November 2020 C.E.T. and EWP forecast contests -- the last month of this contest year

[Froze were the Days]

Sorry a bit late, 6.4c for me.

Edited November 1, 2020 by Froze were the Days

[Man With Beard]

19 hours ago, mb018538 said:

6.6c and 58mm please 

I think you've got this year's title in the bag now. All your main rivals have posted pretty similar values to yours. 

Must admit I'm a bit nervous about going so low considering the first day of the month could be close to a November CET record, I'd have thought?

[mb018538]

1 hour ago, Man With Beard said:

I think you've got this year's title in the bag now. All your main rivals have posted pretty similar values to yours. 

Must admit I'm a bit nervous about going so low considering the first day of the month could be close to a November CET record, I'd have thought?

I haven’t even looked what everyone else has posted up to be fair! Just done my usual - wait til the end of the month, look at all the data available and go from there. Never entered before so I’m quite surprised to be top! Think I’ve been fairly close all 11 months so far.

[davehsug]

Oh dear, late again! 8.5C & 94 mm pls.

[Roger J Smith]

CET November 1981-2010 averages and 1772-2019 extremes

______________________________________________________

 

Date ____CET mean__CET cum____MAX ___________ MIN ______ CET running mean (extreme values)

 

01 Nov ___ 9.5 ______ 9.5 ______ 13.5 (1894,1982) __1.5 (1934) ___ 13.5 (1894,1982) _1.5 (1934)
02 Nov ___ 9.0 ______ 9.2 ______ 14.3 (1821) ______ 0.5 (1880) ___ 13.8 (1821, 1894)_2,3 (1880)
03 Nov ___ 9.0 ______ 9.2 ______ 15.0 (1996) ______ 1.9 (1820) ___ 13.6 (1894) _____ 2.6 (1880)
04 Nov ___ 8.5 ______ 9.0 ______ 14.2 (1946) _____0.8 (1845&48) _ 13.0 (1894) _____2.6 (1880)
05 Nov ___ 8.3 ______ 8.9 ______ 15.4 (1938) ______ 0.8 (1801) ___ 12.7 (1894) _____ 3.1 (1782, 1880)

06 Nov ___ 8.2 ______ 8.7 ______ 14.3 (2015) ______ 0.6 (1868) ___ 12.2 (1996) _____ 2.7 (1782)
07 Nov ___ 8.3 ______ 8.7 ______ 13.8 (2015) _____ --0.6 (1791) ___ 11.8 (1938) _____ 2.4 (1782)
08 Nov ___ 8.7 ______ 8.7 ______ 13.4 (1852) _____ --1.2 (1812) ___ 11.8 (1938) _____ 2.4 (1782)
09 Nov ___ 8.1 ______ 8.6 ______ 12.5 (1988) _____ --0.5 (1921) ___ 11.8 (1938) _____ 2.3 (1782)
10 Nov ___ 7.9 ______ 8.6 ______ 13.4 (1977) ______ 0.1 (1812) ___ 11.8 (2015) _____ 2.4 (1782)
 
11 Nov ___ 8.3 ______ 8.5 ______ 13.9 (2015) ______ 0.3 (1828) ___ 12.0 (2015) _____ 2.4 (1782)
12 Nov ___ 8.2 ______ 8.5 _____ 13.8 (1938,1947)_ --1.4 (1919) ___ 11.9 (1938,2015) _2.6 (1782)
13 Nov ___ 7.2 ______ 8.4 ______ 14.6 (1938) _____ --1.1 (1862) ___ 12.1 (1938) _____ 2.7 (1782, 1919)
14 Nov ___ 6.8 ______ 8.3 ______ 13.4 (1938) _____ --1.1 (1925) ___ 12.2 (1938) _____ 2.5 (1782, 1919)
15 Nov ___ 6.8 ______ 8.2 ______ 13.5 (1821) _____ --2.1 (1965) ___ 12.0 (1938) _____ 2.3 (1919)

