Wiser Orcs (a FfH II mod)

[Bringa]

Hey,

I downloaded the newest version of this today. I get a file named wiser_orcs_033_d_005.zip, inside of which there is a directory named Wiser Orcs 033.c 003. I guess you zipped the wrong dir?

 

[xanaqui42]

Hey,

I downloaded the newest version of this today. I get a file named wiser_orcs_033_d_005.zip, inside of which there is a directory named Wiser Orcs 033.c 003. I guess you zipped the wrong dir?

No; that's just a process problem - it ends up being easier for me to give the same (erroneously-named) directory. Just use the verification date to double-check that it's the right build.

I'll warn you that Wiser Orcs is no Smarter Orcs, yet- I really didn't start attacking the tech tree until today. For that matter, one of the early tech problems in Smarter Orcs (once we changed the dll) I just removed 1/2 an hour ago.

 

[xanaqui42]

Warning: This is based on FfH II 0.33e, which is not the latest version.

Technology choice fixes
There were a couple defects in how the AI chooses technology. Specifically, it could choose a technology that even its algorithm decided was sub-optimal, and it would nearly always choose any technology that was very cheap. As a side-effect of my debugging strategy, you'll note that technology suggestions are modified by flavor, just as they are for the AI.

Intelligent Animals
Specifically, the AI players would use animals the same way uncaptured animals were used - which doesn't tend to bring about coordinated attacks or even city defense. While there was no effort to use them in an optimal manner, they should no longer be forced to attack the closest unit regardless of anything other than the difficulty level.

Strategy Framework
While this is by far the largest change here, you aren't going to notice a lot yet. Mostly, this is the beginning of the framework for future changes.

You may note that Falamar will always rush for Octopus Overlords in the present version. That is because I used him as a test subject for the new Strategy Framework. At the moment, Falamar's rush (and a tiny time-waster per turn) is all that's implemented. however, there are hooks for far more, which we should start to take advantage of in the next few releases.

Strategy Classes
The basic concept is that there are multiple strategies, each which has 0 or more Strategy classes. Strategy classes are used for multiple exclusion - so that you don't have a civilization trying to, say rush Octopus Overlords at the same time that it's trying to rush the Order (given that you give them both the same Strategy class). Note that Strategy Classes are, at the moment, untested. That will be one of the higher priorities for the next release.

Strategy State Engine
There is now the capability for using a strategy state engine. Essentially, each Leader has a block of Python code it fires off during initialization that presumably will be used to determine their strategies (yes, a leader can have more than one, given that they're of different Strategy Classes). Each strategy that a player has will fire off another piece of python code each turn that allows presumably later strategy changes (so, for example, you could switch out of a melee strategy when Copper isn't found). Note that the present code is little more than a test of basic functionality. Also note that I will likely embed more common triggers into their own functions, so that the use of the turn function is less necessary.

Strategy Features
A Strategy can presently do the following:

• Alter any number of flavors for the present player.
• Execute a block of Python code each turn (largely for the state engine, but it could have other uses, particuarly now that there isn't much functionality in here yet).
• Alter the religion weight modifiers for a leader (note that this does not affect if a leader can obtain a religion, unlike the standard religion weight modifier. It does (as well as the leader weight) affect disciples chosen).

 

[[to_xp]Gekko]

thans for all you're doing xanaqui42, a better AI is invaluable for a better game. this should really be in base FFH2 imho. looking forward to the next releases that will bring this to patch G

 

[xanaqui42]

thans for all you're doing xanaqui42, a better AI is invaluable for a better game. this should really be in base FFH2 imho. looking forward to the next releases that will bring this to patch G

I mostly play until the AI frustrates me, then I try to fix it A merge with 0.33g (or higher) will be in the next release. I'm not certain when that release will be, though - I'd like to implement at least a bit of what I've gotten via the religion quiz, and test strategy classes.

 

[xanaqui42]

Next I'd add Arrays of Unit/Building Infos with AI weight values which will add to the basic AI Weight Values.

Question: Which of the following would be better:

1) Modify building values directly through strategy classes
2) Add flavors to buildings, and let all flavors affect building value

 

[xienwolf]

If you are going with the idea to create new flavors for specific path objectives, then the Flavor is probably a good way to go. Then you can have overarching "emphasize culture" commands, as well as specific "Get a Planar Gate" ones without too much issue.


