Difficulty levels

[Seravy]

So yeah, the recent AI improvements definitely made the game harder, but it's very hard to judge how much harder it is. I'm going to need to play several dozen games which will take months before I can safely say the difficulty levels are balanced (or not). I'm going to stick this thread until we are satisfied with the difficulty levels.

The intended purpose of difficulty levels :
Easy - 100% win even for beginners unless intentionally trying to lose.
Normal - 100% win for experts, an enjoyable game with a high but not 100% chance to win for beginners.
Hard - 100% win for experts with a very strong wizard, very high chance to win for them otherwise, none to medium chance for beginners, depending on how badly and what wizard they play.
Extreme - High chance to win with a stronger wizard for experts, medium chance with an average wizard, no chance for beginners
Impossible - Low to medium chance to win with a very strong wizard, very low chance with an average one, pretty much no chance with a weak wizard unless with extreme luck on starting location and enemies.


These are the current settings :

AI abilities
The AI is unable to move before attacking on Normal and Easy.
The AI is unable to use "my unit is faster" and "my unit is invisible" stall tactics on Normal and Easy.

Possible ways of improvement :
The AI could increase the random factor of combat spell decisions on lower difficulties, in other words, play a little bit dumber. (only Easy? Up to Normal? Up to Hard?)


AI picks
Easy and Normal uses the default wizards without modification.
Hard gets 13 picks.
Extreme gets 14 picks.
Impossible gets 15 picks.
Impossible wizards have double chance to pick Warlord, Guardian, Alchemy and Tactician.

Possible ways of improvements
Maybe reduce the extra picks received on Hard and above? As is, there is no mode where the AI uses custom wizards that have the default 12 picks.
Maybe change the chance of selecting the "default" retort for each color template based on difficulty, so Extreme wizards pick the "best match" retorts more often than Hard wizards?

Diplomacy
1. On higher difficulty the AI players conduct positive diplomacy with each other more often.
The chance is (15-2*Difficulty level) per turn.
2. The roll for positive diplomacy between AI players receives a +2 to 5/difficulty level bonus (this was 3-9 in the previous versions) - this affects peace, pact, alliance and trade rolls.
3. The amount of relation penalty received when the player controls too many cities is higher on higher difficulty.
4. Negative relation changes between AI players are reduced by 10%/difficulty level and positive increased by this amount past turn 150.
5. Generic War declaration against the human player receives a -8 penalty per difficulty level.
6. Relation based war declaration against the human player does not happen if the human player is already in a number of wars equal to the difficulty level.
7. (new) There is a -10*Difficulty level penalty to all Treaty offer rolls the AI makes when deciding to offer a Wizard's Pact, Alliance or trade offer to the player. Peace Treaty is not affected. Pact offered due to overwhelming human military is not affected.
8. Relation has a higher effect on the AI's preference to use more curse type overland spells (small effect)

Possible ways of improvements
Probably no need after the last update but testing needed.

AI Resources
1. City Growth.
Easy : 75%, Normal 110%, Hard 130%, Extreme 140%, Impossible 150%.
2. Outpost Growth
Easy : 100%, Normal 125%, Hard 150%, Extreme 175%, Impossible 200%.
3. Production
Easy : 75%, Normal 120%, Hard 140%, Extreme 166%, Impossible 200%.
4. Gold from cities
Easy : 100%, Normal 100%, Hard 140%, Extreme 180%, Impossible 250%.
5. Power income (all sources)
Easy : 75%, Normal 130%, Hard 160%, Extreme 200%, Impossible 250%.
6. Research income (cities only)
Easy : 100%, Normal : 100%, Hard : 110%, Extreme : 120%, Impossible 130%.
7. Maintenance (mana, gold, food on everything)
Easy : 100%, Normal 60%, Hard : 50%, Extreme : 40%, Impossible 30%.
8. Overland spell casting cost
Easy : 100%, Normal 80%, Hard : 66%, Extreme : 50%, Impossible 40%.

Possible ways of improvements
The jump in city growth between normal and hard might be too large?
Lots of testing needed.

Other
Random events are disabled on Easy (even if enabled from the menu).
Events are more frequent on higher difficulty levels.
Fleeing is always successful on Easy
Neutral cities are allowed to grow up to 4+2*Difficulty size.
Frequency and strength of Rampaging Monsters
Frequency and strength of Raiders
How much neutrals prefer to target the human player's cities over other cities.

