[Freeciv-Dev] Re: (PR#2415) autoattack patch

["Per I. Mathisen via RT" <rt@xxxxxxxxxxxxxx>]

On Wed, 4 Dec 2002, Thomas Strub via RT wrote:
> Ok, using a autoattackunitlist should remove the problems of most
> situations like that above.
...
> > This can be solved in server-only autoattack with a simple rule:
> > Transported units should attack before transporters, if at all possible.
>
> Think the list-solution can handle that too.

Not if the transport is the best attacker.

> > > > 4. There are situation where i don't like to loose units because the
> > > > autoattackalgorithm isn't perfect.
> >
> > Please specify.
>
> There a situations with low winchances (< 10%) and high winchances (>70%)
> where your algorithm isn't a good solution.
>
> Winchance 5 %, Chance other unit is winning 100%. You would attack, but
> with the attack the other unit isn't losing any MP + its possible the
> unit is getting Veteran. You lose a unit without doing damage.

(If you have a 1/20th chance of killing the moved unit, it is unlikely you
don't damage it. But I get the point. Note, however, that attempting to
second-guess the algorithm by using thresholds can lead to situations far
worse than situations without them. For example a unit with very good
defense but ridiculously good attack may never by autoattacked because it
falls under the threshold and anyone who try to move into its range get
wiped out by its autoattack - instant invincibility.)

> I think the damage should be higher than the value of a unit.

What value?

> Thats one parameter. And i think the user should have the chance to
> change that parameter.

What parameter? How should the user be able to change it?

> + making more damage attacking than losing the unit (or more units)
> because of the attack of the unit which arrived.

I don't understand.

  - Per

PS Please don't cc: me or the list. RT does this automatically.