Concerns regarding the Eureka mechanism

[stealth_nsk]

Other examples are less obvious directly, but their results can still be recognized - like the fact that AIs were able to see what's happening in the fog in earlier games for example.

Yes, technically you could notice the difference. You need spy in every AI city to calculate its science output precisely. You need to look for it to see which eurekas the AI have and which the AI doesn't (quite tricky job). And with all this info you're like "gotcha - I just saw what the tech with eureka required the same amount of science as the tech without eureka".

But outside the world of flying pink unicorns you could never tell the difference.

From a budget position, sure.

It's much more than that. With AI being too complex, it will always be 2-3 weeks behind actual gameplay rules. At some points it could be critical as development could just plainly stop without testing if complex AI will be unable to play with new rules. In the end you could just loose months of development on it.

That's one of the points why modders could create more advanced AIs. Even if they change rules, they don't have large development team working on tight iterations.

 

[Ryika]

Yes, technically you could notice the difference. You need spy in every AI city to calculate its science output precisely. You need to look for it to see which eurekas the AI have and which the AI doesn't (quite tricky job). And with all this info you're like "gotcha - I just saw what the tech with eureka required the same amount of science as the tech without eureka".

But outside the world of flying pink unicorns you could never tell the difference.

Do I need to have a spy in every city to see how much food they're producing to understand that Deity AI has growth bonuses? No? Case closed.

It's much more than that. With AI being too complex, it will always be 2-3 weeks behind actual gameplay rules. At some points it could be critical as development could just plainly stop without testing if complex AI will be unable to play with new rules. In the end you could just loose months of development on it.

That's one of the points why modders could create more advanced AIs. Even if they change rules, they don't have large development team working on tight iterations.

We're not talking about complex AIs with custom scripting for X situations here, we're just talking about adding to the Flavor-System that already exists. In terms of Civ 5 you'd simply have the game check which technologies the AI will get in the near future and then add the flavors of the Eurekas associated with them into the Flavor Evaluations of the City AI. That really shouldn't be too hard, and it's open for adjustments, because you just have to tweak some numbers and that's it.

 

[stealth_nsk]

Do I need to have a spy in every city to see how much food they're producing to understand that Deity AI has growth bonuses? No? Case closed.

The example is irrelevant. You can tell if AI has huge bonuses, but you can't tell if in some cases AI bonuses differ a bit. Especially for such composite parameter as science output.

We're not talking about complex AIs with custom scripting for X situations here, we're just talking about adding to the Flavor-System that already exists. In terms of Civ 5 you'd simply have the game check which technologies the AI will get in the near future and then add the flavors of the Eurekas associated with them into the Flavor Evaluations of the City AI. That really shouldn't be too hard, and it's open for adjustments, because you just have to tweak some numbers and that's it.

No, we're talking about setting additional goals for AI. That's one of the worst things you could do for strategic AI (tactical AI have worst challenges, of course).

 

[winddysphere]

The thing that I'm concern the most : Will the eureka condition will be the same each game? I mean, it could be at least 2-3 minimum eureka conditions to make each game less likely to be the same ol. In my opinion it's like the players will know the tactics and keep doing the same over and over again.

 

[Ryika]

The example is irrelevant. You can tell if AI has huge bonuses, but you can't tell if in some cases AI bonuses differ a bit. Especially for such composite parameter as science output.

Of course you can still tell. It's less obvious, yeah, but the effect is still there. Again, that's the example with being able to see in the fog. Chances are you'll not realize in your first hundred hours, but over time you realize inconsistencies.

No, we're talking about setting additional goals for AI. That's one of the worst things you could do for strategic AI (tactical AI have worst challenges, of course).

No, we're not, that's nonsense. The flavor-system is created in a way that flavors get multiplied by the default values of the Civ, so any bonus that does not match their overall goal will never come into effect in the first place.

You can even make it so that the code checks for the AIs default flavor and only adds the Flavor for Eurekas if the AI has a high natural flavor, which means that the system simply somewhat reorders the things that the AI would build anyway by availability of Eurekas.

Anyway, it's obvious that we disagree, have started repeating points and won't convince the other party that we're right, so I'm leaving this discussion here.

The thing that I'm concern the most : Will the eureka condition will be the same each game? I mean, it could be at least 2-3 minimum eureka conditions to make each game less likely to be the same ol. In my opinion it's like the players will know the tactics and keep doing the same over and over again.