16 Nov ___ 6.8 ______ 8.1 ______ 13.4 (1997) _____ --2.9 (1901) ___ 11.9 (1938) _____ 2.2 (1919)
17 Nov ___ 6.4 ______ 8.0 ______ 13.1 (1997) _____ --2.0 (1841) ___ 11.8 (1938) _____ 2.4 (1919)
18 Nov ___ 6.7 ______ 7.9 ______ 12.2 (1978) _____ --1.6 (1887) ___ 11.6 (1938) _____ 2.6 (1919)
19 Nov ___ 6.4 ______ 7.9 ______ 12.7 (1994) _____ --2.1 (1815) ___ 11,4 (1938) _____ 2.9 (1919)
20 Nov ___ 6.2 ______ 7.8 ______ 12.8 (2009) _____ --1.2 (1829) ___ 11.2 (1938) _____ 2.9 (1786)

21 Nov ___ 5.8 ______ 7.7 ______ 13.7 (1947) _____ --1.4 (1880) ___ 10.8 (1938) _____ 2.9 (1786)
22 Nov ___ 5.8 ______ 7.6 ______ 14.4 (1947) _____ --2.3 (1773) ___ 10.7 (1938) _____ 2.8 (1786)
23 Nov ___ 5.8 ______ 7.5 ______ 13.3 (1947) _____ --4.2 (1858) ___ 10.6 (1994) _____ 2.7 (1782)
24 Nov ___ 5.9 ______ 7.4 ______ 12.7 (1980) _____ --4.6 (1904) ___ 10.6 (1994) _____ 2.6 (1782)
25 Nov ___ 6.3 ______ 7.4 ______ 12.1 (1964) _____ --1.5 (1923) ___ 10.5 (1994) _____ 2.5 (1782)

26 Nov ___ 6.2 ______ 7.3 ______ 12.1 (1983) _____ --1.7 (1904) ___ 10.6 (1994) _____ 2.4 (1782)
27 Nov ___ 6.0 ______ 7.3 ______ 11.3 (1818) _____ --2.6 (1915) ___ 10.6 (1994) _____ 2.3 (1782)
28 Nov ___ 5.8 ______ 7.2 ______ 12.0 (1828) _____ --4.0 (2010) ___ 10.4 (1994) _____ 2.2 (1782)
29 Nov ___ 5.5 ______ 7.2 ______ 12.2 (1818) _____ --2.6 (1801) ___ 10.2 (1994) _____ 2.3 (1782)
30 Nov ___ 5.8 ______ 7.1 ______ 12.4 (2001) _____ --1.7 (1973) ___ 10.1 (1994) _____ 2.3 (1782)

_________________________________________________________________

[West is Best]

Well I think we've just started the month with the warmest ever recorded November night. The highest minimum is always a bit of a strange one to measure but by my reckoning the record has just been smashed. We await the Met Office to confirm.

[mb018538]

3 hours ago, West is Best said:

Well I think we've just started the month with the warmest ever recorded November night. The highest minimum is always a bit of a strange one to measure but by my reckoning the record has just been smashed. We await the Met Office to confirm.

I already think my 6.6c guess is looking like absolute garbage! The models have really pulled away from HLB scenarios in the last couple of days, and through next week there are a lot of days forecast in the low to mid teens, with nights not especially cold. We could get to mid month with the CET up at around 9-10c at this rate!

[feb1991blizzard]

29 minutes ago, mb018538 said:

I already think my 6.6c guess is looking like absolute garbage! The models have really pulled away from HLB scenarios in the last couple of days, and through next week there are a lot of days forecast in the low to mid teens, with nights not especially cold. We could get to mid month with the CET up at around 9-10c at this rate!

You're top of the league - you can still change it - i would do that given what is at stake - you'll get a penalty but the penalty would pale into insignificance compared to having a massively wrong forecast.

[mb018538]

27 minutes ago, feb1991blizzard said:

You're top of the league - you can still change it - i would do that given what is at stake - you'll get a penalty but the penalty would pale into insignificance compared to having a massively wrong forecast.

Didn't even know you could change it! How does that work anyway? Never changed forecasts any so far!

[Summer Sun]

12.9c to the 1st

3.9c above the 61 to 90 average
3.4c above the 81 to 10 average

___________________________

Current high this month 12.9c on the 1st
Current low this month 12.9c on the 1st

[Earthshine]

1 hour ago, Summer Sun said:

12.9c to the 1st

3.9c above the 61 to 90 average
3.4c above the 81 to 10 average

___________________________

Current high this month 12.9c on the 1st
Current low this month 12.9c on the 1st

Toasty!

