Difficulty levels

[antdog]

I read a while back that the difficulty levels were reset so that Warlord is Prince, Prince is King and so on. Is this change still in effect?

 

[Thalassicus]

TBC wasn't really a difficulty above vanilla - my original post was an oversimplification of a somewhat complicated topic. Basically, in Civ 4 I played on Emperor difficulty setting. In Civ 5 with TBC, I still play on Emperor. I feel it's about the same challenge level as before, which was my goal. "Challenge" is different from "AI handicaps," however. How challenging a game is depends on two things:

1. Complexity
2. AI handicaps (difficulty setting)

TBC increased the complexity of Civ, so I suggested lowering the handicap to keep overall challenge the same. Firaxis then increased the complexity of vanilla by adding lots of this mod to vanilla, so they also suggested lowering the handicap. In other words, vanilla has caught up to the mod.


There's also another side to this:

• The AI cannot handle some of TBC's complex topics as well as the human (like Granary's local resource bonuses).
• Originally the AI handicap curve was irregular, with unequal jumps between difficulties.

I solved these two problems in March by doing this:

• Raised the curve to compensate for the AI's disadvantage when playing a complex game.
• Smoothed out the curve.

AI bonuses are technically part of "difficulty levels," and that is what I was discussing in the original post. I've copied it below.



=======================

These are the AI handicaps, from settler (1) up to deity (8).

 

[antdog]

thanks for the clarification.

 

[The_J]

Moderator Action: Moved to Thals main forum.

 

[timbuktu]

It is a bug when some AI have 5 cities at the beginnig of the game (4000BC) on emperor difficulty level? In consequence: around 2000BC these AI's enter into medieval age, around 800BC into renaissance. Sounds a bit unrealistic.
Secondly, around 1000BC usualy I get report on surplus of happinness and there is always a AI player which have about 50 whereas the second greatest figure is about 15. Irrationally huge gap.
These observations were made while playing on largest map, 12+ players and 24+ puppets, no technology victory.

 

[Arksa]

So Vanilla immortal is less handicaps than the mods emperor?

I made a topic about how I'm getting my ass handed to me on emperor with this mod but was able to cope in vanilla. I guess I need to go back to king after all .


And yeah on emperor I noticed China having 2 cities very close to start, both capital and the other with 5 pop, while my capital was growing in pop 3 with and I was making my second warrior or something...

 

[timbuktu]

I have to add more precise information to my previous post.
Actually, "at the beginning of the game" meant first dozen turns when I first met others civilizations. Only then one can check how many cities AI have. So don't know how many cities/settlers they have in 4000BC.
I used to play on King Lv. and encoutered the same issue. Once time I even managed to conquer the capcity of the AI with highest hapinness surpluss and I got about 40 happinness added. For more than half of them I could'nt find out where they came from, the rest had sources in buildings, wonders, new luxury resources, etc.

 

[Sneaks]

Ai starts with free extra settler workers scouts and warriors at high levels as well as different happiness scaling.

 

[Thalassicus]

As Sneaks mentioned, AIs get various bonuses at higher difficulty levels, and like Arksa said this mod provides more of a challenge than the vanilla game. On Emperor difficulty and up militaristic AIs start with an extra settler and a few other things to compensate for their dumb decision-making in battles. Peaceful AIs usually surpass them anyway.

 

[timbuktu]

Just played on King and there were at least one AI started the game with more than 1 settlers. Generally, I'm glad that it is not any bug but intended change.

 

[Thalassicus]

This is the full list of A.I. starting units/buildings. I know it's complicated, but I've done my best to tweak these things over the past 6 months so each AI personality has a decent chance of success:

