Jump to content
  • Member Statistics

    17,502
    Total Members
    7,904
    Most Online
    Weathernoob335
    Newest Member
    Weathernoob335
    Joined

January 2018 temperature forecast contest -- "Regional Rumble" begins


Recommended Posts

The 2018 version of the long-standing forum temperature forecast contest will go under the title of "Regional rumble" and this will create a new dynamic for the contest, operating in sync with the individual contest we have always had, as best score from each regional forum will count towards a Regional Championship throughout the year. 

To facilitate this, I have posted "portal" threads in regional forums where those members who may never venture into this general interest forum can enter directly. 

For the regulars here, please identify the forum that you are playing for -- I will take central/western even though I probably post more often in GL/OV and MA. If any regional forum wanted to have two teams representing different geographic divisions, that could be accommodated but I am not proactively going to arrange that (you'll need to inform me about it and who is on which squad). 

So the rules remain almost the same, only the paradigm of competition has been amended.

I will relax the late penalties slightly for 2018. And in particular, the late penalties for January won't start until the end of Wednesday Jan 3rd (in PST that would be 4th:08z), after which they will run at 2% per hour to 10z Jan 6th.

Otherwise, here's a summary of the old and the new late penalties. I can't go any lower than this without giving an advantage to late entries.

(note: the deadline will be the same, 06z of the 1st day -- starting with February)

hours late ______ 2017 penalty ____ 2018 penalty

first 36h ________ 1% per 2h _____ 1% per 4h

next 82h ________ 1% per 1h _____ 1% per h for 91h

when 100% reached __ 6th 04z ____ 6th 13z

________________________________________________________________________________

Predictions will continue to be temperature anomalies (F deg) relative to 1981-2010 averages, for these nine locations:

DCA __ NYC __ BOS ___ ORD __ ATL __ IAH ____ DEN __ PHX __ SEA

________________________________________________________________________________

Scoring rules (these are not different but gathered from various earlier posts and codified for smoother application in unusual situations) ...

1. In a "normal" month with the outcome between -5 and +5 F anomalies, and at least one forecast scoring 60 points, all forecasts receive a score based on 100 minus 2 per 0.1 error. The one forecast scoring above 60 in a poor set of forecasts will be rewarded more than cases where all scores are low (see 3).

2. In an "abnormal" month with the outcome either lower than -5 or higher than +5, scoring errors are relaxed as follows: between -5/+5 and any forecast more extreme, only one point is deducted per 0.1 error. That 0.1 grade is matched between zero and correct sign forecasts that are equally removed from zero as actual is removed from -5/+5. Example, if the outcome is +6.8, then reduced error zones are 0.0 to +1.8 and +5.0 to +6.8 and in addition an equal zone above the actual which would be +6.8 to +8.6. In this system, two-point reductions still occur from +1.8 to +5.0. When the actual reaches +10.0, all forecasts with correct sign receive one point per 0.1 deg predicted (e.g. +7.0 is scored 70). In cases with even larger anomalies, this system adjusts by proportion, e.g., for +12.0, a forecast of +9.0 is scored (9/12) of 100 points or 75.

3. In a "poor scoring" month where no forecast receives 60 points using these standard systems, a boosted "minimum progression" of scores is substituted, in rank order, 60, 55, 50, 45, 40, 35, 30, 27, 24, and this turns into 2 point differentials where the number of forecasts requires it to reach zero for low score. One slight difference from previous years is that there won't be a second lower set of boosted scores for cases with no score over 40, no matter how bleak the actual scoring, we will go with this max of 60 each month. And if anyone scores higher by systems 1 or 2 than they would get from this minimum progression, they will earn that score instead (this rarely happens but it can in theory). Remember, with just one raw score of 60 or higher, this all goes out the window and everyone else is stuck with their poor score. That will often mean considerably more zero scores than if this system were applied. 

Please note, the late penalties are still applied to these minimum progression scores, if the 40 score has a 5% late penalty, then it changes to 38 and this does not affect any other scores. 

One new feature will be tracking your average error per station. That will be revealed in a separate table perhaps at the end of each three month period. If every month was scored by rule 1 above, then the average error would be simple to calculate, it would be (100 minus your average score) divided by 20. If you average 70 as a score, your average error is 1.5 deg. But with some cases scored with smaller error points, that may not apply, so to give you the absolute feedback on how you're doing, I will provide those averages (and we'll see if that runs exactly the same as the scoring order or not). 

I hope you'll continue to support the temperature forecast contest and spread the word in your regional forums. But look for threads there by about the 28th. I will count entries posted in those threads and bring the forecasts over here for our tables of forecasts etc. 

The 2017 annual contest has become very, very close and depends almost literally on how cold it gets on Dec 30-31 at various locations, check the December thread for developments. 

 

Link to comment
Share on other sites

That's good, in general terms I am hoping this new paradigm will revitalize this contest, and also make it more accessible.

