Jump to content
  • Member Statistics

    17,609
    Total Members
    7,904
    Most Online
    NH8550
    Newest Member
    NH8550
    Joined

January 2013 Forecast Contest, Temperatures


Recommended Posts

  • Replies 73
  • Created
  • Last Reply

Here's my suggestion on late penalties (after this month). Given the late start and deadline of end of 3rd, in this case and any other month when a late start is required, the time penalty would increase so that the first day of time penalties "catches up" to the late penalty schedule for normal starting times on first. That normal penalty should be 3 pts per hour late, per group of three stations, in other words, if you're four hours late, you're playing off 96 rather than 100 points for each station you forecast. Etc ... but if the late penalties start after the first, then using this month as an example, the time penalties would be four times that rate on the 4th to catch up to the normal schedule. By the end of the 4th (all contest times eastern) the penalties would then be equal to what they might be in a normal month. And that would be 96 points per station so you would be playing off 4 points per station. By noon on the 4th you would be playing off 48 points per station.  At this rate, the futility barrier is always going to be 0400h on the 5th. At that point, your forecast is scored from zero. Another rule would be that no penalty reduced score could be negative.

 

And here's a suggestion for easier scoring than standard deviation but which would give probably almost identical results. In general it would be the method I used to approximate, that is, 2 pt deduction per 0.1 error. But when actuals swing wider than 5 deg, there would be "free" points introduced so that from 5.1 to 5.5 you had 0.1 free error point, from 5.6 to 6.0 two free points etc. This would expand to as much as 1.0 deg for a case like last March's 10 deg anomalies. To explain what that means, the more anomalous months are probably going to result in lower scores and also it seems unfair to penalize as much for a guess of +8 against +10 actual, as it would be for +1 against +3 actual. So in that instance, the +1 forecast scores 60 but the +8 would score 80 and also anyone within 6 deg instead of the usual 5 would score something. Perhaps to make it even fairer, the rule could be more complex, the free range near the actual, and a lower drop off near the low score zone, calibrated like this, 1 pt instead of 2 pt drop for forecasts on the right side of normal. That would run as far as necessary to complete the equation so to speak, in the extreme case of +10, it would need to run from 0.1 to 7.9 (then 8.0 to 8.9 score 80 to 98 and 9.0 to 11.0 scores 100, etc, then the progression reverses although who would forecast +12 or higher even in this contest? ) ... For a more moderate case like actual +6.5, the free point zone is +6.2 to +6.8 (anything in there scores 100) and the two point zone runs from +2.4 to +6.1 as well as +6.9 to +10.6 (scores would be 24 at outer edges of these zones) so from +0.1 to +2.3 as well as from +10.7 to +12.9 scores would be 1-23. This system would help even out a month like last March when I would say the scores generated were too low and moderately good forecasts scored only a few more points than totally weak ones. We seem to get one or two of these months every year (who knows, maybe we will get a negative one).

 

This is the system I would use if I were scoring, and I think it would converge on the method we are (in theory) using now.

 

I also think that if the contest is announced in a timely way, late penalties should be automatic no matter what the calendar brings by way of holidays or weekends, it's my observation in this and other similar contests that most regular contestants show up on time regardless and it seems more "professional" to expect on-time forecasts no matter what the calendar says. However if the contest is announced late, or something like Sandy hits, then an extension is understandable.

Link to comment
Share on other sites

I don't wish to start managing the scoring without a group decision, perhaps some of the more senior members could give me the word on what we should do, I see that Mallow has accepted the idea of a change. Frankly with my proposed system all that would change would be the delay factor between end of month and scoring results, I have tested out normal months and found very minor differences between the two systems.

 

With the proposed tweaking of unusual months (anomalies > 5.0) you would see scores that reflected minor successes a bit more than was the case in March 2012. The method is easy to monitor for contestants and you can make sure your scores are correct (that was never a problem before, by the way, just saying that as I will probably have scores posted without a connection to a data file you might want to check the math, although they would be done in a data file so any errors would be transcribing errors, I have some experience scoring contests elsewhere and so I'm used to double-checking everything).

 

I just checked the thread and everyone so far has entered all six forecasts. I would still propose to keep scores for the "classic" and "expanded" contests.

 

I'm going to store the posted forecasts fairly soon unless I hear otherwise, and will place them in the thread in a way that you might find more interesting than alphabetical. Come back tomorrow and see what that is (the suspense may be killing you).

