August 2019 C.E.T. and EWP forecast contests

[damianslaw]

2 hours ago, cheese said:

Rainfall:

Sunshine:

Ties in with what I thought. We finished the month with just under 50mm, compared to a LTA of 58mm. So sunnier than average, drier than average and warmer than average - August actually turned out to be the best month of the summer here despite the pessimism earlier in the month.

As suspected a very wet month over Cumbria, indeed a very marked north west / se split. Despite the warmer than average temps, the near average sunshine and well above average rainfall cancelled out the warmth, meaning August 2019 will go down as another rather poor August here.

[J10]

22 hours ago, J10 said:

Spreadsheets to follow tomorrow.

August 19 CET.xlsx

Spreadsheet as promised.

[Relativistic]

Not displeased with my forecast at all - 17.3C with a hot spell some time after the 20th... and yet I only managed 18th!

[feb1991blizzard]

9 minutes ago, Relativistic said:

Not displeased with my forecast at all - 17.3C with a hot spell some time after the 20th... and yet I only managed 18th!

yes, both the last months went as i thought and got a combined error of 0.3 from both months, yet my 0.1 error this month i managed 12th place!

[feb1991blizzard]

I do agree with MWB that people are getting better to a certain extent, don't know if conscenscous reveals this or not or whether the ridiculous forecasts at the extreme ends skew against that or not, however, i still believe that also because the last few months have been what everyone expects these days (ie - warmer and wetter than average), that is why it has been so competitive at the top of the leaderboard, when we start getting some significantly colder than average months, whether that be this contest or next years, the skill will be who spots them first, that will separate the boys from the men.

[Weather-history]

EWP is 87.3mm

 

[Roger J Smith]

Yes I just noticed that value posted (87.3), it will change the ranks and therefore the shared scores of any tied values in the preliminary scoring table (earlier post) but probably not much will happen as a result of that in the annual scoring. I will work on the updated excel file tomorrow (my time, much later today your time) as I have been out in the hot sun here most of the day and can't stay awake. ... On the point about consensus and wild forecasts near the extremes, the EWP contest uses a median consensus so it doesn't really matter how extreme the highest and lowest few forecasts are, the median is the middle forecast of all those submitted. Average might sometimes be affected. And when I do a CET table of entries, my consensus value is also a median but the spreadsheet from J10 has the average. I think sometimes there is a difference of 0.1 so in most cases it doesn't matter much which type of consensus stat we use. 

If you are unfamiliar with statistics, mean and average are the same thing and are the sum of all values divided by the number of values ... median is derived from the value closest to the middle of the ranked set ... for an odd number of values, it will be the value equal to (n+1) / 2 where n is the number of values (entries in the case of our contest, and that excludes any normals we post). When there is an even number of values, median is the average of two values (entries) on either side of the decimal that results from taking (n+1) / 2 ... as an example, with 48 entries, the median is the average of 24th and 25th ranked entries (as (n+1)/2 is 49/2 or 24.5). For 49 entries the median is the 25th value. (50/2). 

How much can a very high forecast change average more than median? Let's take the usual example of the LG punt for a value 2 deg above an all-time record. That is often 7 degrees above normal. So let's say consensus is one degree above normal from median and all but that one forecast are randomly distributed on either side of the median. With 49 forecasts, the six degree anomaly of one entry will therefore give an average of 6/49 above the median, or 0.12 roughly. So it only makes 0.1 difference to the average compared to the median (if those assumptions are correct -- if there's one oddly low forecast to balance out, then even less of a contrast between mean and median). 

EWP is a bit different in that it's easier to go much higher than much lower compared with normal or consensus values. And a few people do go much higher. So I would imagine that the average EWP forecast is maybe 3-5 mm higher than the median forecast. But as average EWP forecast is not ever utilized in any sense, that doesn't affect the contest. (this is why I chose median and not mean in the first place)

 

[Roger J Smith]

Here is the updated scoring file for the EWP contest. 

EWP20182019AUG.xlsx

[J10]

Sorry to seaside 60 whose entry was accidentally missed.

Seaside 60 came 22nd in the monthly comp and was 0.2c out.

Seaside 60 came 19th in the seasonal comp and 19th overall.

Here is the revised spreadsheet.

August 19 CET revised.xlsx

Edited October 3, 2019 by J10

[feb1991blizzard]

10 minutes ago, J10 said:

Sorry to seaside 60 whose entry was accidentally missed.

Here is the revised spreadsheet.

August 19 CET revised.xlsx 457.77 kB · 1 download

Is the updated annual CET spreadsheet available yet incorporating September's results.