Spoiler :

local isBold = (GameInfo.Leaders[player:GetLeaderType()].Boldness > 5)
-- 1 settler
-- 2 chieftain
-- 3 warlord
if handicapID >= 4 then -- prince
  M_InitUnitClass(player, "UNITCLASS_WORKER", plot)
  M_InitUnitClass(player, "UNITCLASS_ARCHER", plot)
end
if handicapID >= 5 then -- king
  if isCoastal then
    M_InitUnitClass(player, "UNITCLASS_WORKBOAT", plot)
    M_InitUnitClass(player, "UNITCLASS_TRIREME", plot)
  elseif isBold then
    M_InitUnitClass(player, "UNITCLASS_WORKER", plot)
    M_InitUnitClass(player, "UNITCLASS_ARCHER", plot)
  end
  capitalCity:SetNumRealBuilding(GetUniqueBuildingID(player, "BUILDINGCLASS_MARKET"), 1)
end
if handicapID >= 6 then -- emperor
  if isBold then
    M_InitUnitClass(player, "UNITCLASS_SETTLER", plot)
  end
  M_InitUnitClass(player, "UNITCLASS_ARCHER", plot)
end
if handicapID >= 7 then -- immortal
  if isCoastal then
    M_InitUnitClass(player, "UNITCLASS_WORKBOAT", plot)
    M_InitUnitClass(player, "UNITCLASS_TRIREME", plot)
  elseif isBold then
    M_InitUnitClass(player, "UNITCLASS_WORKER", plot)
    M_InitUnitClass(player, "UNITCLASS_ARCHER", plot)
  end
end
if handicapID >= 8 then -- deity
  if isBold then
    M_InitUnitClass(player, "UNITCLASS_CATAPULT", plot)
  else
    M_InitUnitClass(player, "UNITCLASS_ARCHER", plot)
  end
end

AIs also have other bonuses which depend on difficulty level, such as these on Emperor:

Spoiler :

050 StartingLocPercent
080 AdvancedStartPointsMod
008 HappinessDefault
100 NumCitiesUnhappinessMod
080 PopulationUnhappinessMod
000 Gold
000 GoldFreeUnits
100 ProductionFreeUnits
050 ProductionFreeUnitsPopulationPercent
002 ProductionFreeUnitsPerCity
100 RouteCostPercent
100 UnitCostPercent
100 BuildingCostPercent
100 ResearchPercent
100 PolicyPercent
100 ImprovementCostPercent
002 CityProductionNumOptionsConsidered
002 TechNumOptionsConsidered
002 PolicyNumOptionsConsidered
0-1 AttitudeChange
030 NoTechTradeModifier
025 BarbCampGold
000 BarbSpawnMod
010 BarbarianBonus
010 AIBarbarianBonus
030 EarliestBarbarianReleaseTurn
007 BarbarianLandTargetRange
018 BarbarianSeaTargetRange
000 AIStartingDefenseUnits
000 AIStartingWorkerUnits
000 AIStartingExploreUnits
100 AIDeclareWarProb
075 AIWorkRateModifier
100 AIUnhappinessPercent
080 AIGrowthPercent
057 AITrainPercent
100 AIWorldTrainPercent
057 AIConstructPercent
100 AIWorldConstructPercent
057 AICreatePercent
100 AIWorldCreatePercent
057 AIBuildingCostPercent
050 AIUnitCostPercent
040 AIUnitSupplyPercent
050 AIUnitUpgradePercent
<!--80 AIInflationPercent-->
0-4 AIPerEraModifier
160 AIAdvancedStartPercent
030 AIFreeXP
003 AIFreeXPPerEra
PROMOTION_HANDICAP_I AIFreePromotion

 

[manarod]

After playing many games i know many problems that the AI has and i would like to suggest and idea. You could use some script events activated at some point when the AI is doing very bad?
For example:

1 - If the AI as reached the renaissance and has only 3 or less cities make an event to give 1 or 2 settles, instead of hoping that the ai will build them

2 - At higher level at every finished reasearch make all Ai units upgraded automatically. In my current game at renaissance siam attack me with archer! (and he is more advanced)

3 - After a certain number of turns check if the ai as enought military units and give them some (for example 20 turns on normal speed. In long games the ai never build so much after the beginning)

I hope to has been clear

 

[Sneaks]

1. Doesn't work because the AI doesn't always want more cities. In fact, it is hard coded not to build more than X cities if going for culture victories

2. This is an issue that we cannot fix properly yet. Will need core .dll

3. This isn't an issue. At high levels, the AI is swamped with units.

 

[Ahriman]

I'd really hate to see "fake" fixes like this that arbitrarily gave AI's extra units or upgrades or cities just because they were losing. Handicap modifiers is one thing, but that is still a level playing field across the AIs, and no AI player is rewarded for playing poorly. Just giving free gifts is something else altogether. Also, I never see the AI not upgrading units it has the tech and gold to upgrade, and the AI is usually swimming in gold. If they attacked you with an archer, they probably didn't have Machinery yet. Very possible to reach Renaissance without Machinery.