 

---------------------

 

On the subject of scoring the months where actuals are >5.0 (plus or minus) this would be the procedure (slightly revised from what was in an earlier post) ...

 

Scores would drop off by only 1 pt from the actual until forecasts below 5.0 were encountered, from there down to a meeting point, they would drop by 2. The meeting point would be determined by a 1 pt upward progression from 0.1. Example for actual of 6.0, scores are 90 to 100 for forecasts of 5.0 to 6.0. They are 1 to 10 for forecasts of 0.1 to 1.0, and 12 to 88 for forecasts of 1.1 to 4.9. At the extreme point of an actual 10.0, scores are all determined by 1 pt differences. For an actual greater than 10.0, the score zone moves up with that anomaly (e.g., 11.0 provides scores from 1.1 to max). Any forecasts on the high (low for neg) side of these cases would be scored by mirror image of those rules. Example, a forecast of 8.3 for actual of 6.0 scores like 3.7.

 

Looking at a March2012-like case, this is what three sample forecasts would have scored using my method unrevised, and what they would score in the revised version:

 

ACTUALS of 9.7, 10.2, 8.8 ........... Standard scores ...... Revised scores

 

Person A ... 6.4 .. 7.2 .. 5.4 .... 34..40..32 ... 106 ....... 67 .. 70 .. 66 ... 203

 

Person B ... 4.1 .. 3.0 .. 4.5 .... 00..00..14 ... 014 ....... 41 .. 28 .. 52 ... 121

 

Person C ... 2.0 .. 1.7 .. 2.2 ... 00..00..00 ... 000 ........ 20 .. 15 .. 22 ... 057

 

I think those revised scores give a fairer assessment and also normalize the month so that scoring range will be closer to most other months (it will still probably be lower as it should be if most forecasts are off by several degrees).

 

 

 

 

Link to comment
Share on other sites

January forecast contest -- entries

_____________________________________

 

Forecaster ....................... DCA ... NYC ... BOS ... .... .... ... ORD ... ATL ... IAH  .... .... pen (per fcst)

 

HOCKEYINC .................... +6.1 ... +5.6 ... +4.1 .... .... .... .... +2.2 ... +3.0 ... -1.2

 

H2OTOWN_WX ............... +5.3 ... +2.7 ... +2.6 .... .... .... .... +1.5 ... +1.1 ... -0.6

 

DEREK Z ........................ +5.3 ... +4.9 ... +4.6 .... .... .... .... +5.8 ... +4.3 ... +3.3
 

MIDLOSNOWMAKER ....... +4.7 ... +3.8 ... +2.6 .... .... .... ....  +1.0 ... +.3.5 ... +0.7

 

QVECTORMAN ............... +4.3 ... +4.5 ... +4.1 .... .... .... .... -4.0 ... -2.0 ... -2.0
 

MALLOW ....................... +3.2 ... +2.7 ... +2.8 .... .... .... ....  +1.8 ... +3.6 ... -0.2

 

RODNEY S ..................... +3.2 ... +3.1 ... +3.2 .... .... .... ....  +2.3 ... +2.9 .... 0.0

 

FRIVOLOUSZ21 ............. +2.9 ... +2.5 ... +1.2 .... .... .... ....  -1.3 ...  +1.7 ...  -1.1

 

INUDAW ....................... +2.5 ... +3.1 ... +2.8 .... .... .... .... +1.3 ... +0.9 ... +0.2

 

BLAZESS556 ................. +2.4 ... +2.2 ... +1.6 .... .... .... .... +1.3 ... +1.9 ... +0.3

 

POTTERCOUNTYWXOB.. +2.3 ... +1.4 ... +1.8 .... .... .... ....  -2.3 ... +2.6 ...  0.0

 

SACRUS ........................ +2.3 ... +1.4 ... +0.6 .... .... .... .... -2.1 ... +2.0 ... +0.5

 

NYCSUBURBS ................ +2.2 ... +1.7 ... +1.1 .... .... .... .... -1.5  ... +2.4 ... +0.5

 

SKIERINVERMONT ......... +1.8 ... +2.2 ... +3.0 .... .... .... .... -2.0 ... +1.0 ... -3.0 .... .... 23

 

STEBO .......................... +1.7 ... +1.8 ... +1.0 .... .... .... .... +0.8 ... +1.9 ... -0.4

 

