BNW Deity Tier List

...Civ III used to have an exploitable bug that gave you double resources for sacrificing population, Civ IV you used to be able to steal workers and chop forests. I don't think I would have been able to beat any of them on the highest difficulty without exploits. I couldn't find any easy exploits in Civ V outside of the AI being atrocious at war...

Thank you for the quick reminder why I was so disappointed with III and IV, and why I think V is so much better. Since II, I just played the game as it was laid out by the developer, not looking for exploits. II could be won straight-up, at the highest difficulty setting, without caravans or other exploits. SMAC as well. That was not true for III and IV, and players really did need to chase exploits. I hated that, and V is much better in this regard. Note that the AIs in III and IV were also atrocious at war, it is just that stacks-of-doom masked the defects.

Coming back to this tier list, science being so much more important than religion or production, it makes perfect sense that Celts and Rome don't make the cut. But do check out consentient's list as he lays out a methodology for numerical ranking. Other tier lists, including this one, end up being much subjective.
 
Thank you for the quick reminder why I was so disappointed with III and IV, and why I think V is so much better. Since II, I just played the game as it was laid out by the developer, not looking for exploits. II could be won straight-up, at the highest difficulty setting, without caravans or other exploits. SMAC as well. That was not true for III and IV, and players really did need to chase exploits. I hated that, and V is much better in this regard. Note that the AIs in III and IV were also atrocious at war, it is just that stacks-of-doom masked the defects.

Coming back to this tier list, science being so much more important than religion or production, it makes perfect sense that Celts and Rome don't make the cut. But do check out consentient's list as he lays out a methodology for numerical ranking. Other tier lists, including this one, end up being much subjective.

I think Civ II could probably have been won without the Caravan -> Wonder exploit, but there were so many wonders that felt like must haves I couldn't bear to keep playing if I missed The Pyramids/Michelangelo's Chapel/Leonardo's Workshop/Adam Smith's Trading Company/etc/etc/etc. The power level for Wonders in Civ II was insane.

Personally I wish they'd make the AI a little better on the higher difficulty levels but tone down the production bonuses. I want to see the AI doing better because they make better choices, not because the bonuses let them do well regardless of what choices they make.
 
Personally I wish they'd make the AI a little better on the higher difficulty levels but tone down the production bonuses. I want to see the AI doing better because they make better choices, not because the bonuses let them do well regardless of what choices they make.

We all wish that. But they do not know how to program the AI to play significantly better. So the production bonuses and happy penalty has to fill in. It works well enough, and the tradition goes back to II at least.
 
Attila is not God tier??? They are 100% unstoppable if you start next to them. Horse archers dominate everything until xbows.

France and Brazil are both terrible and should be much lower. France is basically a blank civ with a musket UU and Brazil gets horrible jungle starts with no production.

Venice is so bad they were deleted from NQ games. Spain is not considered God mode? This list has some glaring issues IMO.
 
We all wish that. But they do not know how to program the AI to play significantly better. So the production bonuses and happy penalty has to fill in. It works well enough, and the tradition goes back to II at least.

That tradition goes back to Civ I, that's why Chariots were so powerful. The main difficulty in winning on Emperor in Civ I is how quickly your towns reach their happiness limit. So the exploit was playing on Pangaea, researching The Wheel as your first tech and cranking out Chariots. Every city you captured you set to producing Chariots. Not only did this strategy make happiness penalties almost completely irrelevant, they let you win the game in BC which gave you a huge score. There were a lot of silly exploits in Civ I, but this was the best one.

As for making the AI better I'm sure they could do it, I think they don't because most people enjoy winning against impossible numbers by 'out-thinking' the AI.
 
Well even if they spend more time with the AI it is irrealistic to imagine they'd do different codings for different difficulties. If anything Prince would get harder and either the bonuses would be reduced or all difficulties would get harder.

Still AI programming is a job of patches, fixing and improving as players play the game. Balance too.
And I'm 100% convinced difficulty vs an AI is both a matter of AI and balance. The AI does not know the game is imbalanced and that some options are noob traps. It is never programmed this way. And while improvements could be made, an AI not taking Rationalism is still crippling itself. Even if its decision made sense because of how flavors work. That's where balance come, so that when an AI decides on a strategy at least this strategy works as described or as intended by the developper. And then I end up failing a lot against the deity AI of my balance mod as a result ><
The AI programmer has the task to try to make the AI "think" and in the case of civ5 they opted for an AI that follows flavors (This is good for gold, this is good for science etc) but it's the job of the designer to make sure your options are valid choices.

Or you could do it like an AI mod I saw on Steam... where the guy just picked Artificial Unintelligence and put all flavors to 999 for key techs, buildings and policies. So that the AI just beeline the same stuff everygame. Challenging yes but boring.

Sadly Firaxis has never been a very frequent patcher and Civ5 balance and AI are probably set as is now...
 
And then I end up failing a lot against the deity AI of my balance mod as a result ><

Really? I read your mod changes superficially and they don't seem to be so big.

Can you explain in few sentences what makes deity so much harder in your mod?
 
Sorry, but it is very naive to believe such a thing.

