Buffed AI for BNW

[glider1]

When you finish your next game and if you want to try this BuffedAI mod I'll upload the latest version. There has to be at a bit of interest from at least one CivFanatic to make the effort worth it. I like it because instead of giving the AI heaps of free techs etc, Buffed AI just gives the AI proper per-era bonuses and also makes combat a lot harder. I just tried it out and on prince difficulty the AI won science on turn 350 and I had to build a lot more troops to make conquest work. With the offensive bonus the AI gets, it's really nice for the first time to experience my frontline having to retreat because the advancing enemy frontline is actually too strong to attack. I'm normally an immortal player without this mod.

 

[Marshall Thomas]

What do you have the AIPerEraModifier set to? I know deity uses a -5 (I currently have it set to -4). The AIPerEraModifier seems like a useful rating, as most players seem to agree that the AI starts too strong compared to the player -then becomes relatively weaker as the game progresses through the eras.

I have mixed feeling about the disabling the victory conditions. I do think that they are somewhat arbitrary and often cheat the most balanced AI civ out of a victory. Also, the AI can't handle them nearly as well as the player; so yet another advantage over the AI. Yet, at the same time, a few civs and some game mechanics seem to be based around them; so turning them off may need to be balanced with more modding. What do you think of enabling/disabling victory conditions?

Also: it seems as though most players think that marathon mode gives the player an additional advantage over the AI (which is the last thing we need) because it encourages military capability. However, if we mod things so that the AI is stronger militarily, could the reverse be true? I personally have only ever played on epic speed, though I'm considering marathon; especially if the military AI is enhanced. What's your opinion of epic speed vs marathon speed?

Finally, I noticed that AI inflation can be modded. What do you think the results of making inflation for the player more severe would be? As mentioned above, the player seems to grow stronger through the eras compared to the AI. Wouldn't increased player inflation combat this?

Thanks in advance.

 

[glider1]

AIPerEraModifier is the biggest factor by far. -4 is immortal and is good for most experienced players. In BuffedAI mod AIPerEraModifier also boosts the AI's science output which makes it more competitive generally. BuffedAI then cuts the AI's other bonuses somewhat. If anything, BuffedAI has reversed the normal trend. The human starts out fractionally stronger and the AI very consistently pegs back and the endgame could actually favour the AI too much right now (there is a lot of testing needed with just me testing it!)

I went through exactly the same philosophical debate about disabling the victory conditions. The problem is the AI is just not competitive enough for them. Disabling all VC's except time means we get a game that is classic sandbox. The AI acts according to it's personality and how it sees things on the map. In my "victory_by_score" mod still in beta, the AI will play unleashed according to these things and same with the human. A massive religious spread and domination by anyone is scored appropriately at the end of the game.

I actually like both ways of playing but with VC's turned on, I need BuffedAI to make the AI competitive for them. It all boils down to how many turns does the AI typically win on VC's? In BuffedAI mod I try to adjust the AI's bonuses to push that time down to match what humans are doing it in.

I had exactly the same philosophical debate about marathon speed. It is a great way to play but it needs special adjustments. The problem is that without adjusting AI combat, the AI will loose it's units and cannot replace them. Deity marathon players think that the AI remains competitive in combat by massive carpet of doom attacks. In the BuffedAI mod still in beta, the AI's combat is compensated by offensive attack and this actually is the way to balance AI combat independent of game speed. It works at whatever speed you choose.

As for epic v marathon, standard speed is special for what it brings and marathon is special for what it brings. Epic is a halfway house that is a compromise if you can't mod the game to make marathon better which BuffedAI does.

On AI inflation, as it currently stands Firaxis have totally disabled it because it is badly implemented. There is a possibility of reintroducing it if I or WE think that the economic boosts the AI is getting are either too much or too little.

I've got to do another playthrough on BuffedAI today because I've changed it so that players can play on whatever skill level they are used to.

 

[Darkwave09]

I play on deity for the challenge but I'm bored of conquering the other civ for wonders. I really like the idea of your mod. I want to build some wonders too and avoid the carpet bomb of death. I too think the AI starts out too strong then gets progressively weaker as the game continues. I like the idea of the same happiness to start with the AI. If I loose to the AI I want to loose because of game play not AI buffs. If you need someone to help test Beta version, I will help

 

[glider1]

Hey Darkwave, I've uploaded the latest version of this mod at the first post. Try it out at Deity and see how you go! I think it should be roughly the same difficulty but humans have more creative options with what they can do.

Suggestions are welcome. I'm also working on other ideas so appreciate any help I can get.

 

[Aristos]

Man, we really need to start working on a BetterVAI... at least start with the TacticalAI. Holidays?

 

[glider1]

So true. I'm on Civ even it is just 1 turn a day so holidays are irrelevant. Next step is to start improving the AI's tactics removing the most glaring bugs. Then the AI buffs in this mod can be reduced.

The first issue to fix in my book are when AI units embark onto water and get slaughtered. An algorithm is needed to filter out the stupid embarkations from the sensible ones. An algorithm that covers most situations.

