October 2019 C.E.T. and EWP contests

[Optimus Prime]

1 hour ago, Relativistic said:

 

 

Down as 10.0C in the ranked table

https://www.metoffice.gov.uk/hadobs/hadcet/mly_cet_mean_sort.txt

 

9.95°C (-0.65°C)

Max: 13.3°C (-0.7°C)

Min: 6.6°C (-0.5°C)

[Don]

Just now, Optimus Prime said:

9.95°C (-0.65°C)

Max: 13.3°C (-0.7°C)

Min: 6.6°C (-0.5°C)

We haven’t had a below average month have we?!

[sundog]

19 hours ago, Weather-history said:

November 1919 was a really cold one in Scotland. A mean of 1.1C 

I think the winter of 1919/20 was nothing to write home about though. 

[Weather-history]

3 hours ago, sundog said:

I think the winter of 1919/20 was nothing to write home about though. 

It was pretty mediocre.

[Beet]

So close. I was 0.2 off last September, and now 0.1 off. So, as November is my third entry, I was 0.2 off, then 0.1, so then 0.0 now right? Third time's the charm.

[feb1991blizzard]

28 minutes ago, Beet said:

So close. I was 0.2 off last September, and now 0.1 off. So, as November is my third entry, I was 0.2 off, then 0.1, so then 0.0 now right? Third time's the charm.

same here except i have now had 2 x 0.1 offs, think if there was a competition for the second half of the CALENDAR year, i would be in with a chance of winning it.

[Roger J Smith]

If 10.0 is confirmed, then the top scoring CET forecasts will be these:

1. Brmbrmcar (10.0, entry 28)

2. Roger J Smith (10.0, entry 32)

3. weather-history (10.0, entry 32)

4. Feb1991Blizzard (10.1, entry 23)

5. Walsall Wood Snow (9.9, entry 31)

6. daniel* (9.9, entry 33)

7. damianslaw (10.1, entry 35)

8. Beet (9.9, entry 46)

9. freeze (9.8, entry 24)

10. jonboy (9.8, entry 40)

11. booferking (10.2, entry 54)

12. CheesepuffScott (10.3, entry 2)

13. DiagonalRedLine (9.7, entry 7)

14. Blast from the Past (9.7, entry 14)

15. Prolonged SnowLover (10.4, entry 11)

16. Kentish Man (9.6, entry one day late)

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

(edit for confirmed EWP 148 mm)

For the EWP, will confirm this tomorrow but looks like 145-150 mm for the final total, (it was 148) therefore most likely leader is Godber.1 at 150 mm followed closely by Steve Murr at 140 mm. Virtualsphere appears to be third at 134 mm, Born from the Void (130 mm) fourth and tied on points with SteveB (167 mm) due to a late penalty to BFTV, then we have coldest winter at 123 mm (sixth), three entries at 120 mm and I Remember Atlantic 252 at 188 mm. 

Godber.1 and Steve Murr may then be tied for combined rank totals, as Godber (10.5) finished 21st due to three earlier entries as well as Steve Murr (9.5) who finished 20th in CET. So their ranks both add up to 22.

However two closer CET forecasts had better combined ranks, freeze and booferking had 2nd and 3rd shots at 120 mm behind Blast from the Past, those look like being 8th and 9th overall, for combined ranks of 17 and 20 in the two contests 

I will report on the misadventures of the three robotic forecasters later. 

 

 

 

 

Edited November 2, 2019 by Roger J Smith

[North-Easterly Blast]

It looks as though we have finally recorded a below average month for the first time since March 2018.  October, unlike most other months, is a strange month in that the 1971-2000 average (10.4) is lower than the 1961-90 average (10.6), although in the 1981-2010 average it rose again to 10.7.  At 10.0, October 2019 was below average although not by much, and the coolest since 2012.

[Isle of Harry]

34 minutes ago, North-Easterly Blast said:

It looks as though we have finally recorded a below average month for the first time since March 2018.  October, unlike most other months, is a strange month in that the 1971-2000 average (10.4) is lower than the 1961-90 average (10.6), although in the 1981-2010 average it rose again to 10.7.  At 10.0, October 2019 was below average although not by much, and the coolest since 2012.