[feb1991blizzard]

4 hours ago, mb018538 said:

Didn't even know you could change it! How does that work anyway? Never changed forecasts any so far!

You can change it up to 3rd of the month with penalties increasing each day - if you change it tonight you get a 0.2 penalty on your score i think although don't quote me on that and tomorrow would be 0.3 - best to ask Roger or J10 but if you think you're gonna be way off and you want to go much higher then im sure it would be better, i wouldn't do it if your only adding 0.4c onto it or something like that but from your tone you seem to think you have massively screwed up - not so sure looking at those EC clusters though, the bottom ones would give a cold continental flow so its a judgement call.

[mb018538]

4 hours ago, feb1991blizzard said:

You can change it up to 3rd of the month with penalties increasing each day - if you change it tonight you get a 0.2 penalty on your score i think although don't quote me on that and tomorrow would be 0.3 - best to ask Roger or J10 but if you think you're gonna be way off and you want to go much higher then im sure it would be better, i wouldn't do it if your only adding 0.4c onto it or something like that but from your tone you seem to think you have massively screwed up - not so sure looking at those EC clusters though, the bottom ones would give a cold continental flow so its a judgement call.

Thanks for explaining. I’ll just leave it be as I’ve done all year then, and see how we go. If it’s good enough it’s good enough....if not then so be it!

[Roger J Smith]

I sent mb018538 a private message indicating my thoughts about changing but in general please note, J10 makes decisions about CET issues, I just tabulate the entries and then run and score the EWP contests. So basically I don't even know if late edits are acceptable in the CET portion, I have allowed them for EWP with late penalties applied, and I have posted one or two requested CET amendments over the years in the tables of entries (it rarely happens), but then I actually don't know if they were allowed or ignored because J10 makes up his scoring tables and then uses my table of entries as a sort of quality control feature to check accuracy of his own compilation, so he would have come across the original forecasts and then the posts requesting a change but whether he allowed them or not, I honestly don't know, as I never looked into the CET scoring tables to find out. Maybe he will let us know if his policy is to allow them with a late penalty or not to allow them. Once I've tabulated the CET forecasts, I make no further use of the information and I just go to the CET tables for annual rank information if I want to cross-reference them with EWP, I don't look at the actual mechanics of how those CET ranks were derived because I don't need to know. I think I understand the system but my impression is that the late penalties are not as punitive as the EWP 0.2 out of 10 per day. Aren't they something like 0.1 per day, but against what, an annual average? This is something I don't factor in when I report on best combined forecast, for instance, the 10.5 set in October had eight forecasters in it and three of them had late penalties, were they really 6th, 7th and 8th in points for the month or did they fall behind one or two of the non-penalized 10.4 and 10.6 forecasts? In the EWP section, being late one day usually means you will drop one rank lower because the scoring intervals are almost always a bit smaller than the daily penalties. A person who is three days late could be 3-4 ranks lower (higher numbers, I hate rank terminology!). 

In general, you'll find that the CET ranks are a little more lenient towards 10-11 out of 12 contest entrants than EWP which simply adds up all scores, today I did indicate averages for those who hadn't entered all eleven, and I might try to work out a scoring system that gives a bit of a boost to those who have a good average and miss one month. On the other hand, if you miss three CET contests you are relegated to second table no matter what your points total, and in the EWP you cruise along with your points and can be ranked fairly well even if you've only played 7 or 8 out of 11-12. There's probably no perfect way to do this, the perfect solution would be if everyone always entered, which is out of our control. 

[Dancerwithwings]

Me personally...keep it simple...once midnight strikes that’s it, “NO changes” only penalties if ya late .   edit; mb018538 I’m lower in my prediction that you  but like MWB said “you’ve pretty much got it” 

Edited November 2, 2020 by Dancerwithwings

[Roger J Smith]

Problem with that approach might be, if people knew they couldn't change after deadline, more people would wait and come in late just to be sure the models weren't going off on one (remember Jan 2008 anybody?) ... with the late edit available, there's no particular advantage to waiting and taking the small late penalty so we tend to get almost everyone in on time. 