[zitro1987]

AI Abilities
Easy - No Stalling tactics (AI and neutrals), No move before attacking (AI and neutrals), Significantly higher random factor of combat spell decisions
Normal - No stalling tactics (AI and neutrals), they can move before attacking, slight random factor of combat spell decisions (not significant as this is a clear fun factor of game, just a nudge)
Hard - No stalling tactics (neutrals-only), best AI-decision making with spells
Extreme - Even neutrals can stall

On Stalling tactics and neutrals - It's odd how their air elemental is even afraid of hitting focus-magic nagas? It takes a lot of fun out of attempting lairs when the criteria to stall is so easy to experience. Please make the criteria to stall higher for neutrals or more weak-unit specific (for example, hitting slingers, but avoiding hitting paladins). I'd find it more fun if the neutral AI creatures attempt to cause 'careful damage' by targeting weakest units as opposed to just running away. Probably too hard to mod.

Picks -
Easy - 12 picks (differ significantly from 'best match', dumb wizard choices)
Normal - 12 picks (closer to 'best match')
Hard - 12 picks ('best match')
Extreme - 14 picks ('best match')
Impossible - 15 picks ('best match')

AI Resources

City growth seems reasonable, the maintenance bonuses on the other hand seem to steep, combined with the other bonuses. Maybe increase normal -- impossibly by 10%?

[Seravy]

It's odd how their air elemental is even afraid of hitting focus-magic nagas?

Air elementals have low health, single figure, and medicore stats for a single figure unit.
Focus Magic Nagas have 50% more total health, somewhat weaker total melee power and armor, but a powerful poison and a ranged attack, which will also include the poison bonus. So overall they are several times more powerful from the viewpoint of the AI. The only reason why Air Elementals aren't annihilated by them is because they are immune to both poison and ranged attacks, but these abilities are not counted in unit ratings. Poison is worth about 3 swords per figure, so a Focus Magic naga is worth as much as [(4+3+3+3)*3= 39 ] an Air Elemental would be if it had 39 swords. (This is a simplified calculation but shows it easily fullfills the requirement of the naga being 2x as powerful.)
It's unfortunate but the AI is too dumb to realize the Air Elemental is good against Nagas and I don't think I can change that.

Picks - sounds reasonable. Not sure if extreme should be 13 or 14.
Maintenance - I don't want to change this one, the AI is unable to comprehend maintenance and slow down production, and it also cannot sell buildings. It's also unable to cancel unneeded spells, and needs much more armies in play to remain a threat than the human player. In other words nothing related to maintenance was improved on the AI. I prefer to improve the balance by changing the gold, power and production.

[GermanJoey]

Adding channeler to the double-chance Impossible picks seems like a good idea to me, as both the in-battle mana savings as well as the upkeep savings are really useful for the AIs.

[Hadriex]

The city growth and outpost growth are a bit much. Same with the output of settlers. They fill the world very very quickly on extreme and impossible with metropolises.

Otherwise I'm pretty much satisfied with the difficulty settings. I mean, they'll need to be more testing for the diplomacy of course, but that aside things are pretty good I think.

[Nelphine]

I disagree about channeler. Consider on impossible 1 AI can already cast ~7 times as many spells overland as a human player, with the same mana income, and channeler is largely wasted. Even alchemy seems undesirable compared to other retorts when you consider the resource advantage. Add in that an impossible AI probably gets a much higher resource income than a human player, and it becomes even worse.

However, AI is still limited in combat. So, having high skill is important (archmage should be doubled) and high mana stores is important (I think it would be reasonable to have an AI build up 25k+ mana, and sit on it, by turn 200).

I would also suggest reducing the resource advantage in general. But I don't have enough experience with 3.x yet, so I can't be sure. But ideally I would think impossible should drop to 150% resource bonuses (or, casting 3 overland spells for every 1 the human casts given the same income). I just don't know if its smart enough to make this work yet. The BIG problem with the resource bonuses is that its too early game focused. Every spell matters early game, and is well balanced (except maybe focus magic) so no matter what the AI casts, its huge and with their resource advantage, they can often just crush you super early. But the later the game gets, the more there are 'bad' choices of spells to cast (or even troops to build), and so the massive resource advantage gets eaten up on useless things (although in combat there are very few bad choices).

So, I'd go with something like, each difficulty starts much lower on resource advantages than the have now (150% for impossible, and balance from there). Then each difficulty would increase that advantage based on time, up to a certain cap. Like impossible could increase by 1% per turn until it reached 300% on turn 150. Extreme might incease 2% every 3 turns until turn 150. Hard could increase 1% every two turns until turn 120.

Make the early game less frantic, but make sure they have the resources for bad choices in the end game.

