A simple solution, if somewhat radical, would be to change the category
distribution. Instead of
4 history,
4 lit,
4 science,
...
suppose it looked something more like:
5 biography
4 events
3 things (animate and inanimate but not people)
2 places
3 documents
1 idea
1 word
1 formula
...
The numbers of course are just illustrative (someone could do a sample
of recent tournaments and see what kind of distribution is typical).
Science, history, lit are of course important and the way things are
studied in school, but they are not necessarily the way things are
stored in the brain. Some people like bio (people who like people),
some like places, some like vocabulary, some like works of creation.
Some cross-check to ensure that not everything was lit related or
science etc. would be important, just as we cross-check now to ensure
not everything is 20th century or Canadian.
Just some red meat for the wolves.
-- dml