Thoughts on A/B/C bonuses

[Milae]

Bonuses I'm referring to are explained at the bottom of this page:
https://civ-5-cbp.fandom.com/wiki/AI_and_Difficulty
Long story short AIs get flat bonuses to all yields which scale with era when they hit certain conditions (mostly founding new cities and entering new era)

Basically I don't like them and I especially don't like the production element of them.
My reasons:
1) Feels like the AIs are in some sense playing a similar game to me but then also have a whole separate game on top which rewards them for completely different things such as founding rubbish cities everywhere just for the bonus yields.
2) AIs have weird power spikes at the same time of every game. And as with most bonuses these seem to have the most effect on the early game.
3) There is little variance between Civs as they will all have any weaknesses plugged by these bonuses. For example a scientific Civ with no emphasis on culture will still have many policies and I think this is a big reason why most of the more powerful AIs are the ones with extra science.
4) Giving the same yield to all different yields seems quite strange as they are not all equal in worth. As mentioned before I think the Culture they get from these is very significant and I would also say that the food they get is very high too as all AI cities tend to have high pop even if they have no real food in the city.

Quite a few of these problems could perhaps be reduced if rather than a flat yield bonus they received a bonus based on their current output so like x turns extra of science/culture/food in each city.

Now why I have a big issue with production:
1) A big part of the AIs bonuses are around reducing production costs already so on high difficulties they are often swimming in production after a while. This means the tech leader will often build every single wonder as the production cost is negligible to them. Also many AIs are constantly producing wealth/Science/culture on cities later on as they have nothing to build.
2) A lot of the times these production bonuses will end up going to waste overflow as they are much larger than the cost of many things.
3) However sometimes they will come at the perfect time and will all go into say a world project which gives them a free win out of nowhere. This happens pretty often to at least one AI so winning world projects is often impossible even if you are far ahead.
4) Also will pretty often complete Wonders for the AIs. This is very frustrating as when you know you are competing for a wonder you can try to use spies or look at the map to see if other civs are building it to make sure you will get it first, but then they just randomly complete it. I know great engineers can do this too but from what i've seen this is much less common and at least they actually had to do something (which I am also able to do) to get it.

I honestly believe just removing the production bonus wouldn't be a huge nerf for AIs as they already get so much cost reduction.

I do know that AIs would need to be buffed in other ways to compensate for any reduction in their bonuses as overall I don't think AIs are too strong right now so am happy to discuss ways that AIs could be buffed in a new thread or is there already one for this?

Also do you guys actually agree with me on this?

 

[Melchizedek]

I am quite new to using this mod, but I think you have laid out a strong argument against these bonuses. Maybe the AI should get smoother bonuses that are multipliers of existing yields, or reductions of cost, like you suggested.

 

[andersw]

So what difficulty are we talking about?

From what I understand the AI needs the production bonus to be somewhat competetive.
And a techer will always grab a good amount of wonders not just because of bonus but also because that player/AI can begin producing a wonder before others.
Several techs also include buildings which improves hammers, fast tech can also help to get production going.
I dont feel its that extreme, I'm more concerned about the AI's not perfoming at all, what I mean is that the playing field tends to be a bit more uneven than I would like it.

Some games I get extreme AI's (on emperor) but "every single wonder" ... just no, that was high difficulty vanilla runaway AI, I get AIs who build a lot of wonders but there is often some explanation behind it (america with Council of elders in current game for example).
The AI bonuses are way less than vanilla.

I do get frustrated about warmonger penalty which too often makes my units garbage, and that is even worse on immortal/diety.

 

[vyyt]

I don't know much about the inner workings, but I quite like the way things seem to work on surface (on emperor/epic), so I voted "Like them", but if sudden bonuses cause AIs building wonders in 1 turn or winning world projects without actually having the production, then I suppose spreading the production bonus (= smaller amounts but more frequent) or just giving a flat discount to buildings would be ok.

I am not against tweaking the A/B/C bonuses, spreading them more throughout the game or creating some exceptions, but I don't to see them go altogether. How else would you create different difficulties.

By the way, the scaling of AI bonuses for different speeds is being looked at for the next version.

 

[Recursive]

By the way, the scaling of AI bonuses for different speeds is being looked at for the next version.

To clarify this statement, the AI's A/B/C bonuses currently do not scale with game speed, making them stronger on Quick, weaker on Epic, and much weaker on Marathon.

Basically I don't like them and I especially don't like the production element of them.
My reasons:
1) Feels like the AIs are in some sense playing a similar game to me but then also have a whole separate game on top which rewards them for completely different things such as founding rubbish cities everywhere just for the bonus yields.

