[Freeciv-Dev] Re: (PR#4387) The Return of the Rand() Moves

["rwetmore@xxxxxxxxxxxx" <rwetmore@xxxxxxxxxxxx>]

Raimar Falke wrote:
> On Wed, Jun 11, 2003 at 04:17:39AM -0700, Gregory Berkolaiko wrote:
>>On Tue, 10 Jun 2003, Raimar Falke wrote:
>>>On Tue, Jun 10, 2003 at 09:37:32AM -0700, Per I. Mathisen wrote:
>>>>On Tue, 10 Jun 2003, Raimar Falke wrote:
>>>>>>Unfortunately, I really don't see a way of computing the real average time
>>>>>>in PF.
[...]
>>The outline implementation will be a nightmare to implement and maintain,
> 
> Can I convince you that this isn't the case if I code the implementation?
> 
>>will slow the PF down significantly
> 
> That is true. I'm not sure about the "significantly" but it will be
> slower. If you want exacter data you have to pay the price. The same
> is true for the trireme.

I have yet to see any credible argument that "exacter data" of a average
quantity provides any measureable improvement to game play.

"Exacter data" is just not work any effort and is in general an approach that
will ruin any chance at developing a credible AI. If the AI can reliably choose
one of the top 3 or 5 "best" moves by any measure of analysis, it will be as
effective, whether it chooses randomly from these, or does some excrutiatingly
exact calculation to make its final pick.

Using CPU resources to measure a broader scope of effects on the off chance
that one of them is dominant, is far more productive than refining the 29th
decimal place in one or two effects that someone thought might be important
but has never really done any hard measurement of how much.

[...]
>>and will not satsify all users either.
> 
> You have to outline this.
> 
>       Raimar

Cheers,
RossW
=====