Each question would be able to specify a number of subquestions with predefined options, such as the colors, any intenger, etc. Or the editor can just write out multiple answers if necessary.
Seeing these should be optional in the sidebar. They could be useful for #47.
Any scoring system should correct for the ease of guessing by weighting more heavily the results from questions that had many possible answers. We can also store all answers ever given to the question to form a probability distribution over possible answers, and calculate an "ease of guessing" metric for each question. (A question with 20 possible answers but only two that get picked is easier to guess than one that's more evenly distributed.)
We could use Benford's Law as a prior on numeric answers, but that won't be necessary once the actual distribution is known.
Test-takers should select their choice of "entirely from memory", "CR allowed but no other resources allowed", and "CR, Gatherer rulings, Google". Store scores seperaty for each one.
While a test is being taken, other instances of RG at the same IP address (and on their account, if we implement accounts) should be disabled to prevent them looking up the question.
Each question would be able to specify a number of subquestions with predefined options, such as the colors, any intenger, etc. Or the editor can just write out multiple answers if necessary.
Seeing these should be optional in the sidebar. They could be useful for #47.
Any scoring system should correct for the ease of guessing by weighting more heavily the results from questions that had many possible answers. We can also store all answers ever given to the question to form a probability distribution over possible answers, and calculate an "ease of guessing" metric for each question. (A question with 20 possible answers but only two that get picked is easier to guess than one that's more evenly distributed.)
We could use Benford's Law as a prior on numeric answers, but that won't be necessary once the actual distribution is known.
Test-takers should select their choice of "entirely from memory", "CR allowed but no other resources allowed", and "CR, Gatherer rulings, Google". Store scores seperaty for each one.
While a test is being taken, other instances of RG at the same IP address (and on their account, if we implement accounts) should be disabled to prevent them looking up the question.