They could definitely make the AI a lot more intelligent if they wished to do so. The problem is that it would cost significant resources being committed to programming the AI.

They would rather commit those resources to things which create revenue more efficiently such as sparkly graphics and DLC's for people to buy. Notice how they chose to create DLC's and expansions while ignoring the buggy MP code for years and years.

They opt to sell us more content for a broken/flawed game rather than fix the game's flaws.

Remember, they are a company and their actions have to make sense financially otherwise they will not exist anymore.

No matter how awesome they make the game it is going to have a small audience compared to games like COD, Halo, SC2 and LOL. They have to work with the audience they have and commit resources according to estimated sales and revenue. The turn based strategy audience is simply smaller.
 
They could definitely make the AI a lot more intelligent if they wished to do so.

It is hard problem in computer science. They could do a few things, but having a computer play like a human is surprisingly close to creating a general artificial intelligence. For something like chess, you can simulate intelligence well enough through recursive look-ahead, but that would be too computationally intensive for a game like civ.

Buffing the computer gets prolly >75% the effect, and maybe .001% as much work. It won’t be changing anytime soon.
 
It is hard problem in computer science. They could do a few things, but having a computer play like a human is surprisingly close to creating a general artificial intelligence. For something like chess, you can simulate intelligence well enough through recursive look-ahead, but that would be too computationally intensive for a game like civ.

Buffing the computer gets prolly >75% the effect, and maybe .001% as much work. It won’t be changing anytime soon.

At the very least they could get it to recognize obstacles and understand the difference between a defensive and defenseless city to attack. Avoiding bottlenecks and single file attacks shouldn't be that difficult.

The AI in SC2 is significantly better at avoiding bottlenecks and single file attacks than the AI in SC1 was. The SC2 AI even recognizes the need to make air units when it can't attack from the ground. Although it's obviously not as quick to realize these things as a human would be.

It has to throw some ground units at the problem first before it realizes that it should make air units. They may have applied some sort of trial and error approach.
 
Really? I read your mod changes superficially and they don't seem to be so big.

Can you explain in few sentences what makes deity so much harder in your mod?

The nerfs to the science curve hurt humans more than the AI, NC and universities strength reduced, rationalism at industrial era, reduced secularism, fixed GS bulbs.
The buffs to suboptimal paths help the AI.
The change in how the tech penalty for cities works is currently set to be more favorable at Deity for the AI than humans.
Melee units are scarrier.
The AI can difficultly be bribed away.
The AI no longer always accept white peace.

Really you could give V2 a try, it will be released in 2 days, I'd appreciate your playthroughs and reports in the mod thread. Not saying you necessarily will always feel trouble but you should feel less at ease than the base game I'd think.
 
This is the biggest issue with Civ:
"Or you could do it like an AI mod I saw on Steam... where the guy just picked Artificial Unintelligence and put all flavors to 999 for key techs, buildings and policies. So that the AI just beeline the same stuff everygame. Challenging yes but boring."

There are just many things that aren't really balanced and in many cases could be considered even broken. This limits options and forces any good player to make many of the same decisions such as beelining certain techs and almost always choosing rationalism. Being able to make an AI better by just hardcoding what to beeline for every game means you have some serious balance issues. The decision should be based on your strategy, starting location, civilization, map, etc and various paths should be stronger based on those variables. I would much rather see them spend resources balancing the game before improving the AI.

Don't get me wrong the AI in Civ5 is terrible but spending time improving it in a game that needs balance fixes to me is a waste of time. I'm actually very interested in AI game development and have been working on an AI for TripleA (Axis & Allies type games). Its simpler than Civ5 since its just a war game but the AI does a much better job of evaluating its options on what to purchase, attack, defend, etc.
 
At the very least they could get it to recognize obstacles and understand the difference between a defensive and defenseless city to attack. Avoiding bottlenecks and single file attacks shouldn't be that difficult.

The AI in SC2 is significantly better at avoiding bottlenecks and single file attacks than the AI in SC1 was. The SC2 AI even recognizes the need to make air units when it can't attack from the ground. Although it's obviously not as quick to realize these things as a human would be.

It has to throw some ground units at the problem first before it realizes that it should make air units. They may have applied some sort of trial and error approach.

SC1 had its ups and downs, but its most impressive AI challenge yet was Omega and the last level with Zerg. I had a friend clear this using cheat codes.

Moderator Action:
General Thread warning:

We seem to have gone way off topic. The topic of this thread is the BNW Deity Tier List.

If you wish to further discuss mods that change the AI of Civ V, please do so in the Creation & Customization forum.
If you wish to discuss the AIs behavior in other games, please do so in the forum for that particular game or the catch all "All Other Games".
 
Really? I read your mod changes superficially and they don't seem to be so big.

Can you explain in few sentences what makes deity so much harder in your mod?

For me, the most important thing that kept me alive in the case of early invasion was the fact that Melee units were useless against cities. Now that cities have been weakened and melee units buffed, an AI can easily storm the cap as I can't possibly have enough units for defense in time.
 
I read your mod changes superficially and they don't seem to be so big.