The only reason I can see NOT to do things with that approach is a desire to avoid creating a ton of Flavors, but I fail to see any downside to having a lot of them, they aren't looped too often from what I have seen. Though Flavors might be so overused now that it becomes difficult to balance the final values against each other and get the end result desired (same issue would exist with direct value adjustment though).

Primary advantage to the array of building weights is that it gathers the data in one spot in your Strategy XML file for ease of modification and balance.

 

[[to_xp]Gekko]

xanaqui42, while you're adding "flavours" for the AI, you could probably address this issue:

the algorithm with which the AI civs choose a religion is still the same as the original civ, in which religion was basically just for making friends. this means that the AIs will happily convert to a very unflavourful religion if it happens to spread to just one of their cities. this doesn't make sense in FFH, and adding "religious flavour" to each leader would be great

 

[xanaqui42]

xanaqui42, while you're adding "flavours" for the AI, you could probably address this issue:

the algorithm with which the AI civs choose a religion is still the same as the original civ, in which religion was basically just for making friends. this means that the AIs will happily convert to a very unflavourful religion if it happens to spread to just one of their cities. this doesn't make sense in FFH, and adding "religious flavour" to each leader would be great

Actually, it already exists (in LeaderHeads). The strategy code allows you to change it further. I haven't looked into the valuation of religions; it's possible that the numbers used are far too small.

 

[xanaqui42]

I started to look at adding flavors for buildings, and I discovered that they are already there (in FfH II). That was easy .

 

[lexington1]

Right now, it seems to pretty much ignore it. And (thought a few people have commented on this already) religions-I had one game where the Luichurp (a good civ, mind you) adopted AV when it was in maybe half his cities and a good number of them also had Empyrion in many of them-the AI should be changed so evil religions (AV, to a lesser extent OO and Esus) are avoided by good civs, and likewise good religions (Order, Empyrion, maybe Runes) by evil civs.

Oh, and the AI can't grasp that the Lanun are a water-based civ, and tries to play them like a normal land-based one, even when its in the middle of a bunch of islands.

 

[xanaqui42]

Right now, it seems to pretty much ignore it.

Presently, our goal is to re-enforce that; concentration on magic weakens an AI player. Once we get to working on Magic AI, we'll re-enable it - and in the meantime, we can work on priestly spells.

I'm hoping to code a group of spells for the next release or the one after it. I'm planning to work on a group of spells that includes Drown (as that's pretty critical for Octopus Overlords).

And (thought a few people have commented on this already) religions-I had one game where the Luichurp (a good civ, mind you) adopted AV when it was in maybe half his cities and a good number of them also had Empyrion in many of them-the AI should be changed so evil religions (AV, to a lesser extent OO and Esus) are avoided by good civs, and likewise good religions (Order, Empyrion, maybe Runes) by evil civs.

The next release will work significantly on religions, although it won't be complete by any measure. However, the description you have (religion by alignment) is only one of the algorithms we're using. Also, roughly 1/2 or so of civilizations will use the old algorithm.

Oh, and the AI can't grasp that the Lanun are a water-based civ, and tries to play them like a normal land-based one, even when its in the middle of a bunch of islands.

Can you describe this problem more specifically? Does it have problems in putting cities near water, in naval combat, in moving troops between islands/continents, researching naval technology, or something else?

 

[lexington1]

Glad to here your working on this, and I'm looking forward to trying it out. Just curious-are you planning on having civs favor specific religions (Order for Bannor, for instance).

The problem I mentioned with the Lanun happened when I played this map (http://forums.civfanatics.com/showthread.php?t=236321). It puts the Lanun on a largish island, mostly covered with jungle but with one good city spot, in the middle of an archipelago. The Lanun AI, instead of doing what a human would do (rush naval technologies and settle the surrounding islands) always founded two or three rather useless cities in the jungle, and never tried to colonize the rest of the archipelago. I chopped the size of the island in half on worldbuilder to try to force the AI to settle the surrounding islands, and yet it still obstinately refused, having only built one other city on its island several hundred turns into the game when every other empire had at least four or five, and I had long since researched optics. I don't know if its a technology problem, or, (as I'm inclined to suspect) the AI can't grasp the concept of the Lanun being a seafaring civ, and thinks it must have a bunch of cities on its starting continent before trying to found any on islands.

 

[xanaqui42]

Glad to here your working on this, and I'm looking forward to trying it out. Just curious-are you planning on having civs favor specific religions (Order for Bannor, for instance).

