The rules for Round 2 of the Blind Prediction Competition are listed below. Note that these may be modified for the betterment of the contestants where necessary prior to 19th March 2021. No further changes to the rules will be made past this date.

 

• ​Contestants may participate as a team or as an individual. However, the same person may not be involved in multiple submissions.

​

• Contestants are to predict the total acceleration history and the relative displacement history of each of the upper floors for Tests #4 and #5 (see "Background" section for test details).

​

• Contestants should submit their results in a spreadsheet provided.

​

• Contestants must provide sufficient information on their model calibration and analysis approach. It does not need to describe the theory behind the approach adopted, but it needs to be detailed enough so that it could be replicated.

​

• Contestants will be provided with floor acceleration results from Tests #1 to #3, as well as the table acceleration recording for Tests #4 and #5. All recordings will be unfiltered. See "Background" section for test details.

​

• Contestants will also be provided with information on the building, such as structural drawings and material test results.

​

• Submissions will be judged using the "sum-squared-error" approach based on the following equation:
  
  
  
  
  Where:

​xpij = the predicted response (either acceleration or displacement) on floor i for test j at time t

xmij = the measured response (either acceleration or displacement) on floor i for test j at time t 

t1j = time corresponding to 5% of total Arias intensity calculated from measured roof acceleration for test j

t2j = time corresponding to 75% of total Arias intensity calculated from measured roof acceleration for test j

​

• As t1j and t2j are based on the measured response for events which contestants were not given the data for, they will be kept secret until the winner is announced. Instead, contestants will have to submit predictions of the response history equivalent to 100 seconds of shaking for each test and response type, where 0 seconds is the start of the provided excitation.

​

• Contestants will be ranked by their accuracy for each response type and each test (four categories total), with the lowest error receiving first place, etc.

​

• A contestant's ranking in each of the four categories will be added together to give the final score. The three teams with the lowest score will be considered the winners of the competition.