[freeciv-ai] Re: ai hp recovery patch v7

["Per I. Mathisen" <per@xxxxxxxxxxx>]

On Tue, 13 Aug 2002, Gregory Berkolaiko wrote:
> On Tue, 13 Aug 2002, Per I. Mathisen wrote:
> > I ran a few pitfight games, and here are the (quite surprising) results:
> >
> >              'hard' : 'experimental' (ie w/ patch)
> > 50%                 8 :18
> > 50% + triggerhappy  7 : 8
> > 25%              18 : 12
> > no changes          6 : 10  (control group)
>
> I am not sure what is meant by the control group.Is hard the same as
> experimental in the control group?If it is so, the 6:10 score when two
> same level AIs are fighting means that all other results can be very
> misleading.

Yes, the control group is an unpatched server running my pitfight scripts
to see how misleading the statistics can be. Turns out the margin of error
is huge.

> It is probably better to select 20 different seeds and to run test games
> for each seed for all four patch variations.Then there will be less
> fluctuations due to theinitial starting position.

Good idea. I have access to a Sun Ultra E450... my only problem is that if
I send civserver into the background, it stops working. Why is that?
There should be a way to run it completely uninteractively.

Yours
Per