew you had preference for early entries, i knew you got penalised for late ones for obvious reasons, so how much advantage do you get by entering on stupid dates like the 19th of the previous month? and also is it 0.2 you have to be within to score in the CET?

[Roger J Smith]

I am working on the first version of the EWP contest results, as NCIC have posted 76.7 mm. That seems a little higher than Hadley will probably be saying (closer to 70, my guess from their tracking). I can post the six highest scores and the updated annual totals, working my way down the annual totals into the seldom enter range so will be posting soon ...

__ OCTOBER EWP SCORES __ NCIC __ 76.7 mm

01 _ VIZZY2004 _________ 76.0 mm __ 10.0

02 _ SUMMER of 95 ______ 78.0 mm ___ 9.8

03 _ MIDLANDS ICE AGE _ 75.0 mm ___ 9.6

04 _ DAVID SNOW _______ 75.0 mm ___ 9.4 (0.2 pen)

T05_ LEO97t, BlastFTPast _ 80.0 mm ___ 9.1 each

then the top five in the annual contest are

SINGULARITY (79.1), J10 (69.7), tie- DON, MIDLANDS ICE AGE (69.3), JONBOY (68.5)

after which it's consensus followed by Polar Gael, Dr(S)No, Radiating Dendrite, Norrance and DAVID SNOW to round out the top ten.

===========================

To answer in part the questions above this post, the CET scoring runs from 100.0 for top forecast to 0.0 for low (highest error), with all other scores worked out from intervals based on numbers of entrants (this past month it was 66, so 65 intervals required between top and bottom. For the EWP forecast contest, the scoring is similar but runs from 10.0 to 0.0 and the only real difference is that I don't give different scores based on order of entry to tied forecasts, all those get the same number of points.

In the CET, tied forecasts (by absolute error) are given descending numbers of points by order of entry. It doesn't matter what date the early entries take place, it's just the order within the tied group. So this past month, six were tied for 0.1 error after two received top scores for 0.0 error, so those six were ranked third to eighth and at about 1.6 points per drop, that's a range of about 8 points (in rough numbers, from 89 to 97 out of 100). Your timing is based on the last revision too, so there's not much point in revising a really early forecast by 0.1 deg although I've just seen that done (why, I'm not sure). 

The late penalty in the EWP is 0.2 per day. In the CET contest the late penalties are a bit more complicated. J10 has a scoring system that only he and NASA actually understand, I tried, but came up a bit short, although it rarely strays much from the order of average forecast error or points per month. There are points for accuracy, then it gets into rank orders, and so on, until there's a list of added up numbers that are then "normalized" so that top score in that list is against 500 (for perfect forecasting). You might want to look into the scoring file that J10 posted above, net-weather is no longer making these files difficult to open as before, they open up right away and you can see how the scoring is done. I couldn't really tell you how much impact the 10 points late per day actually has because I never really tried to calculate what percentage it is of the scores it is reducing, but I don't think it's much different from my horse-and-buggy 0.2 points per day. 

So on another topic, we had two more close calls on the "monthly double" using these NCIC figures, MIA was first in CET and third in EWP, DAVID SNOW had top ten finishes in both. 

We have had quite a few monthly doubles at the futile end of the spectrum (but that's planned, I believe). Ask more questions about scoring if you wish, we have lots of time until the next month gets scored. 

 

 

Edited November 2, 2018 by Roger J Smith

[feb1991blizzard]

3 minutes ago, Roger J Smith said:

 

To answer in part the questions above this post, the CET scoring runs from 100.0 for top forecast to 0.0 for low (highest error), with all other scores worked out from intervals based on numbers of entrants (this past month it was 66, so 65 intervals required between top and bottom. For the EWP forecast contest, the scoring is similar but runs from 10.0 to 0.0 and the only real difference is that I don't give different scores based on order of entry to tied forecasts, all those get the same number of points. The late penalty in the EWP is 0.2 per day. In the CET contest the late penalties are a bit more complicated. J10 has a scoring system that only he and NASA actually understand, I tried, but came up a bit short, although it rarely strays much from the order of average forecast error or points per month. There are points for accuracy, then it gets into rank orders, and so on, until there's a list of added up numbers that are then "normalized" so that top score in that list is against 500 (for perfect forecasting). You might want to look into the scoring file that J10 posted above, net-weather is no longer making these files difficult to open as before, they open up right away and you can see how the scoring is done. I couldn't really tell you how much impact the 10 points late per day actually has because I never really tried to calculate what percentage it is of the scores it is reducing, but I don't think it's much different from my horse-and-buggy 0.2 points per day. 