This is the proposed structure for the Regional Rumble portion, but if there's any discussion I might amend this (by Jan 10th let's say, we should have the rules set before the first month of scoring). The individual scoring will remain the same.

Regions will participate by top scores of their reps in three groups, eastern three, central three and western three. I will retain the concept of the original six in scoring but let's say for NY forum, Don has high score eastern, RJay central and dmillz25 western, then those will be the scores for the Regional Rumble. The alternatives would be either highest scores for each location or the scores of the highest total ("all nine") contestant. I prefer this three-step approach to either of those. What do you think?

Whichever model we adopt, late penalties would apply before selection. 

I hope this will bring in a larger number of forecasts although if the current steady 14 to 16 stayed active, that's enough of a contest to maintain it. But it may be difficult to get responses from some forums, for example w PA / upstate NY, a smaller regional forum and you don't see those folks very often outside their clubhouse. 

The contest used to have more participants in the Midwest than we have seen recently too.

Well, I am hopeful that this will lead to a larger field of entrants, and some fun for inter-forum friendly rivalry. 

Link to comment
Share on other sites

  • Rjay pinned this topic

--2.0 _ --3.5 _ --5.0 ___ --3.5 _ --4.0 _ --1.0 ___ +2.2 _ +3.7 _ +1.8

Central/western forum

will edit in another forecast from that region, it's posted on the regional portal thread, at the deadline which is 18z Jan 2nd.

Anyone above or posting after this and before that deadline, feel free to edit, I won't collect any forecasts until later than the deadline. 

Good luck everyone. 

Link to comment
Share on other sites

Small group....

7 hours ago, Roger Smith said:

--2.0 _ --3.5 _ --5.0 ___ --3.5 _ --4.0 _ --1.0 ___ +2.2 _ +3.7 _ +1.8

Central/western forum

will edit in another forecast from that region, it's posted on the regional portal thread, at the deadline which is 18z Jan 2nd.

Anyone above or posting after this and before that deadline, feel free to edit, I won't collect any forecasts until later than the deadline. 

Good luck everyone. 

Keep trying to get peoole to join.  

@stormtracker, @BxEngine, @mappy, @H2O@jburns, @buckeyefan1...etc etc

 

Stop being chickens!  

Link to comment
Share on other sites

Just now, so_whats_happening said:

DCA -3.0

NYC -2.6

BOS -2.8

ORD -3.7

ATL -2.4

IAH -1.6

DEN +1.1

PHX +1.7

SEA -0.8

 

Almost forgot about this. Thanks again for doing these breaks up the same ole forum crap. I guess I would put myself in the central PA area I frequent other regionals but that here nor there.

Glad you're in!

Link to comment
Share on other sites

There is a later than usual deadline, 18z Tuesday 2nd, so one or two may be waiting to check the charts in the morning. Anyone is free to edit their posts before I collect forecasts for the table in the evening hours tomorrow (most likely, could be on the 3rd). I have one other forecast from a member in the central/western forum, will put that in the table too (it's in the entry portal thread in that subforum now). 

Yeah, would be good to expand this back to the scale of a few years back, maybe 30-40 regular entrants. I've seen a pattern where people try their hand at it for 2-3 months and either get discouraged (not sure why, sometimes they have done really well in those months), or perhaps they lose interest in the forum after the winter or summer depending on why they joined. 

At the moment, it's going to be a real challenge for the rest of us going up against a rather loaded NYC line-up, I think they were a large part of the top ten last year. We have all regions covered now except maybe the SE and TN Valley, if I get one from there, might combine them since TN valley seems to be a much smaller group than all the others. 

Will be bumping all the threads overnight in regional forums except maybe NYC if you want to do that one RJay, but more from your subforum is not our primary goal at this point. ;)

 

Link to comment
Share on other sites

6 hours ago, Rjay said:

@Roger Smith.If you extend the deadline for another day I can try to get more to join.  

RJay's Regional Rumble? Well, I was R.J. for one summer in 1972, the name did not stick (my initial is J). 

You've done well so far with that project, I only managed to shake out one new participant as far as I know in the regional threads. Still, we now have all the forums with at least one person (except TN valley but I am going to combine SE and TN valley in this contest). I don't see Neckbeard93 yet, but I think he's from the southeast as well. 

That's fine with me, I was unable to spend much time on line today so I am a bit behind in my proposed tour of the threads, so let's say anyone who enters by end of Jan 3rd PST (08z 4th) is in without late penalty, and after that it will become whatever it was supposed to be at that point. Thing is, the regional aspect of this means that people who join in Feb or later can still feel like full participants as the individual contest will not be the only thing of interest. 

I suppose anyone could edit if they want to that deadline, I'm not checking and I think there is enough model chaos that the extra day won't give away much information. 

But I am going to ride my horse to the cliff. :)

 

Link to comment
Share on other sites

  • Rjay unpinned this topic

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...