[Roger J Smith]

Just out of interest to provide context, these are the forecasts of the current annual top ten in the CET to illustrate potential for movement in the standings. I show how the order of entry affects possible rank ordering but bear in mind, that also depends on how many people predicted some unknown value that turns out to have equal error value (e.g., result is 7.1, then all the 7.6 forecasts scatter in with the 6.6 forecasts to determine rank). A result between two forecasts will slant the outcome towards equal ranking.

CET contest

1. mb018538 _______ 6.6 

2. Stationary Front _ 6.6 (will be at least one rank higher due to earlier entry, depending on equal errors for ??)

3. Don ______________ 6.7 (will have at least 3 rank advantage or disadvantage due to two others at 6.6)

4. Man with Beard __ 6.3 (could have at least 5 rank advantage or disadvantage rel to mb due to 5 at 6.4, 6.5)

5. dancerwithwings _ 6.4 (could have at least 3 rank advantage or disadvantage rel to mb, one earlier 6.4, two at 6.5)

6. The PIT ___________ 7.2 (a warmer outcome than this would give a 36 rank advantage rel to mb, 34 SF and 28 rel to Don)

_________________________ (note an outcome between 6.7 and 7.2 would be less advantageous or a slight disadvantage)

7. February 1978 ____ 6.5 (only two ranks separated from mb for any outcome, plus any equal error intrusions)

8. sundog ___________ 6.8 (has about 15 ranks advantage over all but PIT if outcome is warmer than 6.8)

9. Mulzy _____________ 6.9 (has a few ranks advantage over sundog but generally will have same outcome rel to others)

10. DR(S)NO _________ 6.0 (for an outcome this cold, has up to 20 ranks advantage rel to mb but only 9 rel to MWB).

______________________________

It looks to me like PIT has some chance of moving well up the table, most of the others are quite clustered and given the complexity of the CET scoring system which factors in a lot of annual average error stats, I can't see much potential for large migrations, DR(S)NO is probably mathematically unable to move past Man with Beard with a cold outcome (so that would favour Man with Beard overall). 

In the EWP contest, where Twilight leads Godber1 by about three points, there is potential for a lead change, as Twilight has 110 mm and Godber1 has 55, almost the entire field between them so their points differential could be 7 or 8, however an outcome in the middle would leave them in same relative positions. Third and fourth place mb018538 and Mulzy have 58 and 62 mm, probably not different enough from Godber1 to pass his total although it would be close for Mulzy. Fifth place Reef can only reach second with his 112 mm (can't score more than 0.4 different from Twilight). Blast from the Past in sixth place is in the same boat at the lower end with same forecast as Godber1. JeffC and Emmett Garland are other top ten EWP forecasters who have high end forecasts similar to Twilight, while Don (90 mm) and Bobd29 (82 mm) could make moves if they are very close to top scores. Nobody outside the top ten in EWP has much hope of moving up past about 8th place, given the scatter of forecasts. 

[Man With Beard]

Not going to lie - I was tempted today to take the penalty and go for high 7s but it's just not the spirit of the competition. I'll stick with my idiotic 6.3C!

I am kicking myself though. Why have I fallen for the trap - again - HLB without a strong trough anomaly over Europe is so prone to collapsing into a Sceuro or East European High!

[Roger J Smith]

So I had a look into the late penalty structure in the CET contest and can see that (a) it is 5 points per day, (b) the total of them over the contest year then participates in the total points structure at the rate of 1/4 of total deducted, so in other words if you had one penalty of two days (10 points) that would take 2.5 points off your annual total which is about two-thirds of the raw total because it is scaled to 500 and highest actual totals would be more like 800.

So that tells me that in any given monthly contest, the effect of the late penalty is almost the same as my method in the EWP, but it is not applied to monthly rankings directly. 

The annual CET ranks are a function of how well you do against the field in several different parameters, average error, average monthly points, total monthly points, total bonus points for accuracy -- but as I mentioned, once you miss a third contest you're out of the table so maintaining a good average alone cannot help you at that point (although it determines some of the outcome for the field). 

