I don't think a surface wind chart has ever caused me so much angst outside of winter cold spells - just look at how the flow takes the longest feasible track across the North Sea before turning to cross the UK!

While there's plenty of dispute regarding the temps modelled by FV3 (most of all), ECM etc - there's no doubt that it won't be nearly as toasty as it could have been for the CET region, even with the western bias to the monitoring stations.

A markedly warmer than average final week was a factor in my estimation for this month, as the tropical cycle looked to be very supportive after factoring in the typical delay of a week or so.

Frustratingly, that cycle has actually played out pretty much as I expected, and the very warm setup has essentially manifested too, but for an unusually strong ridge that could easily have extended less or more east, leading to much less moderation of temps by the North Sea.

You see with the scientific approach, it's certainly possible to get the overall pattern sequence right for the remainder of the month, but the finer detail can still kick you into the gutter anyway.

If one instead uses a mathematical approach, then the best approach for a low average CET errror is actually to routinely estimate close to the LTA unless extreme anomalies are predicted with high confidence (by the NWP models) for the first week or so.

I gather that this mathematical approach is what GW's hoping to deter by withholding the LTA stats until after the deadline.

SHOW EXTERNAL IMAGES

Originally Posted by: Stormchaser

I don't think a surface wind chart has ever caused me so much angst outside of winter cold spells - just look at how the flow takes the longest feasible track across the North Sea before turning to cross the UK!

While there's plenty of dispute regarding the temps modelled by FV3 (most of all), ECM etc - there's no doubt that it won't be nearly as toasty as it could have been for the CET region, even with the western bias to the monitoring stations.

 

A markedly warmer than average final week was a factor in my estimation for this month, as the tropical cycle looked to be very supportive after factoring in the typical delay of a week or so.

Frustratingly, that cycle has actually played out pretty much as I expected, and the very warm setup has essentially manifested too, but for an unusually strong ridge that could easily have extended less or more east, leading to much less moderation of temps by the North Sea.

 

You see with the scientific approach, it's certainly possible to get the overall pattern sequence right for the remainder of the month, but the finer detail can still kick you into the gutter anyway.

If one instead uses a mathematical approach, then the best approach for a low average CET errror is actually to routinely estimate close to the LTA unless extreme anomalies are predicted with high confidence (by the NWP models) for the first week or so.

I gather that this mathematical approach is what GW's hoping to deter by withholding the LTA stats until after the deadline.

I don’t think GW is talking about withholding LTA stats or anything that is already publicly available.  I think it’s his own MO analysis that he’s delaying until after prediction close.  At least that’s how I read it!  

SHOW EXTERNAL IMAGES

Edited by user 24 June 2019 11:09:21(UTC)  | Reason: Not specified

Originally Posted by: Stormchaser

 

So very true Caz 

Whoops, you're right - missed that last sentence in his announcement post!

So the mathematical approach will still be readily achievable. Hmm... 

Not sure whether to give in and go with that, or keep trying to fly the sodden flag of science .

A month is a long time in terms of weather forecasting and as we know only too well, even models struggle to get it right beyond about five days.  So I think sticking your finger in the air to see which way the wind’s blowing is as good a method as any.  Actually, there’s an element of science in that too!   

Originally Posted by: Global Warming

One thing I am toying with for next year is whether to award bonus points (i.e. a reduction in the calculated error) for close predictions when the CET is significantly different to the LTA. Effectively it would work in the opposite way to a penalty for a late prediction or missed prediction.

I don’t want to overcomplicate it. The current system is very simple and has the benefit of just considering actual deviations from the final CET. But as you say there is a potential incentive to just predict something close to the LTA as often that will avoid large errors. 

We should not discourage people from predicting the LTA because sometimes that is the correct answer. But to my mind there is a greater level of skill in correctly predicting the CET when it is say 2C above or below the LTA. So perhaps that should be recognised in the scoring system? If the CET came in close to average in every month of the year then there would be no bonus points. 

If anyone has any thoughts on the above do let me know. We have plenty of time to ponder any potential changes for next year. Happy to stick with the status quo if that is the consensus. 

I’m just as happy with the way it is now though.  I think it works well but if you think it needs to be spread out a bit more, you could ask for all predictions to be to two decimal places.  I’ll go with whatever you decide!  

That idea is one I've sometimes wondered about too GW  - but each time, I've decided not to put it forward because of the inherent complexity that Caz has highlighted very nicely.

Maybe a threshold approach would be manageable though - e.g. if both estimate and observed are differences to LTA (having turned negatives to positives) of >=1.0*C < 1.5*C, bonus reduction of 0.1*C, if both are >=1.5*C < 2.0*C, reduction 0.2*C, if both >=2.0*C < 2.5*C, reduction 0.3*C... and so on.

Depends on how you feel about the extra workings really. I suppose, once implemented in Excel, it would all calculate from same input CET estimates to save effort?

Originally Posted by: Stormchaser

Maybe a threshold approach would be manageable though - e.g. if both estimate and observed are differences to LTA (having turned negatives to positives) of >=1.0*C < 1.5*C, bonus reduction of 0.1*C, if both are >=1.5*C < 2.0*C, reduction 0.2*C, if both >=2.0*C < 2.5*C, reduction 0.3*C... and so on.

That is exactly what I have in mind. Should be relatively simple to implement and manage.

& Nice one Caz 

Originally Posted by: Global Warming

 

That is exactly what I have in mind. Should be relatively simple to implement and manage.

I think you Stormchaser and Caz are all pretty much in agreement on how it could work

From how I read them I think you are proposing something like this as a logical extension to the current rules

Definitions

CET Anomaly = absolute difference between this months CET and the LTA (Using 1961-90 for easy Met Office referral)
CET Estimate = the exact guess of the competitor for this month's CET to up to 2 decimal places expressed as an estimated anomaly
CET = The final value of the Met Office CET (to 1 decimal place as in the "official" version)
CET Error = The absolute value of the difference betwen the CET Estimate and the "Offical" CET as defined above

IF CET Anomaly < 2 then no bonus calcs this month (average month)
IF CET Error >= 1 Then no bonus for this player (not close enough for a bonus)

=Bonus for exceptional months and close guesses

Bonus Adjustment  = -(CET Anomaly-CET Error)/10

This will give a bonus correction in the range of about -0.5 to -0.1 to add to the straight estimating error

If the CET Anomaly is 5C and the estimate is spot on it would be (5-0)/10
If the CET Anomaly is 2C and the estimate is 1C out it would be (2-1)/10

These parameters can easily be tweaked but it is worth noting that negative anomalies over 2C are much more rare than positive ones and this formula does not reflect that fact/

Taking the period Jan 2010 to May 2019 (113 months) there have been 19 positve anomalies of 2C or more but only 3 negative ones

Edited by user 26 June 2019 08:58:43(UTC)  | Reason: Not specified

All this talk of over-complicated systems is doin' my head in!
It's a CET prediction competion. Whoever gets closest to the value wins. There shouldn't be layer upon layer of compexity about it. KISS!

Edited by user 26 June 2019 11:31:51(UTC)  | Reason: Not specified

Originally Posted by: ARTzeman

Met Office Hadley          13.9c.      Anomaly     -0.2c. Provisional to 25th.

Metcheck                       14.02c     Anomaly     -0.13c

Netweather                    14.41c     Anomaly     0.32c.

Mean Of my 10 stations   14.37c.   Difference   -0.96c.

Looking good for my prediction of 14.4C but are we expecting a downgrade this month?

Edited by user 26 June 2019 11:46:40(UTC)  | Reason: Not specified