AI players don't intentionally exploit the A/B/C bonuses, by the way. They're intended to reward them for already good performance rather than act as a "secret bonus system", if that makes sense.

Now why I have a big issue with production:
1) A big part of the AIs bonuses are around reducing production costs already so on high difficulties they are often swimming in production after a while. This means the tech leader will often build every single wonder as the production cost is negligible to them. Also many AIs are constantly producing wealth/Science/culture on cities later on as they have nothing to build.
2) A lot of the times these production bonuses will end up going to waste overflow as they are much larger than the cost of many things.
3) However sometimes they will come at the perfect time and will all go into say a world project which gives them a free win out of nowhere. This happens pretty often to at least one AI so winning world projects is often impossible even if you are far ahead.
4) Also will pretty often complete Wonders for the AIs. This is very frustrating as when you know you are competing for a wonder you can try to use spies or look at the map to see if other civs are building it to make sure you will get it first, but then they just randomly complete it. I know great engineers can do this too but from what i've seen this is much less common and at least they actually had to do something (which I am also able to do) to get it.

I wonder if maybe the Production yields, and just the Production yields, could be removed, perhaps with an increase to the other yields and/or the existing production discounts on buildings and units to compensate. That way they couldn't be used to build World Wonders or World Congress projects. Any thoughts, @Gazebo?

 

[Casworon]

I wonder if maybe the Production yields, and just the Production yields, could be removed, perhaps with an increase to the other yields and/or the existing production discounts on buildings and units to compensate. That way they couldn't be used to build World Wonders or World Congress projects. Any thoughts, @Gazebo?

I am a fairly weak player but from the surface i love that idea. Using a cost reduction method rather than a bonus production yield method would be much better for tweaking the AI's performance in the area's it needs bonuses the most. While keeping Wonders and Projects competitive for weaker players such as me

 

[Milae]

So what difficulty are we talking about?

From what I understand the AI needs the production bonus to be somewhat competetive.
And a techer will always grab a good amount of wonders not just because of bonus but also because that player/AI can begin producing a wonder before others.

Well on Deity it's pretty common for the tech leader to build every single wonder in Renaissance (except Pisa which is ignored for some reason) and it seems like the production cost of Wonders is pretty insignificant even with the penalty for multiple wonders.

How else would you create different difficulties.

I do have some ideas but don't know where to post them. Honestly though I think just giving AI a percentage bonus to their yields would work fine.

AI players don't intentionally exploit the A/B/C bonuses, by the way. They're intended to reward them for already good performance rather than act as a "secret bonus system", if that makes sense.

Yea but a lot of Civs have a tendency to spam loads of cities and it's kinda weird that they get extra rewards for doing this imo.

 

[andersw]

Well on Deity it's pretty common for the tech leader to build every single wonder in Renaissance (except Pisa which is ignored for some reason) and it seems like the production cost of Wonders is pretty insignificant even with the penalty for multiple wonders.

Then I'd guess its more of a diety scaling issue.
(However in the games a diety player I watch on youtube this doesnt seem to be the case.)

 

[tu_79]

AI needs help to beat a good player, doesn't it?
Therefore the question is how to give ai these bonuses.

There are handicap bonuses that are granted to AI that perform well. This is Gazebo's idea for creating drama. Slightly better AI become really threatening. Later we included the ABC system as a way to distribute these bonuses in a way that allows good AI performers to rival with a human at his own level in any age. At least in standard they work much better than the older method, although minor tweaks can be made to the current values. People with experience can tune this easily in their games.

We could argue if they needed the production discount, seeing that they are already getting some bonuses, but the production discount is for all AI regardless of their performance, while the handicap bonuses are for the best.

I argue that both of these bonuses are necessary for the sake of tuning correctly the AI strength, and if ever, you can reduce or increase some of the values for better results.

Are you sure that the handicaps need to scale on game speed? Slower pace has more chances for winning handicap bonuses from winning wars, but that should be it.

 

[Recursive]

AI needs help to beat a good player, doesn't it?
Therefore the question is how to give ai these bonuses.

There are handicap bonuses that are granted to AI that perform well. This is Gazebo's idea for creating drama. Slightly better AI become really threatening. Later we included the ABC system as a way to distribute these bonuses in a way that allows good AI performers to rival with a human at his own level in any age. At least in standard they work much better than the older method, although minor tweaks can be made to the current values. People with experience can tune this easily in their games.

We could argue if they needed the production discount, seeing that they are already getting some bonuses, but the production discount is for all AI regardless of their performance, while the handicap bonuses are for the best.