I think May this year was slightly below 

[Isle of Harry]

7 hours ago, Relativistic said:

 

 

Down as 10.0C in the ranked table

https://www.metoffice.gov.uk/hadobs/hadcet/mly_cet_mean_sort.txt

 

0.3C cooler than bloody October 1816 - the year with no summer  

[Optimus Prime]

35 minutes ago, North-Easterly Blast said:

It looks as though we have finally recorded a below average month for the first time since March 2018.  October, unlike most other months, is a strange month in that the 1971-2000 average (10.4) is lower than the 1961-90 average (10.6), although in the 1981-2010 average it rose again to 10.7.  At 10.0, October 2019 was below average although not by much, and the coolest since 2012.

May was technically below 61-90 average at 11.1°C (-0.05°C)

[The PIT]

Sunny Sheffield finished on 9.2C -1C below normal. Rainfall 154.9mm 187.1% of the monthly rainfall. Our second wettest since 1955.

[Weather-history]

2 hours ago, Isle of Harry said:

I think May this year was slightly below 

January was certainly below its 1971-2000 average. 

[Roger J Smith]

REPORT ON CONSENSUS and NORMALS SCORING for OCTOBER 2019

 

Dec 2018 _____________________________ Jan 2019 _____________________ Feb 2019 __________

____FORECAST _ error __ rank __ points ____ FCST __ error__ rank _ points _____ FCST __ error __ rank __ points

Consensus_ 5.0 __ -1.9 _ 32 to 32 _ 52.3 _____3.5 ___ -0.5 __16 to 18 _ 75.7 to 78.6 _ 4.0 __ -2.7 _ 33 to 37 _ 48.3 to 54.1

1989-2018*_4.9 __ -2.0 _ 33 to 33 _ 50.8 _____4.7 ___ +0.7 __21 to 23 _ 68.6 to 71.4 _ 4.9 __ -1.8 _ 14 to 14 _ 81.3

1981-2010__4.6 __ -2.3 _ 42 to 44 _ 33.8 to 36.9 _4.4 _ +0.4 __14 to 15 _ 80.0 to 81.4 _ 4.4 __ -2.3 _ 17 to 19 _ 74.2 to 77.0

 

March 2019 _____________________________ April 2019 _____________________ May 2019

____FORECAST _ error __ rank __ points _____FCST_error _ rank __ points _______ FCST _ error _ rank __ points

 consensus _ 6.9 _ --0.9 _ 29 to 33 _ 48.4 to 54.8 _8.4 _ --0.7 _ 24 to 28 _ 55.9 to 62.3 _ 12.3 _ +1.2 _ 28 to 31 _ 50.0-55.0

1989-2018 _ 6.8 __ --1.0 _ 34 to 35 _ 46.2 to 47.8 _8.8 _ --0.3 _ 9 to 13 _ 80.3 to 86.9 __ 12.0 _ +0.9 _ 17 to 20 _ 68.3 to 73.3

1981-2010 _ 6.6 __ --1.2 _ 39 to 39 _ 39.7 ______8.5 _ --0.6 _ 20 to 23 _ 63.9 to 68.6 __ 11.7 _ +0.6 _ 11 to 11 _ 83.3

 

June 2019 __________________________________ July 2019 ______________________ August 2019

____FORECAST __ error __ rank __ points _______ FCST _ error _ rank _ points ________ FCST _ error _ rank _ points

consensus _ 15.0 _ +0.8 _ 27 to 31 _ 48.1 to 55.0 __ 17.0 _ -0.5 _ 19-24 __ 62.9 to 71.0 ____17.0 _ --0.1 __ 3 to 13 __ 81.0 to 96.8

1989-2018 _ 14.6 _ +0.4 _ 11 to 15 _ 75.8 to 82.7 __ 16.9 _ -0.6 _ 25-30 __ 53.2 to 61.3 ____16.5 _ --0.6 __38 to 43 __ 33.6 to 41.5

1981-2010 _ 14.5 _ +0.3 __ 6 to 10 _ 84.4 to 91.3 __ 16.7 _ -0.8 _ 34-39 __ 38.7 to 46.8 ____16.4 _ --0.7 __44 to 44 __ 32.0 to 32.0

____________________________________________________________________________________________

 

September 2019 _____________________________ October 2019 _____________________ Average (11 months)

