Re: More revisions to exploring_worth

[Lincoln Peters <peters2000 at mindspring dot com>]

On Mon, 2002-07-22 at 04:52, Hans Ronne wrote:
> I think you need to multiply the percentile fractions, thus:

Thanks.  The percentiles are one part of the code that still generally
eludes my understanding, so I'll take your word for that.

> 
> Of the stuff that has not been factored in yet, I think the building cost
> is the most important one. It should be included in all the ai worth
> functions, but I think it is particularly important here, where you much
> prefer five destroyers to one battleship. It is not easy to do this,
> however (I already tried). The problem is that as soon as you factor in
> building costs there are at least some games where the ai only builds the
> cheapest possible units. The reason for this is that the worth values are
> very sensitive to the number of unit types in the game (at least for combat
> model 0 that uses the hit-chance table). Therefore, the balance between
> building cost and other factors like attack and defense values need to be
> carefully adjusted for each game. Haven't figured out how to do this in a
> simple way yet.

The first thing I can think of would be an inverse relation, such as
worth *= 1/CP.  Although I'd imagine you've already tried something like
that.  What if the numerator was somehow related to the number of known
cells, how many the unit could explore, and how many other types of
units could explore that area?