I argue that both of these bonuses are necessary for the sake of tuning correctly the AI strength, and if ever, you can reduce or increase some of the values for better results.

Are you sure that the handicaps need to scale on game speed? Slower pace has more chances for winning handicap bonuses from winning wars, but that should be it.

If 100 Gold is worth 67 Gold on Quick, 150 Gold on Epic and 200 Gold on Marathon, and the costs of everything reflect that, we shouldn't have the AI always getting 100 Gold, because then they're getting more or less than the equivalent on Standard and the game difficulty goes up or down as a result. Same with the other yields.

 

[tu_79]

If 100 Gold is worth 67 Gold on Quick, 150 Gold on Epic and 200 Gold on Marathon, and the costs of everything reflect that, we shouldn't have the AI always getting 100 Gold, because then they're getting more or less than the equivalent on Standard and the game difficulty goes up or down as a result. Same with the other yields.

That would need to look at logs. Figure out how many times AI get the handicap bonuses for their successful actions for every game pace, and how much they are worth.

Anyways, it's a matter of adapting the values for the real conditions, not removing the system.

 

[Recursive]

That would need to look at logs. Figure out how many times AI get the handicap bonuses for their successful actions for every game pace, and how much they are worth.

I can tell you for sure it doesn't scale with game speed, I've read the code.

 

[Necamijat]

I can tell you for sure it doesn't scale with game speed, I've read the code.

I believe there are a few separate issues being mentioned here that are not discerned properly.

1) The bonus values don't scale with gamespeed. This is known and would probably need to be looked at.

2) Do longer game speeds allow the AI to pick up more of these bonuses due to longer games/more wars (as per tu_79's question)

3) Can the values be switched from bonus Production yields to something else to prevent sudden wonder rushes (the main part of the concerns here that I've gathered)

4) Does the AI need separate bonuses depending on it doing well or not? (see tu_79's post about 'creating drama' and keeping up)

I do have some ideas but don't know where to post them. Honestly though I think just giving AI a percentage bonus to their yields would work fine.

I'd say feel free to mention your ideas here. At least they'll be read and given thoughts.

 

[Milae]

I believe there are a few separate issues being mentioned here that are not discerned properly.

1) The bonus values don't scale with gamespeed. This is known and would probably need to be looked at.

2) Do longer game speeds allow the AI to pick up more of these bonuses due to longer games/more wars (as per tu_79's question)

Not really no, perhaps there are one or two more wars in a longer game that an AI could win but most of the bonuses come from settling cities, entering new eras, and golden ages which will happen the same amount of times on any gamespeed.
Also my personal feeling playing is that Deity on Epic is similar difficulty to Immortal on Standard.

3) Can the values be switched from bonus Production yields to something else to prevent sudden wonder rushes (the main part of the concerns here that I've gathered)

Yes this is my main concern.

4) Does the AI need separate bonuses depending on it doing well or not? (see tu_79's post about 'creating drama' and keeping up)

Maybe I should clarify that my other main issue with the bonuses are that they are flat bonuses to all yields. This means every AI Civ ends up looking similar in a game. I think if we are talking creating drama then we should give bonuses based on what they already produce so that they are better versions of the Civ they are playing but may still have some of the same weaknesses. Currently a snowballing science civ like Babylon will obviously be ahead on tech in mid-game which seems perfectly fair to me. But then they are also like 3 policies ahead as well which doesn't seem fair when you can see that they aren't really producing that much culture in their cities. Also every single AI city has ridiculously high pop even if it produces no food currently.
I suppose another slight point would be that rewarding building new cities is a strange one to me because it isn't really an achievement and seems unnecessary to encourage wide AI civs over tall ones.

I'd say feel free to mention your ideas here. At least they'll be read and given thoughts.

Okay well a few ideas here:
1) As I mentioned before I don't see what would be wrong with just a percentage bonus to their yields.
2) Decrease AI policy/tech cost increase from having more cities.
3) Give AIs a bonus to the instant yields which they receive already (from great people/policies etc).
4) Flat bonus to AI per turn yields in capital when they reach a new era. (Like Arabia Unique Ability)
5) Bonus to great person production.
6) Extra percentage bonus from being in a golden age.
7) Choose different promotions for their units. For example, I always go for logistics/range with siege weapons. Never seen the AI do this before (and would be very scary if they did).

 

[Recursive]

7) Choose different promotions for their units. For example, I always go for logistics/range with siege weapons. Never seen the AI do this before (and would be very scary if they did).

Apparently, the promotion selection AI is braindead.

