Re: hps of classes (class differences)

["Rupert G. Goldie" <rgg@aaii.oz.au>]

> From crossfire-request@ifi.uio.no Wed May 25 06:37:17 1994
> From: KARIM SANJABI <karim@ecst.csuchico.edu>

> Oh, and while I'm thinking about it ;) some monsters give out
> way too many exp.  Like demi-liches.  500000?  I have seen a 25 level
> barbarian wade through like 30 of them in one room no problem, then 
> he becomes 31st level or something.  I think that there need to 
> be tougher monsters.  I have taken monsters like the chineese dragon, liches,
> and the Demon lord, given them almost every attack, protection, a few
> immunes, and it is only a matter of seconds before they die. 
> 

Hmmm, I meant to mention the demi-lich a while ago. I believe it has waaaay
to many xp. If you run Crossfire with the '-m' flag it will print out a
table of all the monsters. The last two columns are, respectively, crossfire
xp and rgg xp. Crossfire xp are what are in the archetypes. rgg xp are 
calculated based upon a complicated formula I created and tuned with the
aim of automatically calculating xp for monsters. As the formula may not
fully encompass all of the nuances of the difficulty of a monster it is
reasonable to not use it as an automatic xp generator, but it is a very useful
metric for comparing the assigned xps of different monsters.
 
 I figure that if the assigned xp is more than about 3 times off the calculated
xp then something is wrong. The calculated xp for the demi-lich is 14683 - an
order of magnitude below the assigned 500,000. Quite a few of the new monsters
seem to have xp that are way out of kilter with the other monsters. Some time
soon I'll go through all the monsters and post a list of the ones I think need
changing.

> Anyone have knowledge in AI?  I would love to see the monsters be much
> smarter.  I myself have no programming knowledge in it, but am taking
> a really great AI class next semester.  A row of bullet walls has 
> scummed many maps with dumb monsters sitting in front of them.  Shouldn't
> bullet walls have friendly flags set so the monsters tear them down?
> 
> Karim 
> 

As I work at the Australian AI Institute I guess you could say I have some
knowledge of AI 8-). The main problem with making monsters smarter is 
computational time. We already have a few heuristics which are quick and have
made the monsters a bit faster (like making them run away when badly injured).
It is probably reasonable to make a few special monsters smarter so we don't 
chew up too much time while adding some challenging opponents. I would also 
like to see a better parser added for NPC interaction, but that will need 
some serious thought and work done on it as we can't change the method of NPC
interaction too often or we lose too much work put into maps.


Rupert
--
Rupert G. Goldie, Research Scientist                rgg@aaii.oz.au
Australian Artificial Intelligence Institute        
/\/\|| Level 6, 171 Latrobe Street, Melbourne, Australia