Announcement

**Buck Birdseed** · September 2, 2001, 08:03

Is this like the 15th time you've posted this idea? I thought it was stupid back then (inflexible and predictable)...

**Gramphos** · September 2, 2001, 08:06

Re: More funnel-shaped spider-sandtraps in player-created maps

Originally posted by Ralf
This is an old idea of mine, but described in a different way then before - and since so many are newcomers here at Apolyton, I have decided to take another shoot at it. If you already read about it before - bear with me, or read another topic. Its one of my favourite ideas.

Yes, I've read it before.

Now - wouldnt it be nice if the map-creator, besides editing contintental shapes, terrain-type allocation, and (if creating a scenario-map) established cities, terrain-improvements, like roads and so on - ALSO had the free optional ability to pinpoint ideal potential AI-expansion city-locations all over the map?

This identifies your idea and makes it recognizable.

I say as I always have said. There must be ways to check if a square is suitable for a city that the AI can adapt without this "if AI moves settler over defined spot then found city" feature.

**Ralf** · September 2, 2001, 08:46

Originally posted by Snapcase
Is this like the 15th time you've posted this idea? I thought it was stupid back then (inflexible and predictable)...

This is because you dont share Stefu's irritaton in his thread Pillage! Pillage! Pillage!. Many civers (including me) can identify with him though. And it is for those civers (and map-creators who wants better AI-expansion control) this idea is meant for.

Not for you, Snapcase.

(inflexible

There is nothing "inflexible" about helping AI-settlers to establish its cities more efficiently, thus saving it as much as possible from its own stupid self.

and predictable)...

- There is nothing "predictable" about editing potential ideal AI-city locations all over the map, that are 100% invisible (and ignoreable) anyway, for the human player. It only concerns the AI-civs. The HP can choose to follow this guidelines also (if added by the map-creator), but thats always completely optional.

- Each AI-civ can always choose to expand randomly in any direction it wants.
- Each AI-civ start-out locations can also be allocated randomly from game to game on standalone maps.

Finally, the whole idea is optional anyway, so whats your problem Snapcase? By comparison, Im not sure I like the "civ-specific allocation-table", but since its optional - so what. Since I no longer feel pressured to use it, I might probably try it out. Who knows, maybe one start to like things if only one give it a chance.

**TechWins** · September 2, 2001, 12:32

Yes, I've read it before and I still think it's a good idea. Even though, it doesn't apply much to me.

**tniem** · September 2, 2001, 13:08

I wouldn't mind such an option added.

(Ralf - read idea before, but nice to see that your still on your crusade and the title of the thread is quite interesting)

After all the computer does many times found cities in absolutely silly spots. Especially three cities all overlapping on the resources that they can get.

But I think it is more important to just have the code make AI placement better. Even if you put these sandtraps into the game, you will have sandtraps that overlap and the AI will put cities in silly places compared to where their old cities were. I just think a better AI code is the best solution especially since I mostly play random maps.

**Sabre2th** · September 2, 2001, 13:12

**Ralf** · September 2, 2001, 16:10

Originally posted by Gramphos
I say as I always have said. There must be ways to check if a square is suitable for a city that the AI can adapt without this "if AI moves settler over defined spot then found city" feature.

The AI-problem isnt so much about checking a single productive square - thats dead-easy. But, why not compare with yourself:

First you probably gonna compare with your already established city-areas and colonies in order to estimate a good enough new location. You take the development-potential of all the surrounding 20 city-area terrain-types and make a general estimation. You also quicky imagine other potential start-out squares 1-3 squares away and around in a cirkle - all in order to valuate the best overal 21 square-solution. You continue with overviewing any nearby resources/luxuries - estimate culture-growth and road-connecting. You also compare your expansion-plans with future available land, limiting coastlines, closeby mini-islands. Not to mention nearby foreign city-locations and their culture-growth and their probable expansion-strategies, and so on.

