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

["Ross W. Wetmore" <rwetmore@xxxxxxxxxxxx>]

At 01:04 PM 02/04/12 +0100, Gregory Berkolaiko wrote:
>On Fri, 12 Apr 2002 per@xxxxxxxxxxx wrote:
>
>> > > - The AI should not attack if it stands a very high chance of losing
(ie
>> > > mech inf on a mountain). Aircraft are expensive. Waste some commoners
>> > > instead.
>> >
>> > I use the standard want equation (meaning that on average we'd be better
>> > off after attacking)
>> 
>> Yes, but it doesn't seem to check if we have a good chance of winning. It
>> will just attack the "best" target, even if "best" is abyssmally bad.
>
>what do you propose as a measure of "good chance of winning".
>
>Examples:
>1. you put 1$ on red and it wins with prob 1/2 and you get back your 1$ 
>plus 0.5$.  Is it good?  Do you play?
>2. you put 1$ on something that wins with prob 0.75 and you get back your 
>1$ plus 0.3$.  It this good?
>3. you put 1$ on number 13 which wins with prob 1/32 and you get back you 
>1$ plus 100$.  Is this any good?

Missing concepts ...

1)  Units should try to stay alive. It is better to kill two $1 units you can
    easily knock out than to die trying to get one $3 one. Unless there are
    compelling reasons always wait for a 3/2 advantage on attack or something
    more than simple 1:1. Attack at less than 1:1 only in critical need or
    when the units are really not needed. In general, better odds will
    usually arise if you delay (safely of course).

2)  Units should decide if they stand a better chance letting a unit
    attack them, than attacking the unit.

3)  Units should take into account critical surrounding force levels before
    doing the above random gamble attacks, i.e. be prepared to lose aka
    adjust the loss value by something other than pure single unit cost.

4)  Units should make subjective target decisions based on strategic
    advantage, i.e. what unit can do the most return damage, vs just how
    fancy it is dressed (i.e. build_cost).
>G.

Cheers,
RossW
=====