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

["Gregory Berkolaiko" <Gregory.Berkolaiko@xxxxxxxxxxxx>]

On Wed, 3 Sep 2003, Jason Short wrote:

> John Wheeler wrote:
> > [glip - Sun Aug 31 18:01:27 2003]:
> > 
> > 
> >>/*************** Current mode (unittools.c/try_move_unit) **********/
> >>
> >>We have 3 numbers: full move points (F) of a unit, current (C) move
> > 
> > points 
> > 
> >>of the unit and the required (R) move points for the move being tried.
> >>
> >>The move will succeed with probability 100% if
> >>either
> >>  F == C
> >>or
> >>  C >= R
> >>Otherwise, it will succeed with probability C / R.
> >>
> >>/************************* Proposal ******************************/
> >>
> >>The proposal is to implement a system where no probability is needed.  
> >>There are three possible flavours:
> >>1. Strict -- only moves with 100% probability will succeed.
> >>2. Fast -- as long as C > 0, the move will succeed.
> >>3. Mixed -- move will suceed if probability > threshold value.
> > 
> > 
> > 4. Negative -- if C <= 0, move always fails, else move always succeeds,
> > and C -= R [which means that C would need to be a signed integer]. 
> > (When a new turn comes, C += F.)  So, for example, a warrior would have
> > to rest for two turns after climbing a mountain.'
> 
> I will add to this the current behavior:
> 
> 0.  Random - move has probability C/R to succeed.
> 
> It seems like we should just be able to have a single server option 
> (with a second option to control option#3) to determine how this is 
> done, and PF and others should then "just work".

Should just work?  Well, that's the whole problem, "just working" and "the
shortest path" are not well-defined and when a good definition comes (like
Chris' "probabilistic average time is minimal"), it is an NP problem!

G.