So in order to mimic ideal/perfectionist city-expansion strategies, theres a hell lot more calculations that have to be done then just checking if a single square is fertile enough, or not. Infact above is really way too much for the AI to handle within a short few seconds, also considering other prioritys. Not to mention the problem with writing such an AI.

Now, Im all positive for improved AI-expansion algorithms, and Im also very excited about the colony-feature. I have also checked out some colored mini-maps, and if the white dots represents AI city-locations (and not MP ones), I must say they have improved the AI-expansion considerably, compared with Civ-2/SMAC. Everything so far seems in many ways "on the right track".

Its just that I truly want this optional feature added to (at least) human-edited maps (NOT random-generated - unless thay come up with truly convincing results) - and not only for AI-reasons. There are scenario-reasons, as well. Read my answer to tniem.

Originally posted by TechWins
Yes, I've read it before and I still think it's a good idea. Even though, it doesn't apply much to me.

Thanks!

Just because one personally dont use it (or perhaps dont like it) one doesnt always have to be instinctively against the option. And if this idea is wortwhile enough to implement - thats up to the Firaxis-team to decide. They have (of course) already decided (positively or negatively) what to do with this particular idea (if they even discovered it). We/I just dont know about it yet.

Originally posted by tniem
Even if you put these sandtraps into the game, you will have sandtraps that overlap and the AI will put cities in silly places compared to where their old cities were.

If these "sandtraps" overlap each other or not, is entirely up the the human map/scenario-creator. Its really up to him: it is he alone who manually position potential AI-city locations everywhere he wants them. The whole idea is about letting the human mind do what it does best; overview and estimate. Also, the map/scenario-creator can choose (or not choose) to add this feature. Either way, the player always have the final GO. You (the player) can...

- freely choose over HP city-expansion, and let the AI calculate its own city-expansions.
- freely choose over HP city-locations, and let the AI be guided by the map/scenaro-creator.
- let both yourself and the AI-civs be guided by the map/scenario-creator.

The third alternative can be interesting if one want to play, for example a Roman expansion-scenario, and one want to follow a historically realistic expansion-pattern (and be sure that other AI-civs do likewise).

I just think a better AI code is the best solution especially since I mostly play random maps.

A better AI-code is perhaps just what Firaxis aiming for in randomly generated maps. This whole idea is mostly aimed for map/scenario-creators who want this added level of control, and for some us players who would truly enjoy playing & testing it.

**Buck Birdseed** · September 2, 2001, 16:28

It's inflexible, because ideal locations are not chosen according to a pattern. The perfectionist style of play is in fact horribly inefficient, with resources and worker-turns wasted on terraforming unsuitable terrain instead of doing something useful. It's predictable, because once you've played through the scenario twice you're going to know where exactly cities will spring up. It's inflexible because different locations may be best suited depending on the situation at the time- this way AI civs will never place a city on that Island off your coast and use it as an air-assault base. It's predictable because once you've figured out that these "sand-traps" are used, you know you're dealing with an anal-retentive perfectionist and you just place a fortified unit in each of the squares (you can measure out the pattern) and the AI can no longer expand. It's inflexible because it prevents the AI from using whatever tactics are hardcoded into it, thus making the AI even worse. It's predictable because you can predict the movement patterns of enemy settlers and intercept them.

There, three of each reason. How about that?

**Gramphos** · September 2, 2001, 16:36

Originally posted by Ralf
First you probably gonna compare with your already established city-areas and colonies in order to estimate a good enough new location.

Exactly what do you refere to as a good enough new location?

You take the development-potential of all the surrounding 20 city-area terrain-types and make a general estimation. You also quicky imagine other potential start-out squares 1-3 squares away and around in a cirkle - all in order to valuate the best overal 21 square-solution. You continue with overviewing any nearby resources/luxuries - estimate culture-growth and road-connecting. You also compare your expansion-plans with future available land, limiting coastlines, closeby mini-islands. Not to mention nearby foreign city-locations and their culture-growth and their probable expansion-strategies, and so on.

