A look at scaled handicap

The way I understand the handicap right now, is that the AI receives certain bonuses to manu, tech, and money to help it remain competitive. I have no issues with handicaps, but what I often see with the strong handicaps is an alteration of the early game.

In Civ...this was the AI able to build all wonders. In Gal Civ, this would be a monstrously expanding AI or a crazy powerful early rush.

 

What I would like to try out for the AI, is a more scaled approach to handicaps. Start the AI on a roughly level playing field, but increase its bonuses over time (perhaps by tech phase, or simply by turn counter).

This would create a more organic approach, early game things are even. As the human uses more and more of its better gameplay to gain advantage the AI naturally counters with more bonuses. Further, as larger maps tend to leave to longer games, this would counteract some of the weakness the AI has on larger maps, which Frogboy listed as an area to attack.

 

 

39,675 views 16 replies
Reply #1 Top


What I would like to try out for the AI, is a more scaled approach to handicaps. Start the AI on a roughly level playing field, but increase its bonuses over time (perhaps by tech phase, or simply by turn counter).

 

Something like this could work IMO. But would need carefully adjusting as when and how bid the bonuses were, Starting at the same "level" as the played could badly    put the AI behind the 8 ball and impossible to catch up.  A extrastarting planet or two would be an advantage that would take time to build up,

Reply #2 Top

Quoting a0152570, reply 1



What I would like to try out for the AI, is a more scaled approach to handicaps. Start the AI on a roughly level playing field, but increase its bonuses over time (perhaps by tech phase, or simply by turn counter).

 

Something like this could work IMO. But would need carefully adjusting as when and how bid the bonuses were, Starting at the same "level" as the played could badly    put the AI behind the 8 ball and impossible to catch up.  A extra starting planet or two would be an advantage that would take time to build up,

Reply #3 Top

Had the same Idea, great to see that it is already posted.

Of course it would need a proper balance test. Maybe it would be possible to try it in the Steam "opt-in". Maybe it could be possible to make 3 AI difficulties for testing purposes, one with a high flat Bonus, one with a fast scaling Bonus and last but not least one with a low flat Bonus and a slow scaling Bonus.

I would really like to see some scaling Bonuses for the AI, because currently it is all about the early game, if I have a good start, the game is already won and becomes boring, unrewarding and a pure AI slaughter. The main reason for this in my opinion is that the amount of Player exploids are really adding up the longer the game goes on.

 

Reply #4 Top

Interesting concept. Really it could effectively get rid of difficulty levels for the most part. If the ai got penalties or bonuses in relation to the player as the game went on. However, that would potentially make the game insane as an empire with a couple of planets left started cranking out ships every turn. It would require a lot of work to balance this.

Reply #5 Top

Quoting Blaze, reply 4

Interesting concept. Really it could effectively get rid of difficulty levels for the most part. If the ai got penalties or bonuses in relation to the player as the game went on. However, that would potentially make the game insane as an empire with a couple of planets left started cranking out ships every turn. It would require a lot of work to balance this.

 

I think we have a missunderstanding here, I think that you think that the AI gets flexible Bonuses based on the strength of the Player. But I think the Idea is about a growing Bonus over time (turns). Currently we have a flat bonus system (I also think that).

Just as an example (the numbers are pure made up and maybe even the whole system works differently): Gifted AI gets 50% more total Manufacturing, genius gets 100% and godlike 200%.

The Problem with the flat system is that if the Bonus is high the AI is to strong in the early game and if it is low the AI is to weak in the lategame.

 

Because of that resons the flat Bonus system should be changed in my opinion.

The Idea would be to change the System to that (again just made up numbers): gifted 25% more total Manufacturing and + 0.1% every turn. genius 50% and +0.25% per turn and godlike 100% and +0.5% per turn

Or more extrem: gifted 0% more total Manufacturing +0.25% every turn, genius +0.5% per turn and Godlike +1.0% per turn.

 

I favor the System with both a flat and a turn Bonus, to give the AI an advance in the important early colonisation race (but not to much) and than later it still can be a strong opponent thanks to the "per turn" bonuses.