At the moment, I just took portions of the first 10 quiz responses that were easy to implement, making the assumption that virtually any human strategy is better than the present one. Eventually, I'll either remove some of the less effective strategies, or place them in the easier difficulty levels.

Here's the present outline:

1. If the leader is Hyborem, Basium, or Barbarian, do nothing (Hyborem and Basium already have access to their favorite religions on creation, and the Barbarian player doesn't use religions).
2. Amelanchier, Arendel, Thessa : 90% chance of focusing on Fellowship of the leaves. Otherwise use the default logic
3. Arturus, Kandros: 90% chance of focusing on Runes of Kilmorph. Otherwise use the default logic
4. Faeryl: 80% chance of focusing on Fellowship of the leaves. Otherwise use the default logic
5. Falamar, Hannah: 33% chance of focusing on Octopus Overlords. Otherwise use the default logic.
6. If none of the above, use the default logic.

Default logic:

1. Do nothing if Agnostic. Otherwise:
2. If the difficulty is Warlord or easier, there is a 33% chance of focusing on Octopus Overlords if 1/3 of the initial visible tiles are water.
3. If the difficulty is Noble, there is a 20% chance of choosing religion to make friends (without giving someone else an easy religious victory), and a 13% chance of selecting at least 3 religions, starting with one of the early religions, at roughly equal odds.
4. If the difficulty is Prince, there's a 17% chance of choosing Runes of Kilmorph, and a 16% chance of choosing The Order.
5. If the difficulty is Monarch or Emperor, there's an 11% chance of choosing Fellowship of the Leaves first (as a gateway to Council of Esus or Empyrean), an 11% chance of choosing Runes of Kilmorph (as a gateway to Council of Esus, The Order, Octopus Overlords or preferably Empyrean), and an 11% chance of choosing religion by alignment (ignoring Fellowship of the Leaves)
6. If the difficulty is Immortal or Deity, there's an 11% chance of attempting to found at least one religion, starting with Kilmorph if creative, Octopus Overlords if starting with a lot of water, Council of Esus or Empyrean. There's an 11% chance of going for Fellowship of Leaves (or Runes of Kilmorph if there are no Forests visible), and making a mild attempt at grabbing the Shrine of Esus. There's an 11% chance of going for the Empyrean.

The problem I mentioned with the Lanun happened when I played this map (http://forums.civfanatics.com/showthread.php?t=236321). It puts the Lanun on a largish island, mostly covered with jungle but with one good city spot, in the middle of an archipelago. The Lanun AI, instead of doing what a human would do (rush naval technologies and settle the surrounding islands) always founded two or three rather useless cities in the jungle, and never tried to colonize the rest of the archipelago. I chopped the size of the island in half on worldbuilder to try to force the AI to settle the surrounding islands, and yet it still obstinately refused, having only built one other city on its island several hundred turns into the game when every other empire had at least four or five, and I had long since researched optics. I don't know if its a technology problem, or, (as I'm inclined to suspect) the AI can't grasp the concept of the Lanun being a seafaring civ, and thinks it must have a bunch of cities on its starting continent before trying to found any on islands.

This sounds like multiple problems:

• The AI isn't recognizing that coastal areas are good for the Lanun.
• The AI isn't focusing on transport technology and/or building transports when its growth is limited.
• The AI isn't settling on non-primary areas.

 

[[to_xp]Gekko]

nice logic for the AI xanaqui42, although I think that Faeryl should be 60% CoE - 40% FoL or something - those svartalfars are tricky, you know

btw, the AI definitely has an issue with everything concerning a navy. getting a settler on a boat to colonize islands is a concept they just can't grasp, this is not an issue unique to the lanun. it's just that with them it's worse cuz it takes away their primary strenght.

edit: actually the religious AI logic for Faeryl could be a bit more complicated. like, FoL in the early game, then try not to build religious FoL heroes and switch to CoE as soon as it becomes available. this gets them a lot of useful ancient forests btw. then, if they happen to build FoL religious heroes ( maybe cuz they need them for an offensive war, maybe cuz they are under attack ) , of course they should stick with that instead of switching to CoE.

 

[lexington1]

At the moment, I just took portions of the first 10 quiz responses that were easy to implement, making the assumption that virtually any human strategy is better than the present one. Eventually, I'll either remove some of the less effective strategies, or place them in the easier difficulty levels.