So on another topic, we had two more close calls on the "monthly double" using these NCIC figures, MIA was first in CET and third in EWP, DAVID SNOW had top ten finishes in both. 

We have had quite a few monthly doubles at the futile end of the spectrum (but that's planned, I believe). Ask more questions about scoring if you wish, we have lots of time until the next month gets scored. 

 

 

Thanks very much Roger, Just one more question but no rush to answer it as i know you and J10 have your hands full with organising all this, does the fact that i dont enter some months (summer months mainly) mean that i would be disqualified from the annual competition?  either way at least with the seasonal one i have a chance - the winter one i am motivated to win!

[Roger J Smith]

In the CET, you are only in the main table if you miss zero, one or two forecasts. Then the people who have entered fewer are ranked below that in a second section, against those with similar numbers of entries. 

In the EWP, your total score is your total score and that's your ranking. I don't deal in average scores partly because the months cannot be compared as easily, in the CET a part of the scoring is absolute error, in EWP it's solely based on rankings within each month. 

I am probably going to modify the EWP scoring to give some credit for being first in with a forecast amount, thinking maybe I will drop the scores by just 0.1 but preserve the larger drop that would take place for the next level down. Also I am looking at some way of normalizing the scores so that actual error has something to do with the magnitudes. This was an experiment to see what participation might be like, and it has been encouraging in that two-thirds (roughly) of the regulars have chosen to support the EWP contest too. Thanks everybody, hope you're enjoying the expanded forecasting nightmare I mean experience. 

==========================================

EWP results are posted now (see previous post for summary)

https://www.netweather.tv/forum/topic/75427-cet-competition-the-scores-ongoing/?page=10

 

 

[J10]

Looking at the queries above - Every point deduction in the overall comp is pretty much the final points deduction after adjustments, this is by accident rather than by design. However this remains a pretty severe penalty.

In terms of advantages of early entry, this is generally pretty small especially in month with less entries, although it could accumulate if you enter very early each month.  

All players are ranked in the Overall competition, but if you miss 3 entries or more you cannot finish higher than anyone who misses only 2 entries.
As an example, Optimus Prime is the highest ranked player with 3 missed entries is 54th.

Thereafter the first tie breaker is the number of entire, so someone with 7 entries will be raked higher than someone with 6 entries, regardless of points scored.

[Roger J Smith]

The Hadley chart for October now showing 74 mm for the month. Final numbers are not yet available in the Hadley tables.

(NCIC version 76.7 mm EWP has already been scored, see previous post for link to scoring report in the competition thread).

Looks like I may have minimal editing to do in order to create the Hadley scores, when this is finalized.

[damianslaw]

After the very warm April-July period, a cooling trend to near average has become established it may seem - an impressive nosedive in the CET this October - a common feature of months this year..

Whether we see a similiar fate occur this November - who knows, but there are signals a possible marked downturn may once again occur this month.

 

[Weather-history]

Hadley UKP has October as 75.5mm

5th consecutive time that October was drier than the previous August.

[Roger J Smith]

Hadley as mentioned above came in at 75.5 mm.

Scoring has been posted for this version of EWP.

https://www.netweather.tv/forum/topic/75427-cet-competition-the-scores-ongoing/?page=10

The order of scoring for October is only slightly different, in this version vizzy2004 and Midlands Ice Age are tied for top score, so that qualifies as the elusive "double" for Midlands Ice Age. 

Singularity leads the annual scoring in this version also. 

All other details can be found at the link.