How Difficulty Bonuses work for AI

[ElliotS]

Defending, yes. Intervening, no. Because you are small, you can't reach many city states, you can't send many trade routes, once someone declares war on you. Great people is your best weapon.
You can hit your neighbour and retreat, but if some other civ is snowballing farther, you can't do anything.

To be clear: Is your argument "Tall is stronger than Wide in player's hands but not without any weaknesses" or "Tall is weaker than Wide in player's hands."?

I agree with the first and think the second is totally incorrect.

 

[tu_79]

To be clear: Is your argument "Tall is stronger than Wide in player's hands but not without any weaknesses" or "Tall is weaker than Wide in player's hands."?

I agree with the first and think the second is totally incorrect.

The former.

 

[Gazebo]

Overall thanks a lot. While i was wrong about exact mechanics, i was right in that AI gets more bonuses from wide play, which explains why experienced players were asking for wide buff and you were observing wide snowballing.

Also here is the thing: Wide play, especially warmongering, should be stronger that Tall. Why? It requires investments, it is more risky than Tall. You build army and go to war, if you loose, you're screwed, because Tall was building infrastructure while you spent you hammers for army. So there should be higher reward to justify risks


This actually means "the more science AI has - the more bonuses it gets", which perfectly explains ridiculous runaways that happen from time to time. The whole idea leads to the situation that we have right now: every AI starts about the same, but by the time you reach Industrial - only 2 or 3 of them are competitive, all others are behind and have zero chances to catch up.

Maybe it should be replaced by flat bonus, like cost reduction? This will lead to more AIs remaining competitive in later stages of the game. Right now is rich gets richer, poor remains poor (or smarter).

In my most recent game i had Korea (that was actually a TALL Korea) that was 4th in science in Industrial , but then it took Rationalism and literally, without any exaggeration, researched 25 techs in LESS than 25 turns. I was checking it every turn, he was researching 1 tech per turn, sometimes 2 (probably because of Great Scientists). He slowed down only in Informational and started to research 1 tech every 2 or 3 turns.

You are making assumptions about my testing methods. I normally test balance on games with no AI bonuses at all. I then do test games with different levels of bonuses to see how they affect the game.

G

 

[Gazebo]

thats also a problem with yield %bonuses in general. thats literally all it does, if youre doing well the %bonus gives you a big bonus, if youre doing poorly it gives you garbage.

vanilla had %bonuses that were problematic, then CBP came along and remove virtually all of them, and then as time went on in VP they kept getting added back in. not just to wonders here and there but now to basic buildings and policies all over the place as well. its no wonder only 2-3 AI remain competitive at industrial and the others will never catch up no matter what resolutions get passed.

That’s a broad and baseless point, to be honest. The quantity and impact of % yields in VP is drastically lower than those of vanilla. The primary reason you see snowballing is that civ is a zero sum game and wide play + AI bonuses makes it easy for them to snowball.

G

 

[Deadstarre]

That’s a broad and baseless point, to be honest. G

i gave some supportive facts, like how % based bonuses means people with lower numbers get low bonuses and vice versa. if youre going to say that, you should back it up with something other than pointing to how bad vanilla was?

 

[Gazebo]

i gave some supportive facts, like how % based bonuses means people with lower numbers get low bonuses and vice versa. if youre going to say that, you should back it up with something other than pointing to how bad vanilla was?

You prefaced your argument with vanilla having a bunch and VP removing them. You made the comparison, not me.

G

 

[Gazebo]

So I'm going to do three big changes here:

Use @ElliotS 's exponential era scaler.
Incorporate Historic Events into the difficulty bonus system
Reduce the difficultymodifier value from 7 (at the highest) to 5 at the highest.

Thoughts?

 

[ElliotS]

So I'm going to do three big changes here:



Thoughts?

Sounds like a great point to start. Can't wait to try it.

Can we make all bonuses not apply on first city and remove free tech on Deity? (I think it's more elegant than giving a free tech.)