____FORECAST __ error __ rank __ points _______ FCST _ error _ rank ___ points _______ abs err __ rank __ points

consensus ________ 0.0 _ 01 to 01_ 100.0 ________ 10.9 _ +0.9 _ 29 to 34 _ 45.0 to 53.3 __ 0.94 __ 22 to 25 _ 60.3 to 66.5

1989-2018 ________-0.1 _ 02 to 11 _ 84.6 to 98.5 ___11.0 _ +1.0 _ 35 to 36 _ 41.7 to 43.3 __ 0.85 __ 21 to 25 _ 63.7 to 67.1

1981-2010 ________-0.3 _ 20 to 25 _ 63.0 to 70.7 ___10.7 _ +0.7 _ 24 to 26 _ 58.3 to 61.7 __ 0.93 __ 26 to 28 _ 58.5 to 62.3

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

None of the robots did particularly well in October (green type above) and the contest averages for eleven months only changed slightly, with our consensus performing a bit better than 1981-2010 overall and not quite as well as 1989-2018. That applies more to scoring than absolute error where consensus and 1981-2010 are virtually tied. 

The average error (not absolute) for the best-performing 1989-2018 is +0.3, meaning that predicting 0.3 above that statistic would generate the least robotic error. The improvement is only 0.05 deg on absolute error (to 0.80). That is probably better than most of the field but not the top ten. Taking any other standard approach based on normals would introduce larger errors going in either direction. An average of 1989-2018 and consensus (if you knew that beforehand as you might going at the deadline) would yield an average absolute error of 0.76 degrees. 

Consensus this contest year has been too high (warmer than reality) on three occasions (May, June, October), spot on in September, and too low (cooler than reality) the other seven times. The 1989-2018 average (which played as 1988-2017 in Dec) was too high on four occasions and too low on seven although very close in September. January was the one month that 1989-2018 was different in sign from consensus. 1981-2010 matched 1989-2018 in this regard although never getting closer than +0.3 in June and -0.3 in September. 

The only month where consensus really did better than the recent normals was August, I think any improvements in any other months were so marginal as to be scientifically insignificant. Thus I can give you the good news that on the whole, we have no demonstrated skill (the leaders might, the field does not). And since I use median to derive consensus (which reduces the impact of outliers) it would probably look even worse on us if we used mean unless the outliers are equally preposterous in both directions. 

So how does that compare to EWP forecasting? Over there, consensus is scoring a bit better than the recent normals. We are showing more skill at precip forecasting than temperature forecasting. I suppose one reason for that might be that obvious early events can create insurmountable advantages for rainfall. I would argue that what the field is really forecasting is what they see on the 16-day GFS (with hunchcast projections beyond that limit) so to the extent that the 16-day GFS handles precip reasonably well, the first half totals are already in view of forecasters who then need to add on what they think will come beyond the event horizon. If that known quantity is twice normal, then consensus is very likely to outperform normal. If it looks bone dry, there is also an advantage. Other cases are probably less advantageous for consensus. An event like the late September pattern change can give normals the advantage. 

Edited November 1, 2019 by Roger J Smith

[ItNeverSnowsInWindsor]

4 hours ago, Optimus Prime said:

May was technically below 61-90 average at 11.1°C (-0.05°C)

Was it not below 81-10 as well ?

Edited November 1, 2019 by prolongedSnowLover

[Don]

16 minutes ago, prolongedSnowLover said:

Was it not below 81-10 as well ?

Yes it was but October was also comfortably below the 1961-90 average by 0.6C.

Edited November 1, 2019 by Don

[North-Easterly Blast]

As I said before October is one of the few occasions where the 1971-2000 average is lower than the 1961-90 average.  So anything in the 10.4 to 10.7 range is basically an average October, so we have actually recorded our first below average month of 2019 based on all three sets of averages.

[North-Easterly Blast]

12 hours ago, Weather-history said:

January was certainly below its 1971-2000 average. 

January 2019 was one of those months that was just above the 1961-90 mean but just below the 1971-2000 and 1981-2010 means.

[Don]

1 hour ago, North-Easterly Blast said:

As I said before October is one of the few occasions where the 1971-2000 average is lower than the 1961-90 average.  So anything in the 10.4 to 10.7 range is basically an average October, so we have actually recorded our first below average month of 2019 based on all three sets of averages.

