Collect stats on randarts
|Reported by:||magnate||Owned by:||magnate|
One good way to measure the validity of the randart generator is to produce useful stats. This is distinct from either the randart.log file or the spoiler and would include things like:
numbers of each tval (or indeed sval)
min/max/mean/variance for base weapon damage
number of weapons with multiple brands, all three x5 slays, etc. (this should be weighted by rarity)
off-weapon damage boosts, shots, blows, brands, slays
Etc. Ideally these stats would be submitted directly to a central server (along with the code revision) so that we can build up a picture of the degree of anomaly generated over time. It's fine if a 9d10 uber-trident shows up once every million games, but not if one shows up once every fifty games. Might have to pilot this bit with netangband.