[freeciv-ai] Re: [Freeciv-Dev] Re: [RFC][Patch] AI can fly v2 (fwd)

[Raahul Kumar <raahul_da_man@xxxxxxxxx>]

--- per@xxxxxxxxxxx wrote:
> 
> On Fri, 12 Apr 2002, Raahul Kumar wrote:
> > I'm scribed. I know you are, and so is Raimar. Who's left?
> 
> I'm here.

4 people already. Ross, Greg, Petr are the only holdouts.
 
> > I don't want omniscient AI's. I thought the plan was to move away from
> that.
> 
> Sure, but right now we don't have analyzing/profiling code to go with a
> non-omniscient AI. The AI is stupid and players are complaining the 'hard'
> AI is not hard enough, so let us do what we can to make them suffer, ok?

I agree with making them suffer. I suggest a 'Petr' test for the AI. If the
hard AI can beat Petr on a regular basis, it's good enough. (Why Petr? Because
he says he's a bad player).

I suggest we put in analyzing/profiling code. Unfortunately, I can see big
slowdowns happening because of all this extra code. No way to avoid it. Why is
AI so incredibly computationally expensive!

<snip>
> My idea is a little less ambitions, which is simply to put an ai hook
> inside the server attack code, which records losses due to various types
> of units. This profiling data can be used to estimate better the desire
> for city walls, SAM, coastal, antiair, techs, high defense vs high
> movement, etc..

Love it. So we should call update_losses, every time a unit is killed. This
info would be very useful for humans as well.

It might also help in other features too. I can see unhappiness being related
to how many units you have lost vs your enemy in a war. Great way for the AI to
decide if it should quit. Losses 2 X your enemys, give up and sign peace.

> Also I keep thinking that the client side AI should iterate through all
> the enemy units it can see every turn, and record them. Then amortize
> these memories somehow every turn, so that older memories are less
> important than more recent. This is then used to estimate what kind of
> units we need to combat the enemy.

Amortise? I don't know why you want to do that. Just make older memories drop
off as they are replaced by up-to-date info. I see no need to keep info of what
city x looked like 15 turns ago when we have info from 2 turns ago.


> > > whatever it is, the its_time_to_invade() function should be on the
> > > todo list.
> 
> I think it is just some calculations that are screwed up somewhere. Don't
> know where, though.

Guess. Anything at all would be handy. I've looked at f_s_t_k, and
find_beachhead, and I suspect that one of these has the bug. 
 
Aloha,
RK.

People who are rather more than six feet tall and nearly as broad across the
shoulders often have uneventful journeys. People jump out at them from behind
rocks then say things like, "Oh. Sorry. I thought you were someone else."
{Guards! Guards!, 1989, Terry Pratchett}

__________________________________________________
Do You Yahoo!?
Yahoo! Tax Center - online filing with TurboTax
http://taxes.yahoo.com/