The two bugged situations are:
1 - coastal city where AI is attacking it from land and some units flank attack but do so via water (silly)
2 - AI units embarking onto one tile lakes when they could go round (silly).

If we can fix the water embarkation logical bugs, it would be a major achievement just if we did that and nothing else. The coding is the simple part, it's the thinking up of a good algorithm that works in most cases that is the challenge.

First step would be to draw up a flow chart of exactly how CvTacticalAI.cpp decides to move units when they are adjacent to water. If we can draw that, the better algorithm would probably be staring us in the face (that is usually my experience).

 

[Marshall Thomas]

I'm looking foward to starting a new game with this mod. Do you recommend using your "More War" mod with this one? Thanks in advance

 

[glider1]

Hi Marshall
Without the "War Mod"
-> The AI builds up war more gradually like in vanilla.
-> That could make this mod more difficult.

With the "War Mod"
-> the AI is more aggressive.
-> So I suspect that overall difficulty depends on many different factors
-> You can get away with more conquest which could also make it a bit more easy (but more fun).

This mod is currently set for:
AI offensive attack outside of friendly borders against humans = + 25% (including barbarians).
Human nerf to archery line attack = -25%

That is going to make combat harder so it might take some getting used to (especially the archery nerf for humans meaning you will have to bring in more balanced armies). If you think it should be turned down or up, let me know. Think of archery units as just softening up the enemy primarily that is their use.

The refreshing thing about this mod is that finally the AI's happiness and pop growth is the same as us humans but the AI is still a challenge, so enjoy! Play it on the same difficulty you currently use and see how it goes.

 

[glider1]

Released v7

 

[bc1]

First step would be to draw up a flow chart of exactly how CvTacticalAI.cpp decides to move units when they are adjacent to water. If we can draw that, the better algorithm would probably be staring us in the face (that is usually my experience).

This program is quite complicated, it will require significant time and effort to figure things out. I mean come on, the guy who wrote this knows his program better than anyone. And my own experience of "algorithms that stare at you in the face" is that they usually have some shortcoming which actually makes them worse than the original.
But hey, it's the part of Civ V that most requires fixing, I would like to try this if we can get a dedicated team going. Anyone else interested to embark on a major project like this ?

We would need:

• a place to host our work
• people fluent in C++ to analyse and document the existing code
• outstanding players (e.g. Tommynt) to help us understand what AI priorities should be
• system people to design new algorithms
• coders to implement them
• testers to validate

 

[Darkwave09]

Played this mod a bunch of times and works great!! I got wasted the first couple of times I played because I was wonderspaming too much. Now that I'm used to it I can actually built wonders on diety!! The best part of this mod is the reduced AI happiness. The comp doesn't go all city spam crazy. I love that part, the map fills up slowly with cities as civilizations advance, mine included. A small suggestion to fixing AI stupidity. Put melee units in front of range when attacking cities. Pull back units to heal when wounded by city bombardment, espicially siege units. I do the same thing when I attack. Keep up the great work, this is one the best Mods I've used. Thanks. IF you need help testing improved AI tactics I will help. I know nothing about programming but I'm willing to test

 

[glider1]

Thanks Darkwave how do you think deity on this mod compares with standard deity in difficulty any opinions? One of the balancing methods of this mod is that humans have to concentrate more on military expenses which naturally slows down their economy and science rate.

The other thing is that the difference between warlord and deity is much smaller on this mod so it is not for beginners.

 

[Darkwave09]

I have played this Mod quite a few times on Diety now. IMO the AI is harder to kill but less aggressive. It takes more hits to kill a AI unit but they bring less of them. On lesser difficulties this is not a problem, but on Diety I think it is. The AI is just not aggressive or powerful enough to challenge me. I have lots of time to build wonders and defend my empire until I'm ready to expand into their territories. After the first third of the game the AI is basically helpless against my armies. I like the idea of stronger units, but on Diety I should be quivering with fear from a DOW from my AI neighbors. On the lower difficulties absolutely no problems with AI agrresiveness. Emperor difficulty is perfect, but immortal I notice change, and Diety I notice a major change. IF I want to play a passive Dilpo or Sci victory, then no problems. But I don't want to on immortal and Diety. Also the science boost isn't enough on immortal or diety. I am able to keep pace with the AI on research times, but just barely. IMO a small change here is required on higher difficulties. AS I stated before I love the happiness factor of this MOD, the map fills slowly with cities from all civs. That is the best part. Maybe if you gave the AI reduced maintenance cost on their units they could build more at higher difficulties. Or reduce the amount of hammers required for AI units in relation to human players. A 10 or 20 % reduction in build time might help the AI after a DOW from human players. The strength of the AI units is fine, no problems there its just that I can sacrifice wonderspaming a little bit at the start of the game to build up enough defence to basically nullify the AI. I have no fear of a DOW from the AI at that point. At mid and late game the AI has no chance against me, they are basically sitting ducks.I feel like this MOD is almost complete just a few tweaks at higher difficulties. Keep up the great work, I am bid fan of this MOD

 

[glider1]

Thanks for the feedback Darkwave. Quite a few of the settings in this mod are still not adjusted for difficulty so I'll fix that. Agree with you pretty much about immortal/diety. The settings are conservative right now but it is easy to turn up the punishment levels that the AI can meter out! I too need the challenge to keep up the interest in this game.