That is true, but an AI could for instance give each terrain type points on how good the are to have within the city. Then sum all points of the 20 squares around a good (grassland or plain) midpoint. Then get a number if that number is bigger then a minimum number the area can be defined as good. Then it has to check for foreign cities, which it has to do anyway, and see if that point will be swallowed by any culture or if it is a point on a front.
I'll see if I can come up with an example another day.

So in order to mimic ideal/perfectionist city-expansion strategies, theres a hell lot more calculations that have to be done then just checking if a single square is fertile enough, or not.

I said anything of a square, I was speaking of a spot including the surroundings.

Infact above is really way too much for the AI to handle within a short few seconds, also considering other prioritys. Not to mention the problem with writing such an AI.

As I said earlier in this post the problem wit other cities nearby has to be handled anyway.

Now, Im all positive for improved AI-expansion algorithms, and Im also very excited about the colony-feature. I have also checked out some colored mini-maps, and if the white dots represents AI city-locations (and not MP ones), I must say they have improved the AI-expansion considerably, compared with Civ-2/SMAC. Everything so far seems in many ways "on the right track".

I'm glad that you are positive to the game, and how mush you have seen of it so far. I'm too.

Its just that I truly want this optional feature added to (at least) human-edited maps (NOT random-generated - unless thay come up with truly convincing results) - and not only for AI-reasons. There are scenario-reasons, as well. Read my answer to tniem.

I can see why you want this, and as long as it can be turned off (in case it doesn't work as supposed) I'm not against it.

**Gramphos** · September 3, 2001, 13:42

Here is the promised example. (It may have to be balanced, but you will see the principle)

I’ve used Civ2 Terrains and resources for this, but once more it is the principle.
First of All I assigned each terrain a value (may yet have to be balanced). Resources will have their own values that will be added to the base value of that terrain.

Terrain	Value
Grassland	4
Plains	4
Desert	1
Forest	2
Hills	2
Mountains	0
Tundra	1
Glacier	-1
Swamp	1
Jungle	1
Ocean	2
Oasis	1
Buffalo	1
Pheasant	2
Coal	2
Gold	2
Ivory	2
Peat	2
Gems	1
Fish	2
Oil	2
Wheat	2
Silk	1
Whine	1
Iron	2
Furs	2
Spice	2
Fruit	1
Whales	2
Shield	1
River	1

It shall also mind extra about the center square. That square must be plain or grassland. If the square is a river it gives +1 point (not including the ordinary river point). Finally if that square is next to an ocean square (so that you can build ships) it will give you +1 point.
If a Cityarea gets over 50 points it can be concidered as acceptable for an AI city. (Onse more the points and this number may need balancing to be good.) Also the maximum food that can be produced in the city area in one turn might be calculated (and have a minimum value set as well).

For the example I’ve attached an image showing a startup position (yes it is not a superb location, but I didn’t have time to play Civ2 right now).
It the center square grassland or plains? - Yes: go on with full area check.

First count the points from terrain types:
4 plains * 4 = 16
4 hills * 2 = 8
1 mountain * 0 = 0

9 ocean *2 = 18 (that mush?)
3 grassland * 4 =12

Now count all extra terrain:
2 rivers * 1 = 2
1 shield * 1 = 1
1 whale * 2 = 2

Now add extra bonuses for the center square:
The city is next to the ocean: => +1.

Now check the sum:
16 + 8 + 0 + 18 + 12 + 2 + 1 + 2 + 1 = 60
Is 60 > 50 ?
Yes => The area is acceptable for an AI-city

Attached Files

**Ralf** · September 4, 2001, 01:24

Snapcase - first of all:

I havent the slightest interest in trying to pursuade you to play the game with a feature like this enabled. They whole idea is from start to finish meant to be optional. Get that? Its entirely up to the player if he wants to bypass it - also on behalf of the AI-civs. You dont have to create, download or play these special maps/scenarios just because they happen to be available. Its a free choice - IF the idea (or similar) has even been added at all. I havent the foggiest.

