[Freeciv-Dev] Re: stop ai from building useless buildings (2) (PR#1362)

[Jason Short <jdorje@xxxxxxxxxxxxxxxxxxxxx>]

Per I. Mathisen wrote:
> Changes from last patch:
>  -> some formatting fixes & more meaningful variable names
>  -> capital considered as a special case
>  -> SAM restricted to coastline of safe continents
>
> I have also introduced a new handicap, H_EXPERIMENTAL, and a new
> difficulty level, 'experimental'. This is for testing of new AI changes.
> 'experimental' is just like 'hard' but with H_EXPERIMENTAL set.
>
> So, to test new AI code, do like this:
>
>         if (ai_handicap(pplayer, H_EXPERIMENTAL) {
>           /* experimental code */
>         } else {
>           /* old code */
>         }
>
> Run a few dozen autogames to see if things improve or not. And then remove
> old code and handicap check before submitting.
I like the idea behind this (a lot), but I think you're going about it 
slightly wrong.  I have always considered the handicap list as something 
not directly related to the AI, but that could be related to any player. 
 So it should eventually be possible to give human players these 
handicaps (except they're really unhandicaps).  Not all of the handicaps 
fit this description, but most of them do.
I think it would be *easier* to accomplish what you're trying to do by 
using different difficulty levels.  For instance, with something like:
  if (ai_difficulty(pplayer) >= 8) {
    /* experimental code */
  } else {
    /* old code */
  }

you can then face off a difficulty-7 (hard) AI player against a 
difficulty-10 player.  (Note, there is no ai_difficulty(), it's 
something like pplayer->ai.skill_level instead.)
This doesn't work so well if all the difficulty levels are eventually 
used.  But I still feel having "experimental" be listed under the 
handicaps is somehow wrong...
Of course, if you're just using this patch locally you can do whatever 
you want.  But I think you should consider what the "handicaps" should 
be before pushing for inclusion of such a patch.
> Here are my results from autogame pitfights:
>         experimental vs hard:    52:39
>
> Not very meaningful, but:
>         hard vs easy:            25:1
>         experimental vs easy:    38:0
>
> I think that is a nice improvement. I also think it is ready to go into
> cvs now.
Wow, that's nice.  This information is very good to see :-).

jason