 

[Moi Magnus]

So I'm going to do three big changes here:
Thoughts?

I like it !
However, unless I misread ElliotS' post, I'm gonna protest on the use of the term "exponential scaling". It is a quadratic scaling.

 

[Gazebo]

Sounds like a great point to start. Can't wait to try it.

Can we make all bonuses not apply on first city and remove free tech on Deity? (I think it's more elegant than giving a free tech.)

Yeah. that's fine by me.

G

 

[LukaSlovenia29]

I like 1. and 3., and as for 2. -> does this mean ALL historic events or just some? What about historic events that are already included (e.g.: winning a war), they won't count double, right? If all historic events, I think that will further the snowballing effect (because the strongest civs have the most historic events).

Would it be possible for the AI to gain the difficulty bonuses when something exceptionally bad happens to it (e.g.: upon losing a city, being declared war upon, becoming a vassal), to help the worse AIs stay afloat in the game?

Or another idea, if not too hard to implement -> would it be possible to factor the score:average score ration of the AI into how much of difficulty bonuses it gets? I.e. the bigger the AI runaway, less (fewer?) bonuses it gets, and vice-versa.

 

[Gazebo]

I like 1. and 3., and as for 2. -> does this mean ALL historic events or just some? What about historic events that are already included (e.g.: winning a war), they won't count double, right? If all historic events, I think that will further the snowballing effect (because the strongest civs have the most historic events).

Would it be possible for the AI to gain the difficulty bonuses when something exceptionally bad happens to it (e.g.: upon losing a city, being declared war upon, becoming a vassal), to help the worse AIs stay afloat in the game?

Or another idea, if not too hard to implement -> would it be possible to factor the score:average score ration of the AI into how much of difficulty bonuses it gets? I.e. the bigger the AI runaway, less (fewer?) bonuses it gets, and vice-versa.

I dislike rubber-banding very much so. Blue shell PTSD.

It wouldn't duplicate it, I'd just flesh out the current set by adding GP birth, wonder construction, and archaeology.

G

 

[ElliotS]

I like it !
However, unless I misread ElliotS' post, I'm gonna protest on the use of the term "exponential scaling". It is a quadratic scaling.

>.>

You're right mathematically, but not so much when it comes to communicating the concept.

Currently it scales linearly. That's well understood. If the patch said "Changed AI bonuses from Linear scaling to Exponential scaling" people would imagine an increase similar to what my formula spits out. (Unless they assumed it was imbalanced and would be spitting out 1000 in renaissance and started doomsaying.)

If the patch said "Changed AI bonuses from linear scaling to quadratic scaling" many people would be confused.

At least that's my take as Marketing Director.

 

[ElliotS]

I dislike rubber-banding very much so. Blue shell PTSD.

It wouldn't duplicate it, I'd just flesh out the current set by adding GP birth, wonder construction, and archaeology.

G

I think that Great Person births (Or even specific ones like Musician/Artist/Writer and Scientist) would be better than all historic events as a starting point.

I don't think the AI needs more bonuses for getting wonders.

Archeology could be good though.

 

[LukaSlovenia29]

Thanks for the reply.

 

[ashendashin]

I'm liking where this is going. Maybe I'll finally succeed in forcing myself into a standard speed Deity game with the new scale. Been trying to recently without Enginseers yield buffing mods and I just get angry at the classical/medieval AI entries...

 

[tu_79]

I foresee a couple of changes in future releases, regarding AI handicap scaling, but it's a step in the right direction.

 

[CrazyG]

Great direction, I'm looking forward to playtesting some of the changes discussed so far.

 

[Owlbebach]

I also agree, this is a very good move!

 

[LukaSlovenia29]

@Gazebo , while the new version is being prepared, I'd like to play Deity on the current one without AI starting with pottery. If I delete the "AIFreeTechs" part of the code in one of the files, will that change anything else or just remove pottery from AI start?