They're always the same as far as we know. But I agree that, if Eurekas are generic, a shuffle-system with random Eurekas would be a very nice thing.

 

[kaspergm]

I'm not worried that there are generic, easy-to-do eurekas since I think that's part of why they are there: to help civs/players that are not focused on science to research some techs easier.

My objection is: By making the Eurekas very easy to obtain, you fail at achieving the goal you outline above (which I fully support, btw.).

If you choose to go early warmongering, I fully support that you get tech boosts in that direction. But when these boosts are super-easy to get, that will not help the warmonger, because the tech-focused person will ALSO get these boosts, on top of all his other tech advantages. I mean, kill three barbarian units, kill a unit with a spearman, own two crossbowmen ... can you imagine anybody NOT doing these things, even if they play tech-focused?

And I know there is a timing question, but for most of these, that's just way off. For instance: Military Tactics (kill a unit with a spearman) is medieval era, and the spearman comes into game late ancient era. That means you have entire classical era and then some to kill that unit. Or Metal Casting (own two crosbowman) is late renaissance era, and the crosbowman comes early medieval. What are odds that people will not have build or upgraded two crosbowmen at that point?

I know it's hard to judge specifics without playing and at such an early build, but I'm voicing these concerns because I feel it's something that really needs to change before final build. Alternatively, eureka bonus needs to be cut significantly down (even if it seems to not always be the full 50 % but rather 30-40 %, that's too much if requirements are so easy to fulfill).

 

[stealth_nsk]

Of course you can still tell. It's less obvious, yeah, but the effect is still there. Again, that's the example with being able to see in the fog. Chances are you'll not realize in your first hundred hours, but over time you realize inconsistencies.

You keep throwing irrelevant examples instead of looking at just this case:
- You could only roughly tell the AI science output.
- You can't tell how many turns AI required for each particular tech as you don't get notification about AI tech discovery immediately. Only later with buildings, units, etc. you could tell whether the AI has particular techs.
- In most cases you can't tell whether the AI have boost or not. On top of normal boosts, there are boosts from Great Scientists and Spies.

With all this do you really think you can notice AI not playing eurekas, or you're arguing just for the sake of arguing?

No, we're not, that's nonsense.

Building a Water Mill before Construction foes halfway is a goal. If we want AI to really play eurekas, we need to set those goals and assign them priority based on civ flavour. But this means strategic AI suddenly have way more goals to proceed. It's not nonsense, it's how AI works.

 

[x2Madda]

To me that system look extremely gamey with its current design.
It will make people do random stuff not because they need it but because of optimizing boosts. Wich i think is unintuitive and contrary to the design of the system.

Like the op I think the rewards need to be lower or the tasks to be harder. It should either be more map based so that it varries a lot from map to map or should reward some strong detours out of an economic game.

I share these concerns, for example the boost to archery requires that a slinger unit scores the kill. On Emporer and above where AI will probably start with an army and reduced maintainance costs, we can assume they will jump into a weakened camp, as they currently do in all levels of Civ5 AI.

Infact if we look at Civ5 AI as a metric for Civ6 science, the AI already has an insane advantage on higher difficulties, lets assume they get reduced science costs for things and/or are able to produce more science due to less restrictions on them ala previous civ games and that they will almost certainly tick all the boxes for military science boosts, I question how a human player is even supposed to survive let alone win!

While in Civ5 a human player can stay small and go tall in order to catch up and then overtake the AI, due to limited land, which the AI will no doubt gobble up, I don't doubt the AI will remain competitive at all levels of civ6 play (above chieftain setting anyway).

I overall welcome the science boost system despite what I say because it has potential to be tweaked. It can be lowered to 30% or 10% and its overall goal is to encourage the player to not just passively babysit his super city with 0 military.
When Civ6 launches, it will be the quests from Civ:BE, a gamey system that undermines the whole experience. But once the numbers are tweaked and the system refined, it will probably be a great thing for future games.

Spoiler :

unless they dump the game like they did Civ:BE

The thing that I'm concern the most : Will the eureka condition will be the same each game? I mean, it could be at least 2-3 minimum eureka conditions to make each game less likely to be the same ol. In my opinion it's like the players will know the tactics and keep doing the same over and over again.

