Closed FilliFilli closed 6 years ago
Tried the "sorting the list" approach but a) they kept calculating almost always a few, high priority units (costly leaders), and b) sorting is expensive and not necessarly needed.
Implemented the "sell more tickets" approach and it works well. During lottery extraction, every unit's priority is checked, and based on various factor extra tickets are assigned to it. This way all units have a chance to be extracted, with important ones having a greater chance.
We'd like being able to process more probably (=more often) units:
This is done by 'overnumbering' the priority list lottery, as for selling 4x tickets to the Tiger 1 😄 (unfair stuff, basically)
Need a data structure to hold
Need to redefine extraction process.
Basically gets implemented by elarging the 'number range' to "guys who should win", and enlarging the lottery size as a consequence (some camouflaging, against grunts and poors)