Re: [glob2-devel] Full Nicowar behaviour assessment

[Bradley Arsenault]

On 3/3/06, Kieran P <address@hidden> wrote:
>
>
> > Lately, I've been trying to modify this so that it plays better on maps
> such
>
> >
> > as Garden 3, but these changes only make it play worse on open maps.
>
>
>
> > Would anybody mind observing its behaviour on Garden 3 on the most
> > recent commit? And compare that to how it behaves on Big pond, most
> > recent commit.
>
> For Garden 3, it is better, but not good enough. Nicowar prevously refused
> to build a swimming pool and racetrack in the area where you start off
> because there wasn't enough space, and thus in order to build one, had to
> travel far. By the time it was made, and training warriors, I was already
> dieing (as you probably already know). Now with the latest commit, it has
> built both within a good range because it doesn't worry about lots of
> spacing (still require some but not a lot) but it died because of an
> overwhelming castor force :( So some work still needs to be done. :(
>
Yeah, I noticed the same thing, which is why I think Castor handles
space better, because castor managed to fit a good sized base in its
small starting area.

> For Big Pond, it did very well. It won. No suprise as its a bigger map than
> the others. It used more spacing than Garden 3 did.
>
Heh heh, I do 99% of my testing on Big Pond so its no wonder it plays best here.
>
> > Its very hard to get a middle ground, unlike Castor, who seems to do
> > this quite effectivly.
>
>
>
> > It seems Castor plays a better conservative game than Nicowar, a much
> > better conservative game,
>
> > I hope for some feedback, as much as possible, about how to get it to play
>
> These things you will know already, but theres no harm in pointing them out
> again. Nicowar obvioulsy has larger maps conquored (with fruit) so no point
> disuccsing them. I compared the 2 AI's when running on small maps, and
> theres 4 main differences I could find.
>
> Building Spacing: Castor doesn't worry about making buildings that can't be
> upgraded, such as inns. This means it builds heaps next to each other and
> thus saves more space. So far, I have not yet seen a building placed by
> Nicowar that has been obstrcuted by something around it preventing an
> upgrade later. Also, Nicowar leaves lots of space around buildings. Not a
> single buildings white outline touches another buildings. While this allows
> units to move a lot more freely, it takes up more space, which could be used
> for more important things.
> ~ Suggestion ~
> Dont be fussy about spacing when nicowar can't find enough room. Build them
> where you can.

Sometimes this is exactly the difficulty. Nicowar builds its buildings
following fairly strict rules. It always leaves enough space to
upgrade the building + 1 square away from other buildings (not
ressources). So yeah, castor places Inns in pairs, like it does other
buildings when it has the space. I have seen it place barracks,
racetrack, and swimming pools in pairs). I have been thinking about
sperating spacing into a building specific attribute, so that a
building can be placed really close to an Inn, but has to be more
spaced for central buildings like the Swarm.

The when you can't find enough room part is iffy to implement. This is
exactly why I praise the Castor programmers like I do, because they
are quite good at doing "threshhold" based algorithms like this, where
as I tend to avoid threshholds all together, as they come with a vay
of inaccuraccy.

>
> Defence Tower Usage: Defence is another major thing. Nicowar would do a lot
> better if it had some type of programming telling it to build towers if it
> has limited space, so that they have more defence then they would with only
> warriors. Smaller maps would have weak, low level warriors as barracks would
> be scarse. Towers are small and powerful. They need to be implemented.
> ~ Suggestion ~
> Copy castors tower placement code if you need too, just change to have them
> build when barracks placement returns not enough space to build one, so it
> compensates for lack of warrior strength.

I don't need to copy castors placement code, my existing code works
just enough. The towers thing has been iffy in the past, it has almost
uselessly takin up workers when Nicowar could sufficiently defend with
warriors, including on maps like Garden 3. The change to have towers
constructed is incredibly simple, a couple of header file changes, but
requires much experiementation.

>
> Expansion: Another feature Nicowar includes that, while great on big maps,
> could present problems on smaller ones. It likes to expand to new lcoations
> all the time, which means it has to send it units out to build. Thats means
> that more units are being used for jobs other than whats needed.
> ~ Suggestion ~
> Workers should stay close to base, working on upgrading the barracks to
> level 3 (to hold lots of globs) and making warriors, otherwise, if they
> dont, the forces of castor will be too much. 1 level 3 glob can take out
> what? 4 level 1 globs?
Your suggestion is, then, to give more units to upgrading and less to
constructing, yes?


>
> Conversion: Your probably thinking, conversions are good. It helps make the
> army larger (or not, I dont know how you think). However, what if fruit is a
> fair bit away. A worker would have to travel to get the fruit and just like
> expansion, thats 1 less worker builing around the main base where the most
> workers are needed.
> ~ Suggestion ~
> A check needs to be put in place saying that if fruit is more than say 20
> sqaures away, forget about going to retrive it. That way you have more
> workers working on whats needed back at the swarm and surrounding buildings.

I don't control the retrive thing. If the workers can grab the fruit,
they will, so there is no control whatsover. The exploration flags are
put on the three nearest fruits of each type, which is a
reccomendation made from many people on this thread.

>
>
>
> > Castor plays very well on limited amounts of food or wood
> >
>
> Would the fact that Nicowar uses conversion have anything to do with why it
> does play well with limited amounts? Does converting units cost wheat/fruit
> or is it just there for show?

Converting is a very important aspect of the game, it doesn't cost
anything but the harvesting of fruits. I don't see conversion as a
negative aspect in *any* circumstances, considering that if it didn't
convert enemy units, it would go on happilly. If you do convert enemy
units, but they starve to death, you haven't lost anything, you still
have all those swarm made workers you would have if you had done no
conversion, but your enemy has lost something, his workers have just
converted to his opponents team and died, so no losses here or there.

>
> _______________________________________________
> glob2-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/glob2-devel
>
>
>