I have no problem with sometimes some AIs being weak, particularly if they have lost land or wars. I think that models real history pretty well. Its not like the Aztecs or Incas or Iroquois or Zulu or Indians did a very good job resisting the Spanish or English.

 

[tlaurila]

My main beef with difficulty levels on civ all the way along have always been that AI bonuses are too start-heavy. It all falls down to a game where you have a huge challenge, closing towards the impossible, at the beginning. BUT if you make it past the initial bump in AI bonuses, the rest of the game will be scooting along for the inevitable victory. It has the effect of the latter part of the game being dull to the point of a chore.

So I'd be all for no (or minimal) minimal flat-out bonuses for AI to begin with (units, tech, gold, bonuses against barbarians, whathaveyou), but rather more scaling bonuses like +X% to {resource} produced. If possible, preferably even those bonuses could scale up as the game progresses in eras. Finally, AI bonuses to production and economy could be far bigger than science, so that the player isn't left to tech dust, but has to cope with bigger rivals.

 

[Thalassicus]

I agree with you completely tlaurila. The same thing has bothered me in a decade of playing Civ games and their spinoffs. I think the root problem is the human is much better at planning. We can look at the land around our initial capital, figure out where good city sites will be, and how the course of the game might change depending on our initial decisions. On the other hand, the AI thinks only one step at at a time. Where to place my next city?

I tried reducing initial bonuses and increasing long-term ones, but it ran into that problem. It makes the early game easy for the human player, and since the early game has an amplified effect on success as a whole, it makes late game easier too.

We can't solve the planning problem so we have to focus on its side-effects. The problem right now is I cannot find a way to get militaristic AIs to rush citystates instead of the human player. The AI would be much more successful if they did so. Unfortunately, when Firaxis added the "rush" strategy to AIs a few months ago they appear to have hardcoded it to only consider consider major civs as targets. Since the more vulnerable citystates are omitted, it fundamentally damages the AI's early game, and I've been struggling to find ways to compensate for this.

 

[Thalassicus]

Version 110 added a science bonus for AIs based on difficulty level, and reduced other AI bonuses.

Previously the AI production and gold bonuses were 25% per level. Now they're 20%, with a new AI research per-era modifier that ramps up over time. With a King-level AI:

v109
150% 100% : All Eras

v110
140% 100% : Ancient
140% 107% : Classical
140% 114% : Medieval
140% 121% : Renaissance
140% 128% : Industrial
140% 135% : Modern

In the future I plan on making the production bonus scale up as well, but it's more challenging from a technical perspective.

 

[Txurce]

Version 110 added a science bonus for AIs based on difficulty level, and reduced other AI bonuses.

Previously the AI production and gold bonuses were 25% per level. Now they're 20%, with a new AI research per-era modifier that ramps up over time. With a King-level AI:

v109
150% 100% : All Eras

v110
140% 100% : Ancient
140% 107% : Classical
140% 114% : Medieval
140% 121% : Renaissance
140% 128% : Industrial
140% 135% : Modern

In the future I plan on making the production bonus scale up as well, but it's more challenging from a technical perspective.

Is this a step-by-step approach to my "time release" proposal for AI bonuses? It looks like it - and once it's all in place, then it should be pretty easy to tinker with it.

 

[EsoEs]

Yeesh, 35% to Modern era science levels seems like a lot. I know people have been ragging on the "boring late game" but I'd rather people just bump up their difficulty level rather than buffing AI across the board.

 

[wobuffet]

Yeesh, 35% to Modern era science levels seems like a lot. I know people have been ragging on the "boring late game" but I'd rather people just bump up their difficulty level rather than buffing already strong 7-8 level AIs.

The thing is, the game becomes very boring once you're out-teching all the AI players, and the AI systematically lets that happen. I strongly support something like the time-dependent AI science bonus that Thal describes (until the AI can be modded to competence, that is).