ORIGINAL: Haree78
Some questions for those who design the perfect ships.
1) What mistakes does the AI make in ship/base design?
2) What do you always update on the AI designs because the AI design isn't how you would do
it?
3) What balance of weapons do you use? If you are using a race that has specialist weapons
such as the Boskara torpedoes how does this change your designs?
4) Do you use certain ships for specialist roles? Such as keeping capital ships very long range
while keeping destroyers as close up weapons?
5) How close to the max size do you keep your ships and bases?
6) Which ships/bases do you arm and not arm?
7) What designs do you put troops on? Or fighters? How many?
8) How fast do you do your various ships?
9) What is the right amount of armour or shields?
10) How much do you use Ion weapons?
11) What components do you never bother with?
Any other comments on AI ship design, or how you prefer your designs would be really useful.
Well that's not me but I'll try to give some feedback anyways...
First some specifics of my design weirdnesses and then on to the important stuff.
Personally I am a compulsive thourough-going micromanager with my designs. I redesign
everything at the start and update every appropriate design whenever a new technology
comes along.
My ships/bases are all specialists. (Love my monster frigate, yep I only usually build one, which
has missiles, no shields and 3 armour.) Often I will have several destroyer designs current, each
with a different weapon/defenses/range set up. Usually have system defence (collectors but
few fuel tanks) and long range (no collectors lots of tanks) fleets. Start with two explorer
designs (namers and surveyors) and add two more later (spies and scanners). Later game I will
sometimes have a specialist 'nebula' fleet without shields but loads of armour. And on and on...
Mining bases are tailored to their location.
Most military ships I keep near the maximum size. They will sometimes be a bit smaller so that I
don't have to add another life support/habitation step, or so that I can get an extra point of
speed. Explorers I go with about 175 (2 habs). Freighters tend to be around 250-300,
efficiency is key here and 8-10 cargo is pretty good. Miners are bigger but top out at an
efficiecy limit. Contructors I keep as small as possible (often 350 at start game) but I sometimes
have extended range/cargo varieties as the game requires. Passenger ships stay near the
maximum. Bases on the other hand I like to keep minimal (usually around 340, 2 habs, in the
early game) except for resorts which vary according to demand . Spaceports are likewise kept
minimal, 300ish for small 500ish for medium and whatever is needed for large (my one or two
major ship-builders), although I will increase size for extra weapons in trouble spots.
Bases are generally unnarmed except for spaceports which get fighters. Explorers, constructors,
miners, passenger ships and freighters are unnarmed. Mainly I use energy weapons backed up
by torpedoes and carriers packed with fighters (as many as I can fit). Each ship specialises in
one weapon type with secondaries only included for balancing space/power loadouts. I
virtually never use Ion weapons except for carriers and at colonies. I never bother with area
weapons or bombard, very rarely use rail-guns and don't generally research missiles past the
initial one which is needed to get to point defence. If somehow I end up with the technologies I
generally still don't find any reason to use them...
I will increase torpedo ships if I am Boskara but only to an extent. If I am Wekkarus I'm looking
for the point to switch over and don't change my ship balance. Particularly early on I find that
non-weapon specials (Haakon, Quameno) influence my designs more by allowing much more
usable space. Often I end up with them anyways... at least if the game lasts long enough.
Troops are only carried on troop transports, usually 3-6 on most and the occasional 'as many as
I can fit' if there are appropriate targets (I don't usually invade homeworlds... if I did I'd have
several of these).
The question of speed (as much as possible) and shields/armour (as little as possible) really I
can't answer outside of the context of a given game. Early game I try for around 39 speed for
most ships (yep, including freighters...) but I want most of my military faster than most enemies.
(My monster frigate though doens't need anywhere near this speed to run away from ardilus
and slugs...) Similarly with shields and armour I want my military to have enough but no more.
Usually start with 3 shields and maybe 4 or 5 armour. Might bump the armour a bit if I plan on
hunting kaltor lots with these maxos chumps. (Races differ too, Kiadians get by, easily with 2
shields...) My namer explorers on the other hand sometimes start with as little as 9 speed (I think
that might be the Quameno version, 11 for most races?). The surveyors are about 45-50 at
start game. Most civilian ships get 1 shield 1 armour but even these are luxuries really.
Constructors usually get 2 as they are slow and more valuable. (If I played with graveyards
this would change...) Passenger ships sometimes get a bit more protection later in the game
when space is more readily available.
The major problem with AI ship/base design is that it does not integrate research; a secondary
problem is its lack of 'awareness' of the game environment. The final problem is balancing out
the needs of different components (i.e. how many shields vs. weapons, etc).
To deal with the second problem the game could log and analyse data about the
ships/bases/monsters a given cpu empire has encoutered. Size, shields, armour, weapons, and
so on (in both type and number) logged by date and owner empire, and maybe even galactic
location (although the latter is likely to be more difficult to adequately quantify). Then put this
information to use in research and design by assigning somewhat flexible research and design
'builds'. A simple case: if the cpu empire is encountering fighters only it should set about
improving its point defense, and probably armour as well, by researching and then adding this
to their designs.
The more major problem of research AND design (the AI is currently pathetic and erratic
because these are not connected appropriately) could be addressed by assigning each race
(and probably each government type) a default research and design 'build' order. This order,
for replayability, should not be entirely static but have a series of weights which kick in through
the encounter data and a genuinely random component (which should not be overpowered...).
A good way to get an idea of various 'build' orders for the different races would be through an
analysis of human strategies. Barring this running computer vs. computer simulations would
provide decent information to weed out terrible 'builds' and establish some decent candidates
for testing on humans. Remember that humans behave very differently to cpus when playing
the game. Also remember the human playing experience should be 'fun', whatever that entails.
The design stuff proper really needs some sort of percentage of available hull size assignation. So
you might have 40% for primary weapon 20% secondary, 20% shields, 10% armour and
10% other. It needs subroutines outside of this for assigning adequate power, life support and
so on. What I suggest is some sort of schema (I remember somone mocking up a screen of this
when the custom AI ship design was being added*, and quite frankly it was what I was
expecting and I was severely underwhelmed) for each ship role/type.
Thus when designing a ship the AI would select the appropriate schema, access data on the
game environment, and then plug in the most appropriate components. It would research (semi-)
intelligently based upon race characterisitics and what it needs in the game. There are two
major parts to this - first writing the code for the routines, and secondly plugging in the
appropriate weighting of the various elements used as information for the AI decisions.
Obviously we can't have the second part without the first. If there were robust routines then
allowing the player base access to the weightings of various potential elements (for example
the percentages of the schemas, or how much impact frequently encountering missiles should
have on the resarch and design routines) might be the best approach.
Edit: one thing I forgot to mention (lost to a teeth gnashing occurence) is that the cpu should also have 'awareness' of the resources needed for components. For example if it is planning a missile heavy early strategy it should prioritise mining krypton; and if it is going for fast ships it should have a substantial weighting on acquiring carbon fibre sources. This should also run the other way: weighting the cpu to adopt a fast ships strategy if carbon fibre is particularly abundant.
*Edit 2: it could be I was misremembering this images by Arcatus and Shadow Tiger in this thread
http://www.matrixgames.com/forums/tm.asp?m=3065137&mpage=1&key=design