Jump to content

  • Member Statistics

    17,576
    Total Members
    7,904
    Most Online
    BlueSkyGA
    Newest Member
    BlueSkyGA
    Joined

2013 North Atlantic tropical season contest -- provisional contest results

Roger Smith

Recommended Posts

Roger Smith

Roger Smith

Posted
Posted

This thread is open for seasonal forecasts of the 2013 North Atlantic tropical storm count, as usual, predict the numbers of all named tropical cyclones, hurricanes and major (cat 3+) hurricanes.

 

Then add your forecast for June, same parameters. You can use decimals to indicate your estimate of uncertainty.

 

A forecast should look something like this, last year's actual count: 19/10/2

 

Now there has been some discussion about who might run this contest and I hope it won't be me, but I will promise to run the thing if there are no volunteers. Preferably, somebody else can offer to manage the contest and give their scoring rules ... we don't need to know the rules to start recording the forecasts.

 

The default position on penalties for round one (June and seasonal) is that the contest deadline is a lenient end of Monday, June 3rd (EDT) with penalties on some basis after that. But the organizer who steps up might want to revise that.

 

My own forecast will be 17/10/4 and for June 1/0/0.

 

All forecasts can be edited up to the deadline.

 

Just a thought, maybe we should add a forecast for the most significant landfall in the U.S. with category and date. If so, I would go with cat-3, e NC, Sept 18th.

 

Final contest rules (edited June 11, 2013)

__________________________________________

 

Note that these final rules are identical to the proposed rules except for any details added today in bold italics. The additions are mainly for clarity, the overall concept remains the same as in the proposed rules.

 

 

1. Seasonal forecasts will count for 50% of total score, monthly forecasts also 50% and the "major hit" portion will work as a 5% bonus score (at maximum). Standings will be posted at the end of the contest for the regular portions and the overall with bonus.

 

2. For the seasonal portion, 70% of that score will be derived from your June forecast, 15% from July, 10% from August and 5% from September. There will not be a requirement to update your seasonal forecast on October 1st. Late penalties apply to seasonal forecasts on a 50% reduced basis , see rule 5 for the late penalty.

 

3. For the seasonal scores, the baseline points (e.g. 70 for June seasonal) will be reduced by the average of total error squared and total error (where total error is the difference between your forecast and actual numbers).

 

4. For any later entrants who don't post a June forecast, a July 1st seasonal will score 60 per cent, the August 1st will score 15 per cent and the September 10 per cent. This represents a 15 per cent late seasonal entry penalty. For any super-late entrants who start with an August seasonal, that counts for 50 per cent, and September for 20, for a 30 per cent seasonal penalty. No entries will be accepted after August 5th at 10z (the time at which August monthly forecasts have zero scoring value). In earlier cases, a seasonal forecast made in July after the monthly absolute deadline (July 5, 10z) will count as an August 1st seasonal forecast, and the same for June after June 8th at 10z, that would count as a July 1st seasonal.

 

5. For monthly forecasts, June will be worth 8%, July 15%, August 25%, September 30%, October 20% and combined Nov-Dec 2%. Forecasts will be due on or before first of each month at 06z of first day. Following that, penalties of one per cent of score will be applied per late hour. Monthly scores will be calculated the same way as seasonal, the average of error and error squared. One minor detail will change from past years, your October monthly forecast will require a "Nov-Dec" total forecast as a separate item, and at that point the contest will be closed. Monhly forecasts will be verified in the usual way, storms acquiring a name in any given month will have their entire history confined to month of naming. Example, a storm is named August 30th then goes on to become a major hurricane in September. It counts as a major August hurricane. My suggested policy on post-season adjustments is that the scoring will be done before that happens and adjusted scores can be posted for your interest but the pre-adjustment scoring will count as official. Post-adjustment scoring would take into consideration changed logic for seasonal updates (you would see scores based on the logic of what that adjustment would mean to your update). Any NHC adjustments made in-season (before scoring) that would affect logic of your updates will carry a guarantee of update adjustment if it favors you. The logic of this must be based on adjusted prior information not post-forecast. No in-season changes can alter seasonal forecasts made before the adjustment applies. Late penalties apply to monthly forecasts on the basis of 1% reduction of score (after minimum progression applied) per hour after deadline, which is 06z of the first day of the month. This implies that the absolute deadline for scoring entry is 10z on the 5th of any month (all dates were shifted three days later for June to encourage entries).