 

Sry if people get anoyed by my many "thinks" but I wanted to make clear that I dont have certainty about that things.

 

Hagen

 

 

Reply #6 Top

Quoting a0152570, reply 1


 Starting at the same "level" as the played could badly put the AI behind the 8 ball and impossible to catch up. 

 

This; the game rewards early expansion heavily. If the AI is struggling in the late-game, it's usually because it didn't do as well early on as it appears to have.

 

Tbh, the thing that will make the AI better late-game isn't any artful tweaking of the difficulty level bonuses; it's making it understand the adjacency system. The player does, and capitalizes on it. The AI doesn't, and so it's late-game planets are massively less productive.

Reply #7 Top

I think main AI bonuses that really makes difference are AI ability to see the whole map (on highest levels), randomly see part of it (on gifted) and under certain circumstances ignore FoW. Second most major bonus is ships range. On gifted it's 20% and it's a real deal together with open map. Other bonuses are more like counter balance to AI inability to specialize planets and use tile bonuses properly. Yeah, sometimes AI leads to overkill - f.e., Altarians with all bonus combined on bigger map with a few races (like 5-6) receive absolute unbeatable bonus to science due to their racial bonuses, AI bonuses and immense bonuses their receive from capping relics (and with open map they rush for them like there is no tomorrow). But i take all AI bonuses x2, if Devs teach it to scout properly instead of have "random insights" about where habitable planets are. 

Reply #8 Top

Quoting naselus, reply 6

Tbh, the thing that will make the AI better late-game isn't any artful tweaking of the difficulty level bonuses; it's making it understand the adjacency system. The player does, and capitalizes on it. The AI doesn't, and so it's late-game planets are massively less productive.

You are right but it is only a part of the full picture.

Because it is also about the ship design. A Player can really exploit it by using sensor ships (cargo ships with only sensors) to get a massive vision and no need for sensors on other ships. And then you can stack up research that reduces the room that ship parts need (60% with research, additionally 10% extra for weapons and armor and if a special event happens a malevolent race can get additional 10%). That means a Player ship can have 3,3 (5 with the event) times more weapons and armor and 2.5 (3.3 with the event) times more of everything else. With a fleet of those ships no AI Fleet can even scratch your ships and they can move 20+ tiles per turn, sometimes killing 5 or more fleets in one turn.

Yes the price of those Ships also increases by the same amount, but that doesn’t matter much, because like you already said, with the stacking of adjacency it is possible to reach insane production numbers (especially on malevolent home planets with the  +10 prod from a malevolent trait, +50 from the Death furnace, than +25% from another trait and then all production gets another 100% Military Bonus from another malevolent trait). And this things also buff each other (the death furnace bonus becomes 62,5 with the 25% trait and then 125 because of the military prod Bonus).

And the adjacency Bonuses can even exploited more, with the "logistic reduce building" and the "ship capacity increase building" and max military adjacency bonus for them you can get more than 50% more space on ships (and your ship parts only need 40% of the regular space, so the bonus is 125% in reality) and you can reduce the logistic needs of the your ships too, resulting in more than twice as large fleets).

If that is not enough than you build a Fleet with Lasers as their only weapons, a Fleet with rockets and one with Kinetic. Because you have so much movement points, the AI cant reach your fleets in one turn while you can decide which of your fleets attacks his fleets (Oh yes and you have endless vision thanks to the sensor ship).

Add this all up and the AI is screwed after the early game.

 

 

 

Reply #9 Top

One other note, a scaled bonus doesn't have to mean until infinity.

For example, it would be trivial to cap the max bonus at 300% or something...to ensure that the AI doesn't become unbeatable in the longest games.

Reply #10 Top

Quoting Stalker0, reply 9

One other note, a scaled bonus doesn't have to mean until infinity.

For example, it would be trivial to cap the max bonus at 300% or something...to ensure that the AI doesn't become unbeatable in the longest games.