Here's the present outline:

1. If the leader is Hyborem, Basium, or Barbarian, do nothing (Hyborem and Basium already have access to their favorite religions on creation, and the Barbarian player doesn't use religions).
2. Amelanchier, Arendel, Thessa : 90% chance of focusing on Fellowship of the leaves. Otherwise use the default logic
3. Arturus, Kandros: 90% chance of focusing on Runes of Kilmorph. Otherwise use the default logic
4. Faeryl: 80% chance of focusing on Fellowship of the leaves. Otherwise use the default logic
5. Falamar, Hannah: 33% chance of focusing on Octopus Overlords. Otherwise use the default logic.
6. If none of the above, use the default logic.

Default logic:

1. Do nothing if Agnostic. Otherwise:
2. If the difficulty is Warlord or easier, there is a 33% chance of focusing on Octopus Overlords if 1/3 of the initial visible tiles are water.
3. If the difficulty is Noble, there is a 20% chance of choosing religion to make friends (without giving someone else an easy religious victory), and a 13% chance of selecting at least 3 religions, starting with one of the early religions, at roughly equal odds.
4. If the difficulty is Prince, there's a 17% chance of choosing Runes of Kilmorph, and a 16% chance of choosing The Order.
5. If the difficulty is Monarch or Emperor, there's an 11% chance of choosing Fellowship of the Leaves first (as a gateway to Council of Esus or Empyrean), an 11% chance of choosing Runes of Kilmorph (as a gateway to Council of Esus, The Order, Octopus Overlords or preferably Empyrean), and an 11% chance of choosing religion by alignment (ignoring Fellowship of the Leaves)
6. If the difficulty is Immortal or Deity, there's an 11% chance of attempting to found at least one religion, starting with Kilmorph if creative, Octopus Overlords if starting with a lot of water, Council of Esus or Empyrean. There's an 11% chance of going for Fellowship of Leaves (or Runes of Kilmorph if there are no Forests visible), and making a mild attempt at grabbing the Shrine of Esus. There's an 11% chance of going for the Empyrean.

This sounds like multiple problems:

• The AI isn't recognizing that coastal areas are good for the Lanun.
• The AI isn't focusing on transport technology and/or building transports when its growth is limited.
• The AI isn't settling on non-primary areas.

Wow, that sounds really impressive-keep up the good work, and I hope this gets folded into the official FFH releases someday. If I may make one suggestion, I think you should get the Bannor to go for order (and get civilizations to not adopt certain religions, so you don't get contradictions like Empyrean Calabim). But don't get me wrong-I really appreciate what you've done so far.

 

[xanaqui42]

Wow, that sounds really impressive-keep up the good work, and I hope this gets folded into the official FFH releases someday. If I may make one suggestion, I think you should get the Bannor to go for order (and get civilizations to not adopt certain religions, so you don't get contradictions like Empyrean Calabim). But don't get me wrong-I really appreciate what you've done so far.

To be clear, I'm describing the algorithm in detail so that I can get feedback - particuarly negative/corrective feedback. One change I'm already considering is to have the "Religion by Alignment" algorithm available at each difficulty level.

I think that if I continue to develop Civilization Strategies (as I presently plan to) , it could help a lot with scenarios. Hopefully, within a couple more releases, it will also be a significant boost to the AI - for example, I hope to attack Copper Rush after getting the Religion logic more settled; it would be really nice to be able to rush the melee line until you find you can't get a metal, then switch to a different military strategy (as an example).

Unfortunately, at the moment, it looks like the Drown logic won't make the next release.

 

[orangelex44]

To be clear, I'm describing the algorithm in detail so that I can get feedback - particuarly negative/corrective feedback. One change I'm already considering is to have the "Religion by Alignment" algorithm available at each difficulty level.

I'd say this is reasonable - from the individual civ's standpoint, it may not necessarily be the best plan, but from a lore standpoint, it definately makes sense. And it's just a weight, so the AI can still occasionally wank off and adapt some completely insane religion (say, an OO Elohim) - which is fun to see as a player, but annoying to see a lot as a person interested in Erebusian lore.

 

[arcticnightwolf]

What's status of this project? Active, abandoned, forgotten, delayed, in very long progress ??

 

[Vehem]

What's status of this project? Active, abandoned, forgotten, delayed, in very long progress ??

xanaqui42 xanaqui42 is offline
Last Activity: Oct 30, 2008 02:54 AM

He'll be back I'm fairly sure - but not sure when