SRAIN ........................... +1.7 ... +1.3 ... +0.5 .... .... .... .... -2.4 ... +1.8 ... -1.9

 

LITCHFIELDLIBATIONS ... +1.6 ... +0.9 ... +1.1 .... .... .... .... +1.3 ... -0.1 ... -0.4

 

BKVIKING ...................... +1.6 ... +1.2 ... +1.4 .... .... .... .... -1.1 ... +1.8 ... +1.2

 

UNCLE W ....................... +1.5 ... +1.4 ... +1.3 .... .... .... .... +0.8 ... +0.6 ... +0.5

 

WESTWIND .................... +1.5 ... +0.7 ... +0.5 .... .... .... .... -0.7 ... +1.1 ... +0.3

 

ELLINWOOD .................. +1.5 ... +1.2 ... +1.0 .... .... .... .... -0.5 ... +1.5 ... -0.6

 

 

Consensus ..................... +1.5 ... +1.1 ... +1.0 .... .... .... .... -0.8 ... +1.1 ..... 0.0



MN TRANSPLANT ............ +1.3 ... +1.2 ... +1.2 .... .... .... .... +0.9 ... +1.5 ... +0.8

 

SD .................................. +1.3 ... +2.0 ... +1.0 .... .... .... .... -0.8 ... +2.3 ... +0.6

 

CHICAGO WX ................. +1.2 ... +1.0 ... +1.2 .... .... .... .... +1.2 ... +1.3 ... +1.0

 

SHADES ......................... +0.9 ... +0.5 ... -0.2 .... .... .... ....  -2.2 ...  +1.7 ... +1.5

 

WEATHERDUDE.............. +0.8 ... +1.1 ... +1.7 .... .... .... .... -1.5 ... +2.0 ... -4.4

 

CANDYMANCOLUMBUSGA +0.8 ... -1.1 ... -0.9 .... .... .... .... -0.2 ... -1.1 ... +0.4

 

WXDUDE64 ................... +0.6 ... +0.3 ..... 0.0 .... .... .... .... -0.5 ... -0.2 ... -0.3

 

WHITEOUTMD ............... +0.5 ... -1.0 ... -0.5 .... .... .... ....  -0.9 ... +0.2 ... +1.0

 

TOM ............................. +0.5 ... +0.2 ... +0.3 .... .... .... .... +0.9 ... -0.5 ... -0.3

 

TSTEEL ........................ +0.3 ... -0.5 ... -0.5 .... .... .... .... +0.5 ... +1.3 ... -0.5

 

GOOBAGOOBA ............... 0.0 ... -0.3 ... -0.5 .... .... .... .... -1.2 ... +0.9 ... -1.5

 

 

Normal ........................... 0.0 .... 0.0 ..... 0.0 .... .... .... ...... 0.0 ..... 0.0 ..... 0.0

 

 

CHICAGO STORM ........... -0.3 ...  -1.0 ... -1.0 .... .... .... .... -3.0 ... -0.3 ... -0.7   

 

DONSUTHERLAND1 ....... -0.5 ...  -1.0 ... -1.0 .... .... .... .... -0.8 ... -0.5 ...  -0.6

 

NZUCKER ...................... -0.5 ... -1.2 ... -1.4 .... .... .... .... +0.9 ... +1.2 ... +0.7
 

BRAD1551 .................... - 0.8 ... - 1.2 ...- 2.0 .... .... .... ....  -1.5 ... -0.9 ... +1.6

 

CTBLIZZ ........................ -0.9 ... -1.7 ... -2.0 .... .... .... .... -1.0 ... +1.5 ... +1.8

 

HUDSONVALLEY21 ........ -1.2 ... -1.3 ... -1.5 .... .... .... .... -1.8 ... +0.2 ... -0.8

 

ISOTHERM ..................... -1.3 ... -1.9 ... -2.3 .... .... .... .... -2.5 ... +1.1 ... +1.5

 

ROGER SMITH ............... -2.2 ... -3.0 ... -1.3 .... .... .... .... -4.5 ... -3.7 ... -2.0

 

OKIE333 ........................ -6.4 ... -4.1 ...  -2.0 .... .... .... .... -8.4 ... -3.9 ... -3.0

______________________________________________________________________

 

Notes:

 

Entries are arranged in order of DCA temperature forecasts. Check yours.

 

41 entrants, median is 21st and this determines consensus forecast.

