T-hawk Plays Alpha Centauri

[antisocialmunky]

Have you tried any crazy variants with HoMM3?

https://www.youtube.com/channel/UCm-6vLD...azXSiKzOXg
His single Hero run was pretty interesting.

[T-hawk]

So this was fun enough that I wanted another trip through it. This time as a full standalone writeup rather than a forum thread.

http://dos486.com/alpha/morgan/

[LKendter]

Morgan LOL
I still want to do revenge of Morgan after my loss with him.

[Modo]

Thought the next one we'd see out of you would be after the mod is finished (have to dig the thread up but I recall someone wanting at least to mod this game) and now not only a new one but a teaser for potentially more...yes, please and thank you.
Old loves die hard...or habits never die or something.

[induktio]

Certainly an interesting micromanagement-heavy way to play the game... those base numbers reminded me to post about the details on how the tables are stored internally. This game doesn't use any fancy data structures, instead the bases and units are stored in statically allocated arrays and counter variables are used to keep track how many positions are in use. Let's say we have these bases:

0 1 2 3 4
A B C D E

Addition of new bases is easy, just write to the end of array and increment the counter. However, what happens when, say, base B is destroyed? In that case the game will shift each following base backwards by one position like this:

0 1 2 3
A C D E

This preserves the relative ordering by the founding date but makes the delete operation really costly in terms of data structure performance. It also has to adjust each unit home base id following position 1, and probably it's bugged in some way when sometimes the units home to a wrong faction's base.

[T-hawk]

I'll stress that I played that way because it really is optimal, at least in terms of transcendence finish date for Morgan, though it also would be for most any goal with most any faction. Any less ICS than what one thinks is optimal is a self-imposed variant restriction, whether or not one actually says so in such terms.

Modo, I don't intend to play any mods. I'll follow the development with interest and lend suggestions, but ultimately what I care about playing is the original unmodded reference platform. It's always a slippery slope and moving target, which the only way to avoid is to not start.

Interesting info on the base ordering. I was indeed surprised when new bases didn't occupy slots left by deleted bases (I intentionally disbanded a couple to rebuild at size 3 via PTS.) I could've sworn they did (it matters for bureaucracy drone assignment, though I paid no attention to that this game), but maybe I'm remembering Civ 3 or something else. The approach makes sense - bases get looked up in the array hundreds of times per turn, but deleted only a few times per game, and even that cost is probably still measured in milliseconds even on SMAC's original hardware.

[induktio]

Modo, I don't intend to play any mods.  I'll follow the development with interest and lend suggestions, but ultimately what I care about playing is the original unmodded reference platform.  It's always a slippery slope and moving target, which the only way to avoid is to not start.

I wasn't even talking about mods, but that's the way I sometimes feel about them! "Balancing" game mechanics is such a topic that it is really easy to fall into the rabbit hole and modify the game into something unrecognizable. With Thinker, the goal has been mostly just AI improvement and it has to be said rebalanced mechanics are really an orthogonal issue to this. Sometimes they might help AI but it really is an unrelated issue to the question how well the AI actually plays given some set of rules to work with. Something might be gained by removing specific penalties from the factions because humans are generally better at understanding wide scale interactions, but in the end the AI is not behaving any smarter, it just has less penalties. But if you're just speedrunning, then AI modding isn't really needed of course.

With the array deletion behaviour, the performance only really matters if it is something that is performed in a tight loop. Bases get deleted so rarely it is really a non-issue in terms of performance but the changing id numbers probably caused some bugs in their implementation. If this was something that was instead performed in, say, a path finding routine, then the performance would quickly grind to a halt on the larger maps. In that case the routine would quickly exhibit worst case quadratic behaviour which is pretty bad.

[Modo]

"Base #5 was a nasty pink-dot drop over here in Gaia's face. The AI in SMAC isn't aware of such aggressive moves and won't degrade relations or declare war over it. I took advantage of the rule that a tile equidistant between two rival bases goes to the ownership of the newer base. This move stole several rainy tiles, including the highlighted one, with an energy bonus, where the Gaian former is currently working. I correctly anticipated what would happen: the Gaian former didn't stop its job but continued to finish that solar collector for me."

That is disgusting and you should wash yourself after making this play.

"it's been a good long while since my cockiness got so viciously punctured"

This is painful even when reading about it.
Also, you can gauge how to ICS correctly when the resolution increases from screenshot to screenshot.

[T-hawk]

Here's another, since this game is still just so amazing to keep playing: http://dos486.com/alpha/un/

[Modo]

"And you guys will hate to hear this, but... I gave up"

I love it. Doing these runs to find out the proper balancing so that you can reverse engineer the improvement for whatever future run might make use of the new findings is the whole point in my view.
And you're already comparing SMAC games like you did in Civ5 so that section of the site looks quite fertile now.