 

[tu_79]

2) Decrease AI policy/tech cost increase from having more cities.

Without this penalty, it becomes impossible to win a game playing tall. Even as it is now, it's quite difficult. There's a time when any city is able to make up for the penalties just with how much they produce. Reducing this value would make wide playing stronger and I am not yet convinced that this style needs the help.

4) Flat bonus to AI per turn yields in capital when they reach a new era. (Like Arabia Unique Ability)

A mixed bonus has been proposed where some yields are granted for free in a per turn basis (general difficulty) and other ones are granted upon successful actions. But the current one is quite similar, since there is a production reduction based on difficulty and then handicap bonuses upon successful actions.

1) As I mentioned before I don't see what would be wrong with just a percentage bonus to their yields.

This is actually interesting. Currently the bonuses are flat and the yields depend roughly on the type of trigger. With this idea, I agree that civs will feel more unique when they are performing well. Potentially balance breaking, but I think it is worthy.
Another difficulty is that civs that are already ahead will gain much more than civs that are behind from the same actions.

What I'd do is to assign a total flat amount of yields, based on ABC handicaps, but distributed upon the current player's yield distribution.

I'll explain. Suppose that the handicap flat bonus is 1500 for the current turn. Then PlayerA do something like conquering a city, that triggers the bonus. We look at what PlayerA raw yields currently are.
Imagine it produces (average per city) 50 food and 80 production, (global) 24 gold, 12 science and 20 culture, plus 30 faith.
In percentage this is 23% food, 37% production, 11% gold, 5.5% science, 9% culture and 14% faith.
So the 1500 yields from the bonus are given in the same percentage: 345 food, 555 production, 165 gold, 82 science, 135 culture and 210 faith.
Food and production are distributed evenly among all cities, while the other yields are global. (Well, actually it makes sense to segregate local from global values).

 

[Stalker0]

So method aside, I think its important to understand how the curve needs to look.

Ultimately, the AI bonuses must remain exponential. Aka they get a smaller amount of bonus at the beginning, and the amount of bonus they receive increases as the game goes on (and not just in total bonus, the bonus as a percentage of the AIs overall yields needs to increase). This ensures that the bonus is not too large early game (when the human can not yet compete with their innate efficiencies), but scales up to compete against an efficient human engine so that late game can feel competitive even if the human is near the top of the stack.

Beyond that, there are lots of ways to tweak the system. One thing that we don't want to do is give AIs "actual yields", its important that most of their handicap bonuses remain instant yields. This is for happiness reasons....instant bonuses aren't applied to the happiness comparisons, so the handicap doesn't make the AI have more happiness innately. If you change that where the AI is actually generating greater yields....than it throws the happiness system completely out of whack.

 

[tothePAIN]

Exponential bonuses are good but changing it so that the AI isn't getting these outrageous instantaneous production bonuses would be very welcome. I've been the world leader in production, by a wide margin, but lost world projects to the AI by unbelievable margins (how does the AI do 28,000 hammers of production when I only do 12,000 when I have seaports / trainstations in factory in 20 cities...).

To give an example, on another world project, the AI only produced 4000 hammers... the additional 24,000 hammers was because of the AI bonuses.

 

[Deljade]

I'm not a fan of the fact that they occur when a city is founded. This fundamentally benefits certain leaders that have a flavour for founding many cities. I also suspect that this mechanic is a bit of a snowball and favours runaway AI, might be wrong here. Are expansionists AIs too powerful? Well I do remember highly expansionists leaders like Hiawatha and Shaka being consistent runaways some months back. Not sure how it is nowadays, haven't seen them in a while.

 

[amateurgamer88]

Just sharing some observations from my games. I find these bonuses tend to benefit AI that goes Tradition far more than Authority and Progress.

Progress for AI, like the human player, tend to take longer to start snowballing. Authority requires quite a bit conquest to be very powerful but it's actually possible that they won't be the biggest threat in my games until very late in the game if everything goes their favor.

What makes Tradition AI so different? Tradition AIs tend to get Wonders and early game is very strong due to Tradition bonuses. Tradition weakness should be weaker secondary cities but the bonuses actually make their secondary cities almost on par with Progress AI cities. A strong capital and decent secondary cities tend to make the leader very difficult to take down, especially when its neighbors couldn't keep up with its snowballing.

While I understand that bonuses are needed to make AIs competitive on higher difficulties, I kinda wish that we can do something about Tradition AIs that tend to snowball more out of control and faster than AIs taking other ancient policy trees. I think this might be why the AI tend to go Tradition as it can let them snowball more out of the gates than other options.