 

I propose to be lenient on the time penalty this first month (disregard penalty x 4 to catch up, the penalty imposed on one late forecast is for 23 hours x 1 -- it means each location forecast would be scored from 77 rather than 100 as for the field).

 

This is not yet official, anything about this arrangement can be changed by group decision.

 



 

Link to comment
Share on other sites

After 11 days and a warm spell for several days but colder air on the way south, here are the anomalies through 11 days with estimates for values after today based on reported anomalies.

 

DCA ... +4.7 (+5.5)

 

NYC ... +4.6 (+5.5)

 

BOS ... +2.3 (+3.1)

 

ORD ... +5.2 (+6.3) ... today +21 but scaled to +15 in expectation of midnight low

 

ATL ... +6.0 (+7.7) ... today +26 with record of 76 F reported (75 F 1890)

 

IAH ... -0.8 (0.0 est) ... estimated from today's regionals

 

To get to +1 values that are close to consensus, the rest of the month (13-31) must average

 

DCA ... -1.8

 

NYC ... -1.8

 

BOS ... -0.2

 

ORD ... -2.4

 

ATL ... -3.2

 

IAH ... +1.6

 

If you predicted -3 for any station, you'll need

 

DCA ... -8.3

 

NYC ... -8.3

 

BOS ... -6.8

 

ORD ... -8.8

 

ATL ... -9.7

 

IAH ... -4.9

 

the rest of the way.

Link to comment
Share on other sites

There's a lot in the bank but here come the customers looking for their money -- this will get very interesting now with most of these anomalies falling off by 3 or 4 tenths a day. Could see some -20 days in the northeast and Chicago next week.

 

It would only take four days at -20 to wipe out a +4 anomaly around the 20th. Record cold days sometimes have -35 anomalies (I think 1-22-1857 managed a -40 anomaly at Toronto and 1-15-1994 was close).

 

Not saying it will get that intense but there are some -15s for sure in the mix now and maybe -18 or even -20 days. I think everyone's in the game at the moment except paradoxically those who are currently near verification.

 

BTW, if you're going to post anomalies (same goes for forecasts), could you kindly post them in the contest order? I recently went over all the entries to make sure my table was error-free and spotted your scrambled order, that was the only required scoring table correction I found.

Link to comment
Share on other sites

After 20 days, the torch is passed to a new administration (the cold administration, we hope ...)

 

DCA ... +6.4

 

NYC ... +6.1

 

BOS ... +5.7

 

ORD ... +4.5

 

ATL ... +8.6

 

IAH ... -1.7

 

so, with a few -10 to -20 days in the mix for the northeast, these should be changing and for the northeast, Chicago and Atlanta, this is what a -10 average 21st to 31st would create (just as an extreme scenario, not saying this will actually happen, but a fairly large negative wallop is coming down the pipes ...)

 

DCA ... +0.6

NYC ... +0.4

BOS ... +0.1

ORD ... -0.6

ATL ... +2.0

 

Another way of looking at it, BOS has enough positive in the tank to equal six days of -19 anomalies (with -25 potential ahead).

 

Houston on the other hand needs an average of +3.1 to reach zero anomaly for the month.

Link to comment
Share on other sites

After 20 days, the torch is passed to a new administration (the cold administration, we hope ...)

 

DCA ... +6.4

 

NYC ... +6.1

 

BOS ... +5.7

 

ORD ... +4.5

 

ATL ... +8.6

 

IAH ... -1.7

 

so, with a few -10 to -20 days in the mix for the northeast, these should be changing and for the northeast, Chicago and Atlanta, this is what a -10 average 21st to 31st would create (just as an extreme scenario, not saying this will actually happen, but a fairly large negative wallop is coming down the pipes ...)

 

DCA ... +0.6

NYC ... +0.4

BOS ... +0.1

ORD ... -0.6

ATL ... +2.0

 

Another way of looking at it, BOS has enough positive in the tank to equal six days of -19 anomalies (with -25 potential ahead).

 

Houston on the other hand needs an average of +3.1 to reach zero anomaly for the month.

I like your thinking, mainly 'cuz that puts all the numbers in my ballpark minus ATL. We will see, that is a lot of positives currently listed  to overcome.

Link to comment
Share on other sites

Archived

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

  • Recently Browsing   0 members

    • No registered users viewing this page.

×
×
  • Create New...