I would welcome this, a pool of 3 conditions where 1 is randomly assigned per game.
Would also make game of the months more tricky too.

 

[Ryika]

Building a Water Mill before Construction foes halfway is a goal. If we want AI to really play eurekas, we need to set those goals and assign them priority based on civ flavour. But this means strategic AI suddenly have way more goals to proceed. It's not nonsense, it's how AI works.

Okay, one last post on this, because you seem to be completely misunderstanding my argument.

Yes, it is "an additional goal", that's the point of the system, not only for the AI, but for everybody. There's more stuff to factor in. You're basically making the point that "The Eureka System adds more things to do, therefor the AI should not have to deal with it." - which is a ridiculous, nonsensical non-argument that could be made for literally ANY gameplay element. Yes, the AI has to factor in more stuff, because the player has to deal with more stuff as well. That's what happens when you add a new system to the mix.

It does however not make the game more complicated for the AI, and the AI does not need to be reprogrammed every-time a patch is being developed or whatever you implied in that "It's much more than that. With AI being too complex, it will always be 2-3 weeks behind actual gameplay rules."-argument. For the AI it's the same thing: Do stuff based on the priority list. The elements on the priority list now have to factor in an additional element, which will make all other elements on the list have a bit less influence, which is perfectly fine, because that exactly emulates what the player does. If an Eureka changes you just change the Flavors associated with that Eureka, or if it's an Eureka that no longer has to do with Production you remove the flavors.

There's literally nothing in the system that would prevent the AI from using the system. It will not be as efficient as the player at gathering up all the Eurekas, but then again, the AI is not as efficient as the player at _anything_. That's no reason to completely remove the system and replace it with a flat modifier.

 

[thecraftybee]

I imagine it will be hard to balance the Eureka boosts on quicker speeds.

If a normal tech takes 10 turns to complete on quick, but 40 on marathon, then it is far less likely you will be able to get the boosts relating to moving your units about and finding/fighting things, as you can only explore the same amount of space in both games.

I'd prefer to see the 50% bonus dropped to something like 10-20% at most, with perhaps a building/policy/UA boosting it a little further.

 

[uppi]

- In most cases you can't tell whether the AI have boost or not. On top of normal boosts, there are boosts from Great Scientists and Spies.

Those boosts are a good reason why the AI should have boosts: It has to get something out of these options, because due to the competition for great people, you will notice the AI getting them. That could be solved with a raw beaker boost, but that would require an additional ruleset for the AI that would need to be maintained and balanced.

And there is no need to disable the boosts for the AI, because it does not have to care about them. It might get them once in a while, and the rest is taken care of with the inevitable bonus the AI gets.

 

[stealth_nsk]

Yes, it is "an additional goal", that's the point of the system, not only for the AI, but for everybody. There's more stuff to factor in. You're basically making the point that "The Eureka System adds more things to do, therefor the AI should not have to deal with it." - which is a ridiculous, nonsensical non-argument that could be made for literally ANY gameplay element. Yes, the AI has to factor in more stuff, because the player has to deal with more stuff as well. That's what happens when you add a new system to the mix.

A small correction - it's system nearly invisible from outside. If your game involves bonuses for human player from solving a riddle, you don't want AI to actually solve the riddle to get those bonuses. On the point

It does however not make the game more complicated for the AI, and the AI does not need to be reprogrammed every-time a patch is being developed or whatever you implied in that "It's much more than that. With AI being too complex, it will always be 2-3 weeks behind actual gameplay rules."-argument. For the AI it's the same thing: Do stuff based on the priority list. The elements on the priority list now have to factor in an additional element, which will make all other elements on the list have a bit less influence, which is perfectly fine, because that exactly emulates what the player does. If an Eureka changes you just change the Flavors associated with that Eureka, or if it's an Eureka that no longer has to do with Production you remove the flavors.

It's not that easy. Priorities aren't fixed, they change depending on conditions. If AI don't have place to settle, it's settler priority should drop until the place is found; priority of military units change depending on which units you already have, which are needed, etc.

So, what happens if we use just static priorities? We say - Spearmen have slightly higher priority than their normal military value, because they have some usage in eurekas. This would work, but it doesn't make the AI actually play eurekas, because it would force it to build spearmen even after the boost is gained or tech completed.