Here's a raw example applied to the contest leader mb018538 who had one late day penalty. His weighted average rank (reverse meaning 167th is best in the total field of 167 who ever entered any contest) is 158.75, so his score would be (158.75/167) x 500 which is 475.3. However with the late penalty of 5, applied to the average rank, his score actually is (1.25/167)*500 lower or 3.7 lower, 471.6 instead of 475.3. 

The late penalty therefore is about equivalent to three ranks in the month it occurred, if I apply this math to the simpler format of total score being the sum of monthly scores. 

Somebody who was three days late every month would rack up quite a total points decrease, around (45/167)*500 or about 135 (in the middle of the CET table that would drag you to near the bottom). A three day late penalty every month in the EWP would amount to 12 x 0.6 or 7.2 points over the entire contest. That is around 10 rankings in most portions of the scoring table. So I'm learning that my EWP late penalties are relatively benign, possibly too much so? In another contest I score, late penalties are 1 point (out of 100) every four hours, and they increase sharply after 36h, somebody who was just 36h late every month would have 108 fewer total points at end of year, about the range from first to nearly last in that contest. That appears to be even a bit harsher than our CET penalty scale. However, in that contest, you can enter up to six hours later than month begins so we're actually comparing 42 hours lateness. 

So my conclusion is that to take a 2-3d late penalty might be risky because it could drop your total score by more than you gain from a modest improvement in monthly score given that a lot of your total score is dependent on averages for twelve forecasts so that improvement will only change your rank by a very small amount. 

Complex math, as a game theory sort of thing I would say take 20-25 improved ranking points as about the equivalent of a 3 day late penalty. Would have to plug in actual numbers, all depends on a number of factors such as how many new entrants there are (this dilutes the differential), and how separated you are in points where your rankings can change. 

You may be wondering, wouldn't the contest leader normally have a higher average rank than 158.75 out of 167, but in each category, some of those one-off people are in the mix, so a person who enters once, gets top score, and never returns, has at least one high rank (average monthly points) that the regulars must fall in behind. 

[Roger J Smith]

27 minutes ago, Man With Beard said:

Not going to lie - I was tempted today to take the penalty and go for high 7s but it's just not the spirit of the competition. I'll stick with my idiotic 6.3C!

I am kicking myself though. Why have I fallen for the trap - again - HLB without a strong trough anomaly over Europe is so prone to collapsing into a Sceuro or East European High!

I tend to agree with your philosophy (as a fellow bearded man) but now that I see how punitive the late penalties actually could be in the CET, I wouldn't want to risk it on that basis alone. You could improve your forecast and drop your total score, no point in that. 

[Roger J Smith]

Hey also it is worth noting that NW has apparently fixed that problem where you couldn't easily download excel files, now they open up (at least for me) without any rigamarole. So I will get back to posting the EWP files as a result. 

I worked out all of the above simply by clicking on cells and looking at formulae, then I deleted the late penalty in question, if you want to try that, be forewarned you can't just put the data back as easily as you can erase it. 

[Roger J Smith]

I don't see this posted so ... 

12.4c to the 2nd

3.8c above the 61 to 90 average
3.2c above the 81 to 10 average

___________________________

Current high this month 12.9c on the 1st
Current low this month 12.4c on the 2nd

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

EWP started off with 7 mm on 1st, looks to have doubled that on 2nd (for 14 mm) ...

10d GFS appears to average 25 mm (for 39 mm)

days 11-16 might add 10 mm although very uncertain outlook with systems narrowly missing to north and south (49 mm total)

Probably just about all forecasts are on the table with that much uncertainty. 

CET projection looks fairly mild, would likely be in mid 8s by 19th given this scenario. Even if so, colder forecasts could verify as we found in 2010.

[Summer Sun]

10.8c to the 3rd

2.6c above the 61 to 90 average
1.6c above the 81 to 10 average

___________________________

Current high this month 12.9c on the 1st
Current low this month 10.8c on the 3rd

[Summer Sun]

9.6c to the 4th

1.6c above the 61 to 90 average
0.6c above the 81 to 10 average

___________________________

Current high this month 12.9c on the 1st
Current low this month 9.6c on the 4th

[ItNeverSnowsInWindsor]

Can’t see November finishing below average now with the mild south to southwesterlies being shown in the recent models.