100% agree

Reply #11 Top

Quoting naselus, reply 6

Tbh, the thing that will make the AI better late-game isn't any artful tweaking of the difficulty level bonuses; it's making it understand the adjacency system. The player does, and capitalizes on it. The AI doesn't, and so it's late-game planets are massively less productive.

i thought frogboy also said the ai cant adjust its production sliders on a per planet basis only on a global basis which means that for the ai to get anywhere (currently) it needs to generalize all its worlds which is a horrible horrible strategy once that gets fixed and the ai can adjust the economy wheel individually then it can specialize its worlds then it can start kicking ass

Reply #12 Top

So, in short, our main problem isn't AI bonuses, but the sad fact how its unable to anything meaningful but rushing to all available space without even "analyzing" that he's doing (like colonizing planets inside players ZoI that goes into rebellion in 4 turns). It can't properly scout and have to resort to tricks, it can't specialize, it can't use tiles placement and other related things properly, it can't even properly use techs to buff ships (in my game Altarians had like x3 more research progress than everyone else, but heir ships were total crap). The only thing it can do  is to get as much planets as possible and try to harass player with it's sheer potential from manufacturing (for evil races) or research (for "good" ones). And even that could be countered by spamming ships - looks like AI use your power rating and not your actual ships to decide if you're dangerous or not.

Reply #13 Top

Yes, sadly, it seems that whatever calculation is used, it looks like a poor one. Sort of like "We have 100,000 guys with spears, all they have is a couple of thousand." Never mind those few have machine guns, cannons and horses.

I would like to hope I am wrong, but so far quantity seem to count more than quality. I need to do more research/testing.

I would love to see the equation(s) and see what I can offer.
 

Reply #14 Top

My first reaction when I read this: I seem to remember a certain mod in another game with a 'flexible difficulty' challenge option once. I wonder if we might see something similar here? :grin:

Reply #15 Top

Quoting Blaze, reply 13

Yes, sadly, it seems that whatever calculation is used, it looks like a poor one. Sort of like "We have 100,000 guys with spears, cannons, and horses; all they have is a couple of thousand." Never mind those few have machine guns.

I would like to hope I am wrong, but so far quantity seem to count more than quality. I need to do more research/testing.

I would love to see the equation(s) and see what I can offer.
 

FTFY

Reply #16 Top

Alright, I've come around to agreeing that Handicaps are the easiest way to arrange the AI's production, so let's give this a shot.

 

Someone come up with ballpark estimates of what the bonus for Genius and Godlike AI's should have at T1, T50, T100 and T200 and we'll work out a decent formula to get there, then we'll see if we can throw it in a mod.

 

 

Currently we've got this for Genius:

<Stats>
<EffectType>ProductionPoints</EffectType>
<Scope>Global</Scope>
<Target>
<TargetType>Colony</TargetType>
</Target>
<BonusType>Multiplier</BonusType>
<Value>0.25</Value>
</Stats>

 

 

i.e. a permanent 25% boost, compounded by a -25% manufacturing cost multiplier.

 

 

And this for Godlike:

<Stats>
<EffectType>ProductionPoints</EffectType>
<Scope>Global</Scope>
<Target>
<TargetType>Colony</TargetType>
</Target>
<BonusType>Multiplier</BonusType>
<Value>3.0</Value>
</Stats>

 

So a 300% boost, compounded by a -50% manufacturing cost (i.e. they can produce 8x as much as a human) from the start, all the way through the game.

 

We'd have to find some way for the AI to itteratively adapt its multiplier each turn. It would basically be something like

 

<Stats>
<EffectType>ProductionPoints</EffectType>
<Scope>Global</Scope>
<Target>
<TargetType>Colony</TargetType>
</Target>
<BonusType>Multiplier</BonusType>
<Value>0.01</Value>
</Stats>

added on each turn.

I don't really know how we make it adapt that yet, but it could be tied to techs - we could make an AI version of the tech tree, where each tech increased its manufacturing multiplier or added a flat rate.

 

Anyone got any ideas for a more elegant solution?