Originally posted by Snapcase
It's inflexible, because ideal locations are not chosen according to a pattern.

Who was talking about patterns? Did I mention anything about patterns?

Whether it shall be a regular stereotype pattern or not is entirely up to the human map/scenario-builder. Most of the time he is forced to consider limiting coastlines, lakes, unhabitable mountains and such. Personally, I would have no problems at all pinpointing these locations so that the overal available land-area got efficently exploited and at the same time make it look nice and natural.

As for scenarios; lets take that Roman expansion one, and lets say we start very early then only the italian peninsula belonged to Rome. Without this feature the AI-civs establish new cities blindly according to some generic expansion-algorithms. Thats OK as an option - but what if one wants to follow historical facts? Well, with this feature the scenario-builder is suddenly given a expansion-control time-dimension as well. He can choose to pinpoint all the exact historic city-locations for the whole damn timeline.

One can perhaps criticise the execution of above idea - maybe where are better ways of implementing it. But not the idea itself. Its too good for that.

The perfectionist style of play is in fact horribly inefficient, with resources and worker-turns wasted on terraforming unsuitable terrain instead of doing something useful.

This topic is NOT about whether the perfectionist playingstyle is inefficient, or not. Feel free to start another topic about that. Most peaceful perfectionists (including me) plays this way simply because its REWARDING to see how the terrain slowly becomes more and more improved & cultivated. They see it as a CHALLENGE to exploit a big, but still limited land-area as efficiently as possible. The reward is also an empire that also looks really good.
By comparison, spurting diarrhea/ICS-style amounts of halfheartedly developed cities/city-areas all over the whole damn map, simply doesnt feel like that much fun to me.

It's predictable, because once you've played through the scenario twice you're going to know where exactly cities will spring up.

This is exactly the idea with scenarios, Snapcase.

You are playing a SCENARIO for crying out load. The very word "scenario" means scripted; guided; edited; prepared setup, or step-by-step example (compare with a movie-script). By the way; how many times have you ever replayed one and the same Civ-2 scenario? 10 times? 20 times? Most of us replays it only a few times (if good), then we downlad a new one. On the other hand; what about games like "Europa Universalis"? Isnt that game infact a huge scenario in itself? With fixed provinse- and nation-locations all over the map? Does this game automatically become boring just because the player knows all the important key-provinces after a couple of sessions? Well, many customers dont thinks so.

The great thing with the Civ-3 game, is that you can have it all. Both random main-games, and very detailed; higly prepared; scripted; pre-edited scenario-games.

It's inflexible because different locations may be best suited depending on the situation at the time- this way AI civs will never place a city on that Island off your coast and use it as an air-assault base.

Ask yourself - what must the AI protect? The city itself + recources, or a desolate 3-4 square mini-island outside your coast? IF the mini-island is worthwhile enough for a city, you can be sure that I would place a "sandtrap" on that as well. Furthermore; the AI can of course adjust any general empire expansion-directions dynamically according to the game-situation at the time.

It's predictable because once you've figured out that these "sand-traps" are used, you know you're dealing with an anal-retentive perfectionist and you just place a fortified unit in each of the squares (you can measure out the pattern) and the AI can no longer expand.

Now you at it again with those stereotype pattern-fantasies.

I think I have expained enough further up this reply/topic. And Snapcase, if you absolutely insist on hindering the AI-civs to expand - why not use the "reveal map" cheat-menu alternative? Then you can chase down unprotected newly founded AI-cities all over the place.

It's inflexible because it prevents the AI from using whatever tactics are hardcoded into it, thus making the AI even worse.

They cannot program the AI too overview & estimate randomized map expansion-possiblities as human players do. Its way too complex. If you understood the general & principal problems with programming an AI in order to achieve above, I would perhaps write a longer reply (or more probable; we wouldnt even have this discussion).

