Optimal star fleets and other stuff.

A thread of 2 messages, 7 October 1991 to 8 October 1991, in the order they were sent.

#2985 Optimal star fleets and other stuff.

Nick Christenson, University of Arizona · 7 October 1991

I've been following the discussion on optimizing fleets and I'd
like to add my 0.02 Cr.

First, invoking Game Theory, since a fleet battle is a two person
zero sum game, the Minimax Theorem *guarentees* there is *one*
optimal strategy (in this case fleet design) which is expected
(barring) luck to win a greater percentage of the time than any
other strategy.  Work has been done at Stanford by the EURISKO
an heuristic evolving computer algorithm, on the TCS problem!  It
seems that a fleet of relatively small ships (I don't know the
sizes) with some Meson Gun spinal mounts and some PA's is a
better fleet than an homogenous one.  Note, that I don't believe
that EURISKO has found *the* optimal fleet for any TCS parameters
as it is unable to consider radical design changes which require
creative use of obscure rules and prescedents in the Traveller
rules.  I think I could beat EURISKO this way in TCS (if I get
some choice in the parameters to be used) but once an heuristic
program allows from some of my ideas, it will probably beat my
fleet pretty handily.  For those of you who are interested, the
only printed reference I have for EURISKO and TCS is pp. 72+
of _Engines_of_Creation_ by K. Eric Drexler.

It is true that there is more to a space based navy than TCS
ships, but it is my feeling that the main battle line would
be designed in this way.  Also, once a major interstellar war
started, production would shift to favor these designs (along
with Troop and *SUPPLY* ships!)

Side comment on programming GM aids for Traveller:
I don't like Graphics Interfaces at all, but a majority of
*buyers* do.  Also, I don't agree that you have to make this
thing horribly portable.  I say make a universal data processing/
storage/retrieval system that you can *add* graphics interfaces
onto for different platforms.  I don't think that *many* gamers
have problems with wanting to run the thing on a Mac some nights,
a VAX others and an HP every other major holiday.  Make a
universal driver and then interested parties can write an ASCII,
Mac, PC, NeXT or whatever driver for it as they see fit.

Nick Christenson
[email removed]

#2991 Re: (2985) Optimal star fleets.

Brent L. Woods · 8 October 1991 · reply to #2985

 In TML article 2985, [email removed] (Nick Christenson)
 writes:
 >Work has been done at Stanford by the EURISKO an heuristic evolving
 >computer algorithm, on the TCS problem!  It seems that a fleet of
 >relatively small ships (I don't know the sizes) with some Meson Gun
 >spinal mounts and some PA's is a better fleet than an homogenous one.

     Now, this strikes a chord in my memory.  Unfortunately, my memory
is somewhat less than perfectly reliable.  You Have Been Warned.  ;-)

     The whole thing happened a few years ago (I'm not quite sure how
long, but it wasn't recent), when there were TCS tournaments at some
large gaming con (GenCon?  Origins?  Something like that).  Anyway, some
people put some *serious* computing resources into devising optimal
fleets (as well as combat strategy, if I remember correctly--*that* must
have been fun to program).

     If the fleet in question is what I'm thinking it was, its
composition isn't immediately obvious (at least to me).  The fleet
consisted of lots of medium sized (say, 50 kton) *planetoid* ships
(don't know if they were buffered or not--hardly matters, I suppose)
armed with Meson spinal mounts (at least *that* part I had already
figured out).  I don't remember what secondary armament (if any) the
fleet had, but that's probably more flexible.

     I remember trying that fleet composition out (I used to play quite
a bit of BCS and TCS with some ROTC friends) and finding that it worked
fairly well.

     In any case, the story ends with the end of TCS tournaments, since
the optimizers kept winning (rules changes didn't help, since they just
ran a new simulation).

 >Note, that I don't believe that EURISKO has found *the* optimal fleet
 >for any TCS parameters as it is unable to consider radical design
 >changes which require creative use of obscure rules and prescedents in
 >the Traveller rules.  I think I could beat EURISKO this way in TCS (if
 >I get some choice in the parameters to be used) but once an heuristic
 >program allows from some of my ideas, it will probably beat my fleet
 >pretty handily.

     Bear in mind that, while the fleet composition (and, perhaps, the
combat strategy) were mostly determined through computer simulation,
there was an actual live human who fought the fleets in the tournaments.
Things would probably come down to the old saying attributed to Manfred
Von Richtofen (paraphrased):  "It isn't the quality of the crate, it's
the pilot flying it."


- --
     Brent Woods

INTERNET:  [email removed]
[address removed]
PHONE:  [phone removed] (voice)