All the best for 2014

 

[SgtCiv]

Trying this MOD out. I like what I've read so far. I can give some feed back also if you want. With the normal game I have started here lately just DOW on the closest AI's as soon as I meet them and I can still go almost the whole game without ever being attacked, CS do a better job of trying to attack and/or defend themselves.

 

[Darkwave09]

Good to hear. I will check back in a few days to see if have modified the higher difficulties

 

[glider1]

I have played this Mod quite a few times on Diety now. IMO the AI is harder to kill but less aggressive. It takes more hits to kill a AI unit but they bring less of them. On lesser difficulties this is not a problem, but on Diety I think it is. The AI is just not aggressive or powerful enough to challenge me. I have lots of time to build wonders and defend my empire until I'm ready to expand into their territories. After the first third of the game the AI is basically helpless against my armies. I like the idea of stronger units, but on Diety I should be quivering with fear from a DOW from my AI neighbors. On the lower difficulties absolutely no problems with AI agrresiveness. Emperor difficulty is perfect, but immortal I notice change, and Diety I notice a major change. IF I want to play a passive Dilpo or Sci victory, then no problems. But I don't want to on immortal and Diety. Also the science boost isn't enough on immortal or diety. I am able to keep pace with the AI on research times, but just barely. IMO a small change here is required on higher difficulties. AS I stated before I love the happiness factor of this MOD, the map fills slowly with cities from all civs. That is the best part. Maybe if you gave the AI reduced maintenance cost on their units they could build more at higher difficulties. Or reduce the amount of hammers required for AI units in relation to human players. A 10 or 20 % reduction in build time might help the AI after a DOW from human players. The strength of the AI units is fine, no problems there its just that I can sacrifice wonderspaming a little bit at the start of the game to build up enough defence to basically nullify the AI. I have no fear of a DOW from the AI at that point. At mid and late game the AI has no chance against me, they are basically sitting ducks.I feel like this MOD is almost complete just a few tweaks at higher difficulties. Keep up the great work, I am bid fan of this MOD

@SgtCiv, glad you are giving this mod a go.

@Darkwave
Thanks for the detailed feedback that does help tune this mod. I'm happy to tune it as much as anyone needs. I basically followed Darkwaves guide but couldn't resist increasing the AI offensive attack bonus a bit as well for my own games where I feel it needs it.

I didn't touch emperor or below, just immortal and deity. You should see that the AI will have slightly more troops, slightly stronger attack, slightly more likely to DOW, slightly more productive and slightly better science rates per era.

-> The things I want to avoid are carpet of dooms. I can prevent that by changing the settings, so keep an eye out for it.

-> also want to avoid too much of a catch up game. I want the challenge to feel consistent from era to era all the way to the end. I can easily tune each era to boost the AI where necessary, but would need more input on how much per era boosts to apply. I can apply boosts to both production and science independently.

-> also want to avoid over doing the AI's attack inclination towards humans. If we turn that up too much, the AI will attack humans with too little consideration of other potential AI enemies and make the game too easy again. Also, if it is turned up too much, artificial dog-piling on humans can start up as well (want to avoid that).

That said, v8 should be more of a challenge. I've seen the AI complete the science victory almost as quick as a human would. I think the settings are still conservative, but better the AI a bit too easy than too hard (kills the fun).

Enjoy, and keep the feedback coming. This mod is for people that don't mind being punished by an AI for the sake of creative game play, even if the AI is administering punishment by being buffed (not intelligence yet).

 

[SgtCiv]

I've been playing on emperor mostly using Thal's MOD's, but his MOD has went down a road that I can't play with - the AI doesn't build buildings in their cities. Plus some Civ's don't expand at all. So, I've been looking for something else. I will have to up my game to immortal or deity.
Thanks.

 

[Aristos]

I've been playing on emperor mostly using Thal's MOD's, but his MOD has went down a road that I can't play with - the AI doesn't build buildings in their cities. Plus some Civ's don't expand at all. So, I've been looking for something else. I will have to up my game to immortal or deity.
Thanks.

That is exactly what I would expect. I warned glider about this, not to spoil the mod/fun, but to "lower his expectations" in order to tempt him to go to the much harder, but more logical field: working the code. Don't take me wrong, I know glider and he is doing this because he likes the game and likes to make it better, and likes to share, but my view is that he is wasting a lot of time and his skills (and probably other's that are willing to join a different approach) by pursuing what basically is an AI cheat (and as the variables he is touching influence so much of the code, it is only inevitable to create huge imbalances that are, basically, impossible to solve).

In other words, I repeat myself:

Man, glider, we really need to start working on the code. The tactical AI file is the best place to begin. Let's do that; much more time consuming, much harder, but if there is a solution to challenge, it will come from a modified CvTacticalAI.cpp file, and not from changing some values only (which again, creates a cascade of side effects that most likely end up being worse than the original).

What do you say?