abr, you need to try something like this:
Materials:
trusted agent
receiver
sender (abr)
Protocol:
1. abr sends ten words to a trusted agent via PM. These words will be in English and will not involve similar concepts (e.g., "truck" and "car" are too similar). They will be spelled according to Webster's Dictionary; words with alternate spellings in different forms of English (e.g., "theatre" and "theater") will not be used.
1a. The trusted agent works with abr (via PM) if there are problems with the list (mis-spellings, duplicate concepts). When both the trusted agent and abr agree that the list is good, continue to step 2.
2. The trusted agent posts a hash of the list here. (This is a process which generates an encrypted file. It is almost impossible to crack*.)
3. abr posts the specific day, time, and duration in which he/she will send the words to the receiver. The time will be specified in both LOCAL and ZULU (GMT) time so that the receiver knows when to be ready to receive. For example, if abr is sending to Jackalgirl and states that he/she will be sending at 6pm local / 10pm zulu, Jackalgirl (who is in Japan) knows that she has to be ready to receive at 7am the next day. Note that duration must also be specified -- i.e., that abr will send for 10 minutes, or 30 minutes, or whatever.
4. The receiver accepts and acknowledges the date/time via a post here. If the time is not good for the receiver, he/she proposes an alternate time, to which the sender agrees (this process should be repeated until both parties agree to the date and time).
5. At the specified date and time, abr psychically sends the word list to the receiver.
6. At the conclusion of the sending period, the receiver posts the words which he/she has received, or feels that he/she has received.
7. The trusted agent posts the words sent to him/her by abr. Forum members can generate a hash of this list: if the encrypted hash they generate matches the one the trusted agent posted, they know that the trusted agent did not change the words.
8. If the receiver correctly identifies seven out of ten words (with the correct spelling), the test will be considered a success. If the receiver correctly identifies six or less out of ten words, the test will be considered a failure.
This is what people mean by protocol and by specifying levels of success or failure, abr.
Edited to add:
*By itself, the encryption algorithm is very strong. However, it's possible to crack these files if the person who generates them provides clues. This allows people to "crib" -- that is, to guess at words in the hash and see if they can reverse-engineer the hash. Solution? Don't provide clues.
