On Thu, Dec 06, 2001 at 01:07:31PM -0500, Mark Mielke wrote:
> The primary benefit being... that cloaking is not an effective technique
> against robots that can see right through your cloak.

Cloaking is not an effective technique against good players, either.

> You are telling me that I am approaching things from the wrong angle. I
> ask you... when was the last time you wrote a bot that could successfully
> win against most, if not all INL teams?
> How would you know which angle was the wrong one?

Scott Drellishak, an early server administrator and author of a
netrek-playing bot, once hacked his server to report heuristics on
things such as % of phaser hits, % of torps dodged, ability to do
things on both sides of the ship, etc.  His idea was to attempt to
detect robots and borgs based on their tactical characteristics.  He
found that top players were indistinguishable from borgs and bots.

Scott's robot's dodging was probably optimal.  Optimal phasering is easy.
Optimal torping is fuzzier, but not that important anyway.  And his
robot sucked at playing netrek.

Every ehnancement you've talked about is to tactical stuff which top
players, and existing robots, ALREADY DO ALMOST OPTIMALLY.  It is
extremely unlikely that any amount of tactical work will create a
robot which could beat sfd's in a dogfight more than 55% of the time.
Perfect tactical play is a SOLVED PROBLEM in netrek, so you can spend
the rest of your life working on perfecting tactical play and not get
anywhere.  If you don't believe me, fine, go waste your time, what do
I care?

The interesting question is, when you've just won a dogfight at the front,
have one kill, 1000 fuel and 20% damage, do you go back to core to pick,
go to a front-line fuel planet to refuel, refuel in place, or try to
mutual with an enemy, what is the proper thing to do?  There is no
one answer, and it is EXTREMEMLY DIFFICULT to codify a reasonable 
decision; there are too many variables.  It's not impossible, but if
this isn't the part of the problem you're working on, you will not get
anywhere with a netrek robot.  The game of netrek is about making this
kind of multivariate decision CONSTANTLY.