View Single Post
Old October 11, 2009, 18:12   #5
Angband Devteam member
Join Date: May 2007
Location: London, UK
Posts: 5,060
Magnate is on a distinguished road
Send a message via MSN to Magnate Send a message via Yahoo to Magnate
Originally Posted by fizzix View Post
I wasn't pointing out a bug, just an observation. The main observation, like you pointed out, is that the items with multiple sustains seem less. I'm not sure if this is true over a large set. It was really hard for me to calculate that from looking at the code. I could look at a large set, and writing a code to generate standart spoilers and analyze them for sustains is actually in my abilities, so I could do that if it is desired.

The other observation is that it seems from the code that the sustains are all treated as equally valuable. Whereas, in the standarts STR, DEX and CON are treated as the most valuable. INT and WIS are significantly less common, and CHR rarely comes up at all. I'm not sure that it's a good idea to mimic these frequencies though. It may not be desirable.
You are right that the code does not track the frequency of individual sustains, only the total number of sustains in artifact.txt. It does value them differently, but ironically this results in the less valuable ones occurring more often (because they're less likely to overpower a randart and trigger a reset). I'll make a note of this too.

Some python code to generate and analyse multiple spoiler sets would be most welcome. (Or any scripting language of your choice, as long as it comes as part of a standard Linux install.) So far I've been doing it by hand - checking each set of changes by generating a couple of dozen spoilers manually and using grep | wc to see if things are about right. Getting more scientific than this would be helpful for fine-tuning balance.
Magnate is offline   Reply With Quote