[zitro1987]

I disagree about channeler. Consider on impossible 1 AI can already cast ~7 times as many spells overland as a human player, with the same mana income, and channeler is largely wasted. Even alchemy seems undesirable compared to other retorts when you consider the resource advantage. Add in that an impossible AI probably gets a much higher resource income than a human player, and it becomes even worse.

However, AI is still limited in combat. So, having high skill is important (archmage should be doubled) and high mana stores is important (I think it would be reasonable to have an AI build up 25k+ mana, and sit on it, by turn 200).

I would also suggest reducing the resource advantage in general. But I don't have enough experience with 3.x yet, so I can't be sure. But ideally I would think impossible should drop to 150% resource bonuses (or, casting 3 overland spells for every 1 the human casts given the same income). I just don't know if its smart enough to make this work yet.  The BIG problem with the resource bonuses is that its too early game focused. Every spell matters early game, and is well balanced (except maybe focus magic) so no matter what the AI casts, its huge and with their resource advantage, they can often just crush you super early. But the later the game gets, the more there are 'bad' choices of spells to cast (or even troops to build), and so the massive resource advantage gets eaten up on useless things (although in combat there are very few bad choices).

So, I'd go with something like, each difficulty starts much lower on resource advantages than the have now (150% for impossible, and balance from there). Then each difficulty would increase that advantage based on time, up to a certain cap. Like impossible could increase by 1% per turn until it reached 300% on turn 150. Extreme might incease 2% every 3 turns until turn 150. Hard could increase 1% every two turns until turn 120.

Make the early game less frantic, but make sure they have the resources for bad choices in the end game.

Great idea, a good alternative to the current version of 'impossible' could be a game that starts near/at the level of 'extreme' then add a few percent AI bonus per year (12 turns) until around turn 200 when these bonuses reach or slightly exceed current 'impossible' version's AI bonuses.

[Seravy]

I disagree about channeler. Consider on impossible 1 AI can already cast ~7 times as many spells overland as a human player, with the same mana income, and channeler is largely wasted. Even alchemy seems undesirable compared to other retorts when you consider the resource advantage. Add in that an impossible AI probably gets a much higher resource income than a human player, and it becomes even worse.

However, AI is still limited in combat. So, having high skill is important (archmage should be doubled) and high mana stores is important (I think it would be reasonable to have an AI build up 25k+ mana, and sit on it, by turn 200).

I would also suggest reducing the resource advantage in general. But I don't have enough experience with 3.x yet, so I can't be sure. But ideally I would think impossible should drop to 150% resource bonuses (or, casting 3 overland spells for every 1 the human casts given the same income). I just don't know if its smart enough to make this work yet.  The BIG problem with the resource bonuses is that its too early game focused. Every spell matters early game, and is well balanced (except maybe focus magic) so no matter what the AI casts, its huge and with their resource advantage, they can often just crush you super early. But the later the game gets, the more there are 'bad' choices of spells to cast (or even troops to build), and so the massive resource advantage gets eaten up on useless things (although in combat there are very few bad choices).

So, I'd go with something like, each difficulty starts much lower on resource advantages than the have now (150% for impossible, and balance from there). Then each difficulty would increase that advantage based on time, up to a certain cap. Like impossible could increase by 1% per turn until it reached 300% on turn 150. Extreme might incease 2% every 3 turns until turn 150. Hard could increase 1% every two turns until turn 120.

Make the early game less frantic, but make sure they have the resources for bad choices in the end game.

Great idea, a good alternative to the current version of 'impossible' could be a game that starts near/at the level of 'extreme' then add a few percent AI bonus per year (12 turns) until around turn 200 when these bonuses reach or slightly exceed current 'impossible' version's AI bonuses.

This idea is good but very hard to implement.
An even greater problem, this system clearly favors early game strategies and penalizes late game ones, which is undesirable. The AI is the most vulnerable in the early game. A stack of hell hounds can still take out an extreme wizard before turn 40 if played right.

[Nelphine]

Bleh. While you're right of course, I find that right now, early game strategies are encouraged, and I dislike them immensely. I suppose I should really just play on a lower difficulty for the style of game I like, except I find my late game is never challenged on hard.

So. I need to play more extreme games. New resolution! (I usually play one or two extreme games after a bout of hard or impossible - after a hard, I always win, and think, OK impossible it is. After an impossible I seem to get wrecked by a 9 stack of sprites and bears and go ugh same early game issue as impossible, and drop to hard.)

[Al_Al]

Random events are disabled on Easy (even if enabled from the menu).

Is this one necessary? It looks like it simply making Easy difficulty level unnecessarily boring even for the beginners.