Thread:Jemann/@comment-4085424-20140217021358

So I coded up an alternative generator for Broad Challenges at Template:BroadChallengeRevisedTest. Rather than just find the first level that exceeds a probability, it then works out the actual probability for that level. Better illustrated by example.

Old list: Broad, Dangerous 11
 * 10 - very chancy (50%)
 * 11 - chancy (60%)
 * 13 - modest (70%)
 * 15 - very modest (80%)
 * 17 - low-risk (90%)
 * 19 - straightforward (100%)

New list:

It's more accurate than the old version for difficulty levels < 60. The downside is that some probability descriptors will get repeated (e.g. chancy and straightforward in the above). Also it's more complicated, relying on a whole sequence of templates calling each other (although still less fragile than the truncation stuff). So... should I substitute it in? 