majors not playing by the same rules as human?

[misterprimus]

I should mention I was previously playing the regular version. I just noticed that there was an open beta and enabled it in steam, so it's possible some of the issues I mention below were addressed by it.

I've only played a few games, but I seem to be noticing a trend. The "easier" the planet (moon being hard, siwa being easy, Seth being inbetween, etc...), the less the divide between my level of development after a few turns and that of my major competitors. I also noticed a trend where most of the videos on YT show people playing on high water planets like Siwa or the one with hostile alien biology since I presume the AI majors can't avoid combat with minors and others.

On moon and planetoid, I am basically scraping by to survive while my competitor majors seem to be doing just fine, perhaps even attaining impossibly fast technology levels given any given round. So I have 2 questions:

1. Playing on normal difficulty, does the AI have certain advantages over me that make it relatively more difficult for me to win against them on "hard" planets like moon/planetoid? I initially started playing on these since I figured the AI being slowed down would help me more.

2. Does reducing research speed on game generation slow down AI research speed just as it slows down mine? I recall reading something about BPs being different for AI, so I was wondering about this.

Seems to me like some of the advantages the AI has (discussed in the manual) would be fairly minor on a Siwa planet, but quite significant on something like a moon. For example, getting surrounded by numerous independents early on wouldn't affect the AI majors much since their towns are not hurt by danger, but for a human player it could be devastating.

I was also considering a Robinson Crusoe and thinking about how to set it up so it's still challenging (since minors are mostly walk-overs once you upgrading your troops even a bit). Anyone have some input on that?

[BlueTemplar]

EDIT : You've already checked the manual, my bad.
The manual lists all the AI rules differences in section 5.18.
It also lists the various bonuses for the player and the AI in section 3.1.2.3 Difficulty Level, but doesn't tell if on Normal you are *actually* on a "level field" (for those things that can even be compared) with AI Majors (just that *you* don't get bonuses starting with Normal).

[zgrssd]

The unfortunate truth is that game AI can not plan ahead. They inherently life 100% in the current turn/boardstate/game tick.

There is no shortage of game mechanics that require a minimum amount of planing ahead, wich means:
- they are so trivial to do, we humans do not even notice we just did some planing
- they require so much planning ahead, there is no way for a AI to do it this side of a full on AGI

Simulating planing on the AI part is a lot of work.
Compensating for this total inability to plan - that is what "AI Cheats" are there for.
Given that this game is developed by a single person, he choose the simple route to use more AI cheats on those mechanics he could not create AI routines for on a reasonable timeframe.

[BlueTemplar]

I'm only aware of TWO developers that tried to make at least an average 4x AI :

1.) Arcen Games, for their AI War series... but there the AI is literally designed to not play from the same position of strength as you, AFAIK your whole goal is to not "piss it off" (otherwise it will crush you like a bug), and slowly increase your own power until you're finally ready to face it and win the game.
So that one doesn't *really* count (even if the AI is probably not bad, it's had to compare).

2.) Stardock, specifically for Galactic Civilizations 2.
They claim that their "normal" AI not only plays by the same rules as the player, but doesn't even get any bonuses.
(I personally find that hard to believe.)

ALL the other 4X developers I'm aware of went the economically sensible road of using a bad AI, because :
Most of their players are never going to play long enough the game to get good at it, *especially* for 4X where 100 hours is considered to be the rough moment when you stop being a newbie. (As a reminder, the median number of hours played for a median game is *way* under 20, though I'd still expect it to be close to that for a median 4X game ?)
So there's no point really of making an average AI and then have to weaken it compared to the median player.
It's much easier to make a bad AI and then give it economic bonuses, or - where it's not too jarring - make it play a different game from the player, where it can rely on its strengths, and not be too hobbled by its weaknesses !

[misterprimus]

ORIGINAL: BlueTemplar

I'm only aware of TWO developers that tried to make at least an average 4x AI :

1.) Arcen Games, for their AI War series... but there the AI is literally designed to not play from the same position of strength as you, AFAIK your whole goal is to not "piss it off" (otherwise it will crush you like a bug), and slowly increase your own power until you're finally ready to face it and win the game.
So that one doesn't *really* count (even if the AI is probably not bad, it's had to compare).

2.) Stardock, specifically for Galactic Civilizations 2.
They claim that their "normal" AI not only plays by the same rules as the player, but doesn't even get any bonuses.
(I personally find that hard to believe.)

ALL the other 4X developers I'm aware of went the economically sensible road of using a bad AI, because :
Most of their players are never going to play long enough the game to get good at it, *especially* for 4X where 100 hours is considered to be the rough moment when you stop being a newbie. (, though I'd still expect it to be close to that for a median 4X game ?)
So there's no point really of making an average AI and then have to weaken it compared to the median player.
It's much easier to make a bad AI and then give it economic bonuses, or - where it's not too jarring - make it play a different game from the player, where it can rely on its strengths, and not be too hobbled by its weaknesses !

Right, I'm aware of this. My question is: did the programmer adjust the AI's advantages down to compensate for the struggles of the player on something like a moon planet. For example, let's say the AI is stupid and specifically lacking in generating BPs the "natural" way. So he looked at what a typical Siwa human player would do in a typical game and how many BPs he would generate in a single turn; and then gave that amount of BP for free to the AI. So on a typical Siwa game, both the AI and I have Solders II by round 10.

On a moon map, I'm struggling to barely have enough food and water for my people to not starve, so BPs are a luxury early on. If the AI still gets Soldiers II by round 10, while I still only have militia, then essentially the programmer ends up overcompensating for the AI's stupidity and lack of planning. Perhaps it's the result of not enough testing going into a moon map, etc...

So that's my question. I see the AI blowing way past me on harsh planets like Seth, Planetoid, and Moon, and was wondering if this was the cause.

[BlueTemplar]

Well, my next game I'm planning to try a Moon map, so I'll get back to you once I do... [:'(]

The manual does say for Major AIs :

§ No Organizations
§ Does need to produce BP
§ BP is directly spent on
buying Techs, Models or
Formation Types

Furthermore the AI also has a number of disadvantages as it is not able to
pick up on a number of extras that players can (from Stratagems, Decisions
and stories) as well as missing out on some bonuses (like high Leader
Relation scores).