 

ADDED July 12th ... As announced in the thread, the July penalty rate was dropped for the first day to 1% per 6h late, due to timing of the weekend at end of June. The August penalty rate and the October penalty rate will be unchanged as with June. The September penalty rate will be situation-dependent as follows. There will be the Sunday and Monday of the Labor Day weekend at the start of September. Because this is the heart of the hurricane season, a large (75% reduction) penalty reduction would be inappropriate to this contest. If there are no new named storms between Aug 31 06z and Sep 3 06z there will be half-scale penalty points for late entries to 06z Sep 3 (meaning 1% every 2h late), followed by normal on 3rd to 06z 4th (meaning 1% per hour late), and twice normal rest of 4th (meaning 1% every 30 min late) to 06z then normal again to the normal expiry time at 10z 5th. At whatever point in time a named storm is declared in that reduced penalty time window 06z Aug 31 to Sep 3, even if it falls on August 31, penalties will then revert to normal and the catch-up on the 4th adjusted as necessary. This is to protect contestants who have entered in late August, and left for the weekend without the same information available. The penalty situation will be monitored and announced in real time during that phase of the contest.

 

6. Before time penalties, a minimum progression of scores must occur in cases where the field scores low -- this minimum progression requires one score of 80% max, one of 70% max, one of 60% max, then one each at 50%, 40%, 35%, 30% on down to 5%. Where some but not all parts of minimum progression are improved upon, the unimproved portions then drop by 3%. Example, somebody scores 82% but nobody else breaks 70 or 67, then the second highest score is 67%, etc. Only one reduction of 3% will be applied to relevant levels, regardless of how many links in the chain are broken by higher scoring entrants. From past experience this minimum progression or mercy rule will apply only in cases where nature serves up a very unusual result and will reward those who came closer to predicting it. Time penalties are then applied to the set of adjusted scores. Scores are rounded to nearest one decimal when calculating decimal-based forecasts or percentages.

 

7. The distribution of monthly scoring maximum values (8, 15, 25, 30, 20, 2) is locked in, if the season has an odd distribution, your scores will improve relative to the field, if you forecast on that basis. This applied in 2012 (see Nov 2012 thread for my version of scoring that contest, August was above climatology and September below, my alternative scores were based on an approximation of this year's rules).

 

8. The bonus question will be scored by rank order, the best forecast will get the 5% bonus, and every other contestant will be ranked with the first 10 getting bonus points, differential of 0.5% per ranking. These ranks will be determined by smallest error in total of date error for max landfall (date by z time, error points on this basis, one for 1 to 3 days out, 2 for 4 to 10 days out, 3 for 11 to 20 days out or when correct month but no specific date predicted, 4 for 21 to 40 days out and 5 for larger time errors or no date predicted) and category error plus location error of 1 for 100-300 miles, 2 for 300-500 miles, 3 for 500-800 miles, 4 for 800-1200 miles and 5 greater than 1200. Where only a state or other non-specific location is given, the implied landfall will be taken as the central point of that state or region's coastline. No specific location will get a maximum (5 pt) score  ...  Max landfall to be determined on November 15th by consensus (a post will be placed on thread for discussion at that time). If there are two storms that are very close to being equal, then I would propose that the scores be based on better verification on either storm. If there is no landfalling named storm in the season, then a person would have to predict that to get the bonus. Otherwise, all contestants will be equally scored at zero. Where more than one forecast of severe landfall is made, the stronger will count towards the determination and the other forecasts will not be used. added June 28th ... contestants who add or revise a "max landfall" forecast during the contest will be considered entered at full score potential less two positions for late entry, before July 15 or the first named hurricane, whichever occurs first, and after that, no new or revised entries would count towards the bonus.

 

9. All further forecasts in the contest will be entered in this thread which will be bumped for visibility around the 25th of each month until October's final forecasts are made. Entrants who post monthly forecasts in any of July, August and September without updating their seasonal forecast will be scored from their most recent previous update. Entrants who do not post monthly forecasts at any of these update opportunities will have their previous seasonal forecast scored without penalty.

Link to comment
Share on other sites
  • Replies 183
  • Created
  • Last Reply
Roger Smith

Roger Smith

Posted
  • Author
Posted

Thanks for entries received so far. In the absence of any offers to manage the contest, I have posted some proposed rules for scoring in the original post so that newcomers will be sure to spot them.

 

Basically, if I manage the contest, it will be done in a very similar way to past years, with a few slight adjustments. There will be no seasonal updates required after September 1st, and instead of a poorly attended Nov 1st update, you'll need to post a Nov-Dec total with your October 1st entry.

 

I don't see any reason why this contest should not reside in one ongoing thread, so make a note, you may just be coming back to this thread to enter monthly updates (I will bump it to the top at around the 25th of each month).

 

If you care about the rules, check my proposed rules in post #1 and discuss, they can be amended at this early stage of the contest. Let's say whatever we have by consensus on June 10th will be final.

 

And I'm still quite happy to hand off responsibility for the contest, as long as there's a promise to provide scores in a timely fashion.

 

Just a reminder, the June portion of the contest goes into time penalty mode at 06z June 4th. All future months will go into penalty phase on the first of each month at 06z. Seasonal late penalties are discussed in the proposed rules.

Link to comment
Share on other sites

Archived

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

Go to topic listing

  • Recently Browsing   0 members

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