Yes, I actually forgot that October's 1971-00 average was lower than the 1961-90 average.  Nice see a below average month across all three time periods, though, albeit not significantly so.

[summer blizzard]

I go by 1981-2010 so Jan, May, June and Oct fit the bill.

[J10]

Unfortunately my broadband has packed up overnight so unable to send spreadsheet over. 

Monthly figures as per Roger above.

Seasonal 

1 beet 488.2

2 damianslaw 485.5

3 diagonalredline 475.5

Overall 

1 quicksilver1989 477.5

2 stationery front 467.2

3 the pit 464.2

Via mobile phone hopefully full figures should be across by midweek.

 

[Roger J Smith]

EWP preliminary total confirmed at 148 mm. Here's the annual scoring file as posted earlier, with updates for November entries and October best combined forecast -- freeze, 9th CET and 8th EWP for combined low total of 17, then booferking, 11th and 9th for 20, Godber.1 was 21st and 1st for 22, equal to Steve Murr (20, 2). 

I haven't done the math but the November forecasts would need to be further apart than they were from the three leaders for any of them to change positions (similar forecasts can only score a certain amount of points different). This is the current top twenty for annual scoring with their November forecasts shown. 

Rank __ Forecaster _________ Points ____ Nov fcst 

_01 ___ Born from the Void ___ 83.1 ______ 128 mm

_02 ___ Reef ______________ 77.1 ______ 120 mm

_03 ___ J10 _______________ 72.4 ______ 105 mm

_04 ___ weather-history ______68.1 _______ 90 mm 

_05 ___ Don _______________67.9 ______ 135 mm

_06 ___ DR(S)NO ___________67.0 ______ 135 mm

_07 ___ Godber.1 ___________ 64.6 ______ 96 mm

_08 ___ Ed Stone ___________ 64.2 _____ 125 mm

_09 ___ Stargazer ___________63.4 ______ 97 mm

_10 ___ Mulzy ______________62.7 ______ 79 mm

_11 ___ seaside 60 __________62.3 _______96 mm

_12 ___ daniel* _____________61.3 _______96 mm

_13 ___ Blast from the Past ___ 61.2 ______ 140 mm

_14 ___ Stationary Front ______61.0 ______ 124 mm

_15 ___ Midlands Ice Age _____60.0 ______ 102.5 mm

_16 ___ JeffC ______________ 59.8 ______ 105 mm

_17 ___ Feb1991Blizzard _____58.3 ______ 138 mm

_18 ___ virtualsphere ________ 58.1 ______ 134 mm

_19 ___ CheesepuffScott _____ 57.9 _______ 73.9 mm

_20 ___ Kirkcaldy Weather ____56.9 ______ 122.2 mm

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

All annual scoring updates in this file:

 

EWP20182019OCT.xlsx

Edited November 2, 2019 by Roger J Smith

[feb1991blizzard]

Another drop in EWP for me - a feeble 17th now - very poor from me - cannot see it improving next month either - was hoping for a top 10 finish in both, its not going to happen in either now, although a very decent last few months CET has hauled me up from 34 to now only just outside the top 10 i would estimate based on a 4th place in October.

[Roger J Smith]

You could move up in the EWP with any outcome near or above your 138 mm, looking at score differentials and quite a few ahead of you have gone below 100 mm. There could be a five point gain in that scenario. 

Scores are fairly close from 7th place down to 22nd really, anyone in that bracket could move quite a few ranks. The top three however are close to locked in, I would say the top two are definitely guaranteed those spots as there is nobody with any mathematical chance of passing them ... weather-history has some chance to move past J10 if the outcome is 90 mm or less. Don and DR(S)NO are locked into their relative positions with the same forecast, but they could both move up one or two spots if they hit. 

[Walsall Wood Snow]

At 10.0c, October 2019 shares the same mean CET with October's:

1659

1660

1663

1668

1669

1670

1702

1709

1722

1753

Obviously pre 1699 as well as a lot of pre 1723 can't be relied on for accuracy been as many of those months were rounded to the nearest .0 or .5

 

 

 

 

Edited November 2, 2019 by Walsall Wood Snow