If we want the AI to actually play the eurekas, we need to throw in more conditions. In the minimal variant, the condition to decrease priority of boost-causing items should be either gaining the boost or the tech. However it will not prevent obviously stupid things like building Water Mill in each city even though that's too early for them and you need only 1 for boost.

So, to make AI follow eurekas and do it in more or less smart fashion it needs the hell of conditions. Probably something like 5-10 universal ones and 1-3 unique for each eureka. Anything less than that will not under any conditions look smarter than just ignoring the eurekas.

 

[Staler87]

I've got to agree that a flat bonus at 30% to each tech would be a lot better than spending months at creating an AI only mildly capable at using the eureka system. Additionally for every goal the AI has to process and balance against its other goals the processing time of turns goes up. I don't think adding in a behavior that will affect gameplay and player opinion of the AI very little and increases processing times is worth it.

 

[Stringer1313]

I've got to agree that a flat bonus at 30% to each tech would be a lot better than spending months at creating an AI only mildly capable at using the eureka system. Additionally for every goal the AI has to process and balance against its other goals the processing time of turns goes up. I don't think adding in a behavior that will affect gameplay and player opinion of the AI very little and increases processing times is worth it.

I agree

 

[Ryika]

Yes, I agree.

I also agree that it's better to make the AI be able to see things in the fog instead of spending 100 years to create an AI that is capable of handling not seeing in the fog.

Because delaying the release for 100 years would not be a good thing to do.

Of course the very idea that it would take months to make the AI "mildly capable" of using the system or 100 years to make the AI "mildly capable" of playing the game without having to see in the fog are both simply nonsense.

 

[Tesludo]

I'd like to see the tech costs raised a lot, so that an Eureka is useful but not too overpowered.

And if this system is moddable, i'm going to turn these Eurekas in tech unlockers, rather than boosters. That way the terrain and your surroundings will really define your civ. Among some other tweaks...

 

[ProMeTheus112]

higher tech cost doesn't make Eureka less powerful if it is a % ! actually that means longer research time so more opportunity to get eurekas (lower their opportunity cost), and higher raw yield value of eurekas.

 

[Staler87]

Yes, I agree.

I also agree that it's better to make the AI be able to see things in the fog instead of spending 100 years to create an AI that is capable of handling not seeing in the fog.

Because delaying the release for 100 years would not be a good thing to do.

Of course the very idea that it would take months to make the AI "mildly capable" of using the system or 100 years to make the AI "mildly capable" of playing the game without having to see in the fog are both simply nonsense.

You ultimately have to make decisions about what to put processing power towards and what to put development time towards. When something can just as easily be simulated using a flat bonus and no additional AI programming (that may be a lot more complicated then you think) that the player won't notice can be done, I think it should. You could even add in that if the AI actually got the Eureka bonus it would increase to the full 50.

 

[Ryika]

(that may be a lot more complicated then you think)

No, it's not. I've seen how the system works (in Civ 5). And I tell you it wouldn't take much to add Flavors for Eurekas. And it wouldn't cost any meaningful amount of processing power either.

Unless they've decided to get rid of the Flavor-based system - which I wouldn't know why they'd do that, it's a great mix of having a universal system that runs reasonably well with many different sets of AI "personalities" - it would not at all be a problem to add flavors for things like Eurekas - at least those that have to do with constructing stuff. The rest would be, like most things with the AI, hit and miss.

I see absolutely no reason to think that they'd not make the AI use the Eureka-System, and the reasons you guys have brought up make no sense. It's basically like saying: "We shouldn't take water from that lake to wash our dishes, because it may run dry". Completely unproportional reasoning.

And again, "that the player won't notice" is just flat-out wrong. Players WILL notice, not immediately, but as they learn the game and the timings more they will realize that stuff can't be right. And once they're able to look at the code they will know 100%.

And the fact that %-modifiers and Eurekas simply not lead to similar results has also already been brought forward and not yet refuted by anyone. Have a Military-AI and a Tech-Focused AI use a %-modifier and the Tech-Focused AI will always come out ahead. Have them use the Eureka-System and the Military-AI may be able to keep up for a while by getting additional, unit-focused Eurekas.

 

[Acken]

At best the ai will factor in eureka boosts for techs it can research when chosing things to build research etc. Maybe through a percent bonus on the weigths calculations. Id be surprised if Firaxis goes further than that.