[Freeciv] Re: Measure the human
[Top] [All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index] [Thread Index]
On Wed, Dec 04, 2002 at 02:32:46PM +0100, Thomas Strub wrote:
> I hope that this wasn't the last contest. And i think that a contest
> could be a good place to try new patches.
Yes. It would be great to have contests on civserver.freeciv.org.
A challenge is an initial savegame plus a function on civgames to optimize
- in this case it was the function: find a savegame with a minimum number
of turns in which the assigned player has researched The Republic.
Of course the challenge is to reach a savegame that meets the predicate
and that can actually be reached from the initial savegame by playing it with
a specific version of Freeciv.
The problem I see is proof. Is it good enough to make the contestants
send in the full sequence of savegames? If it isn't, we need a file that
all the user actions, and that can be played back on request. Perhaps a tool
that provides this can be provided as a proxy, that listens to the
requests and records them to file. A second tool would read the record and
translate it into a sequence of client interactions. That is the hard part.
> There is a small problem with some random things in the game. I think
> for contests we should remove the random function.
A random value is fixed in the savedgame, I think, and it will determine the
result of all the "random" choices, provided that the game is run with
"equivalent" civservers (same version. must be same platform, too??)
But if a challenge starts with a fixed position, no huts, and doesn't
involve barbarians or other features that highly depend on chance,
I don't think randomness can play a very important role.
If challenges are limited to single player games (i.e. there is never more
than one player in the game at all) it will be even easier.