For a forum that prides itself on trying new things, for the love of all that is novel and interesting, please can we have a new QotM? - Krill

Create an account  

 
Master of Orion AI Analysis

This deep dive article into the Master of Orion AI was linked in my Livestream chat today. I'm in the process of reading through it and there's a lot of interesting information about how the MOO AI makes decisions. Well worth a read!

https://docs.google.com/document/d/1zDVp...yA2jB7vswk
Follow Sullla: Website | YouTube | Livestream | Twitter | Discord
Reply

Fun read, though I haven't played any MoO games myself to have a proper appreciation. The core messages that the AI needs to run fast above all else, and exist first to provide entertainment for the player rather than challenge and difficulty, do resonate with me. One of these days I'll have time to get back to work on my Fire Emblem-clone Python SRPG...

Reply

(January 30th, 2021, 11:00)El Grillo Wrote: Fun read, though I haven't played any MoO games myself to have a proper appreciation. The core messages that the AI needs to run fast above all else, and exist first to provide entertainment for the player rather than challenge and difficulty, do resonate with me. One of these days I'll have time to get back to work on my Fire Emblem-clone Python SRPG...

The MOO AI is very well-designed. While it has some notable hang-ups and holes (and the version most of us play isn't the base game, it's patched a lot of the gaps), but it's perfectly capable of steamrolling the player or being an appropriate nuisance. And considering the computing resources they had at the time, it's amazing that it functions as well as it does. Some of that is because some of the mechanics are designed to be simple to implement - missile bases in particular, but this is a game where you can customize ship construction. Ship construction definitely has mistakes but it still functions, and a lot of games don't have that because the AI just can't handle it. And of course on impossible, the AI definitely lives by the creed of "quantity is a quality of its own."

The "Erratic" personality is a brilliant decision in how it's implemented. The game tells you, as soon as you meet an Erratic AI, that it's erratic and you therefore cannot predict how it will handle diplomacy, because they'll get a "message from the gods" or such and decide it's time for you to die. That would be obnoxious if it was 100% of the time, but it's like 20% of the time. So many modern games give you precise-to-the-nth-degree modifiers, and there's a lot of fun in things being a little messier.

The biggest hole in the MOO AI IMO is its tendency to pull a fleet back the moment it captures a world. It'll capture a world in, say, 2402, and like 90% of the time by 2404 the fleet is gone, which makes it trivial to exploit and take worlds the AIs steal from one another or, rarely, the player. If that fleet could be "frozen" for even 5 turns, it would work a lot better (since settled worlds have missile bases, they're not always pushovers, but a newly captured world doesn't have those).
Reply

My favorite part is the one explaining how to compile using MSYS2. Is there a publicly available, already compiled version with Kyrub's patches? If not, maybe I will try to do it by myself.
Reply

I'm torn between paying attention to the actual AI stuff and just thanking the Invisible Pink Unicorn that I never had to program under the constraints those poor sods had to.

Oh, and I am the guy in the office that gets handed the 'C' projects. The youngsters don't know how to handle memory management without RAII. And I'm still horrified at what the MOO developers had to do.
It may have looked easy, but that is because it was done correctly - Brian Moore
Reply

That Strategy Guid Wrote:In MOO2 some players do optimize for the cost of the early scout, but in MOO1 seemingly nobody does that even for ships, and the design screen does display the cost there.

And it’s not like the player can even design the missile bases (may be OP, though, see Stars in Shadow).

Didn't Sirian and/or Sullla mention that in like their first games - the base-as-printed game has scouts cost 10, but you can make them cost 8? One of the minor but greatly appreciated QOL improvements of kyrub's patch was fixing it. I've known this for a decade and I'm not the only RBer.

I just thought that was funny.
Reply

(January 31st, 2021, 16:44)shallow_thought Wrote: I'm torn between paying attention to the actual AI stuff and just thanking the Invisible Pink Unicorn that I never had to program under the constraints those poor sods had to.

Oh, and I am the guy in the office that gets handed the 'C' projects. The youngsters don't know how to handle memory management without RAII. And I'm still horrified at what the MOO developers had to do.

The thing that makes the game work so well, is that they leveraged the heck out of those constraints. The planet management system, the combat system, even the diplomatic system - they all fall in the "restrictions breed creativity" camp.

And a lot of jobs would benefit from knowing when they should, and should not, take shortcuts like the distance formula (if dx > dy, dx + 1/2dy, otherwise dy + 1/2dx). "Quick, simple and wrong" is sometimes the right decision.
Reply

The AI in this game is very impressive overall can give a nice challenge in early and mid game but usually doesn't pose much of a threat in late game. I'm talking about 1.4m so some bugs like 32000 ships and obsolete fleets may have been fixed in 1oom.

First, AI-AI fronts are very static and player can easily win by conquering all planets from less powerful races while powerful AI races don't do the same. One reason for this is that if attacking AI wins a battle and doesn't destroy the colony, then missile bases are left intact and the following turns the attacking fleet combats the same bases again and again, losing more ships each turn until transports arrive and I wouldn't be surprised if those transports have to evade those bases too to win. Another reason maybe is how AI pulls back its fleet from conquered planets leading to AIs spending time grabbing planets back and forth although this affects player-AI dynamics even more than AI-AI.

Second, if AI has survived late game bugs and has fleets that can threaten player systems it sends them on one system missions only and conquering planets in this fashion is usually a very slow process. To be more of a threat, it should move fleets from one player system to another and also not pull back the next turn after conquering player planet, leaving it exposed to ground invasions even if player is badly behind in tech and even if player has no fleets.

Note, AI pulling back might have been original game design decision to allow player to have comebacks. Rarely, the AIs also trades good stuff like black hole generator for useless old techs which I think is a good thing when game is otherwise lost. However, I feel like little changes could really improve the late game.
Reply



Forum Jump: