[Freeciv-Dev] Re: Requirements for gen eff AI

[banjo <banjo@xxxxxxxxxx>]

On Fri, 2004-05-14 at 02:26, Per I. Mathisen wrote:
> The second, speed is a problem. The new code uses exact measurement of the
> effect of several improvements by actually adding them and measuring their
> effect on the city's outputs. We only do this for city improvements that
> have measureable effect, and only every Nth turn (where N is currently
> myrand(10)).
> 

>From looking at what your code is doing, is it possible to 
precompute some effects (or effect metrics) at server startup
(& maybe store this scenario data in the savegame)?

Would it help for roles (or hints) to be added to buildings
rulesets to aid the ai routines?

Also, once a city is commited to building something, should the
analysis be held off until 1) it's completed the task & needs to
choose a new target, or 2) the city is under threat and needs to build
a defensive unit, or 3) the grand strategy has changed and the city
should build a boat, or caravan or ... (this would imply that cities
should tag their build task with a desireability level 1-5 ( critical,
important, useful, nice to have, something to pass the time with)). 

> The advantage is that the result we get are exact, unlike the guesses the
> previous code made, which were by necessity often wildly off the mark,
> being quite simple algorithms.

Nice! but human players don't tend to overly analyse improvements
efficiency, but have rough rules to guide their use of certain
improvements.  Of course rndCiv will (hopefully) mean that each game 
will be different enough and require reanalysing effects - though
after some practise rough meta-rules will prolly get developed.

> So, the question is, how much slowdown is acceptable?

Maybe a modifyer to N can be set on the server, this i suppose
could be considered an AI efficiency modifyer.