Try playing it. It's much harder. For me, Level 6 (which Acken calls Immortal) is harder by the midgame than Deity. I've not played a full game on Level 8 but I can imagine its Deity++.

Not saying you necessarily will always feel trouble but you should feel less at ease than the base game I'd think.

As good as he appears to be, even IronFighterXXX will have some time adjusting, I'm sure.

For me, the most important thing that kept me alive in the case of early invasion was the fact that Melee units were useless against cities. Now that cities have been weakened and melee units buffed, an AI can easily storm the cap as I can't possibly have enough units for defense in time.

Right, and if you want to take the fight to them, Warrior rush is now back on the table! ;)
 
Well even if they spend more time with the AI it is irrealistic to imagine they'd do different codings for different difficulties. If anything Prince would get harder and either the bonuses would be reduced or all difficulties would get harder.

Still AI programming is a job of patches, fixing and improving as players play the game. Balance too.
And I'm 100% convinced difficulty vs an AI is both a matter of AI and balance. The AI does not know the game is imbalanced and that some options are noob traps. It is never programmed this way. And while improvements could be made, an AI not taking Rationalism is still crippling itself. Even if its decision made sense because of how flavors work. That's where balance come, so that when an AI decides on a strategy at least this strategy works as described or as intended by the developper. And then I end up failing a lot against the deity AI of my balance mod as a result ><
The AI programmer has the task to try to make the AI "think" and in the case of civ5 they opted for an AI that follows flavors (This is good for gold, this is good for science etc) but it's the job of the designer to make sure your options are valid choices.

Or you could do it like an AI mod I saw on Steam... where the guy just picked Artificial Unintelligence and put all flavors to 999 for key techs, buildings and policies. So that the AI just beeline the same stuff everygame. Challenging yes but boring.

Sadly Firaxis has never been a very frequent patcher and Civ5 balance and AI are probably set as is now...

Actually modern day research has reached quite far in the artificial learning. With neurological networks one can achieve AI's that learn from their previous mistakes and actually adapt & change.

Using that an AI could change between different runs - where it learns not to fall in the same noob trap over and over again. (The simple rule is: "make a small change in the build pattern", is the outcome better: "yes" keep the change, "no" do a change the other way round and test it). Biggest problem is: you need thousands of iterations before you get anywhere near stable. It could however nowadays be done by using online gameplay.

It actually *is* already being incorporated in at least one game type. Chess: modern day chess engines actually learn from their mistakes.

Moderator Action:
Please heed above mod text:
If you wish to further discuss mods that change the AI of Civ V, please do so in the Creation & Customization forum.
If you wish to discuss the AIs behavior in other games, please do so in the forum for that particular game or the catch all "All Other Games".

Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889
 
Using that an AI could change between different runs - where it learns not to fall in the same noob trap over and over again. (The simple rule is: "make a small change in the build pattern", is the outcome better: "yes" keep the change, "no" do a change the other way round and test it). Biggest problem is: you need thousands of iterations before you get anywhere near stable. It could however nowadays be done by using online gameplay.

That would pretty awesome! I certainly carry (irrational) grudges from game-to-game so it would be great if the AIs were learning to play better overtime. But I am not convinced all the decision points could be tracked. Ratio of melee to ranged maybe? How many units are needed before DoWing? How does the AI know when it is playing against a good player? Things like don't-throw-units-in-the-water seem better to code. Humans pick up on those points pretty quick, but how would the AI learning be programmed to notice? If the developer appreciated the emergent AI behavior ahead of time they would have avoided it in the first place. So you have to know something needs to be tracked in order to learn if it is good or bad -- but if you knew it needed to be tracked you would not have the initial problem!

But it is not like the developers programmed in the noob traps on purpose -- players spotted them after release. But with all the buffs, they do not much hurt the AI.

Or you could do it like an AI mod I saw on Steam... where the guy just picked Artificial Unintelligence and put all flavors to 999 for key techs, buildings and policies. So that the AI just beeline the same stuff everygame. Challenging yes but boring.

Fair point. I think I prefer the variability.
 
Actually modern day research has reached quite far in the artificial learning. With neurological networks one can achieve AI's that learn from their previous mistakes and actually adapt & change.

This sounds pretty cool but we are pretty far from implementing this in a complex game like civ. There would have to be a huge number of neural networks for each decision, and this will also require a lot of processing, so a pretty expensive computer. For now it is a lot safer and economically viable to stick to the classical algorithms like Minimax.

Using that an AI could change between different runs - where it learns not to fall in the same noob trap over and over again. (The simple rule is: "make a small change in the build pattern", is the outcome better: "yes" keep the change, "no" do a change the other way round and test it). Biggest problem is: you need thousands of iterations before you get anywhere near stable. It could however nowadays be done by using online gameplay.

This sounds more like a genetic algorithm, but it will take a huge amount of iterations (not thousands, more like millions) to make it work. By doing it online, the gameplay would have to be uploaded to a central server where it is used to train the AI and then periodically patches with the new AI are released. I don't know how people will react to the thought that their games are recorded and then uploaded, many are already complaining a lot about using steam online, so that would certainly freak out many people.
 
Top Bottom