It's predictable because you can predict the movement patterns of enemy settlers and intercept them.

- Each AI-civ start-out locations can be allocated randomly from game to game on these sandtrap-maps.
- Each AI-civ can expand randomly (or according to game-situation) in any direction on standalone sandtrap-maps.

**tniem** · September 4, 2001, 13:00

Gramphos, nice example. I would assume the AI already currently does this to some extent.

There is a problem though with your formula.

First it never takes in account where other cities are located. Sure the AI probably never will be able to strategically place cities near you to use as spring boards for attack. Or cities that will be used as airstrips but that should be possible. Also where is the punishment for overlap. It should still be possible to build cities in certain locales with overlap, but there should be a negative number in the equation if the AI plans to do so.

Second, with the new resource system, there needs to be equation for where resources are located. But it simply cannot be that the city has access to them. Instead it should have more to do with how close the city is to the resource, thus requiring less culture buildings to expand your borders.

Finally, that city locale is not ideal. So there needs to be some equation where the AI only goes to the highest number. But even that wouldn't be great because strategically sometimes you might want to be closer to a resource or be building north towards an enemy.

All these things need to be taken into account when building the AI city placement. Which is exactly why it is tough for the computer to pick very good city locations. And why the best games you play are of the MP variety.

**Gramphos** · September 4, 2001, 14:29

Originally posted by tniem
Gramphos, nice example. I would assume the AI already currently does this to some extent.

Maybe, it is quite simple. If so I would like the numbers for each terrain to be editable.

There is a problem though with your formula.

I know, I said it wasn't balanced.

First it never takes in account where other cities are located. Sure the AI probably never will be able to strategically place cities near you to use as spring boards for attack. Or cities that will be used as airstrips but that should be possible. Also where is the punishment for overlap. It should still be possible to build cities in certain locales with overlap, but there should be a negative number in the equation if the AI plans to do so.

I forgot the overlap punishment. In the beginning I thought of half the points for all overlapping squares, but never tested it, and forgot about it while posting. If that would move down an existing "own" AI city below the lower limit that position should not be used.
I will add that Ralf's system don't take this in account either.

Second, with the new resource system, there needs to be equation for where resources are located. But it simply cannot be that the city has access to them. Instead it should have more to do with how close the city is to the resource, thus requiring less culture buildings to expand your borders.

That is true, maybe points for resource/distance from citycenter (if distance is < 10) could work. (The resources need to have variable (and high) point depending on how mush the AI needs it)

Finally, that city locale is not ideal. So there needs to be some equation where the AI only goes to the highest number. But even that wouldn't be great because strategically sometimes you might want to be closer to a resource or be building north towards an enemy.

I know, I said it was a bad startup position of Civ2, but I didn't have time to start over and over again to get a good position. I also think that all numbers might need some tweaking, and if they use a system like this one they should have the terrain points editable.

All these things need to be taken into account when building the AI city placement. Which is exactly why it is tough for the computer to pick very good city locations. And why the best games you play are of the MP variety.

I know that too.

**tniem** · September 4, 2001, 22:20

Originally posted by Gramphos
I know, I said it was a bad startup position of Civ2, but I didn't have time to start over and over again to get a good position. I also think that all numbers might need some tweaking, and if they use a system like this one they should have the terrain points editable.

I know but what I am refering to is a formula to decide between multiple possible starting locations. Not whether the one you picked is a good one. It clearly is not the best but works. But how is the pc going to decide and that requires not only most points but also an element of strategy placement in how the game is moving along. I have no suggestion because I have no way of doing this in my own mind. But it is something that should be taken into account.

I know that too.

I know you do. In fact that bit was intended for all, to say MP is always preferable. If you have time to play and finish a game.

Announcement

More funnel-shaped spider-sandtraps in player-created maps

More funnel-shaped spider-sandtraps in player-created maps

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment