Irrigation Rules

Sitting Bull · February 6, 2001, 14:38

In civ2 and ctp there's a rule that you cannot irrigate unless a water source is adjacent to the tile being improved, this is a very bad tradition of the civ series which really should be broken in civ3 because it is inconsistent, unnecessary, and hinders the AI.

1. If the body of water is "ocean", it is highly inconsistent because nobody irrigates with seawater. We are forced to accept that all water is freshwater lake, then we play a cylindrical map, and build wonders like Magellan's voyage. We colonize space with hydroponic farms all the while completely unaware of the giant underground aquifers with billions of gallons of water for anyone who wants to dig a well.

2. The rule is completely unnecessary because even if we are using a river as the source, we must understand that the scale of the map is such that only the largest rivers appear as such on the map. Most large rivers such as the Danube or Colorado have many hundreds of tributaries which wouldn't show up on the map, but would be available for use some distance from the main channel. In some places more, in some places less, in a few places not at all (such as the lower Nile); which is part of the difference between the generalized "grassland", "plains" and "desert". So what exactly IS an improved tile? Since it has a technological prerequisite, it must be improved tools and methods rather than canals and pipelines? Practices that can be applied *anywhere* that farming can be practiced, if the civ so evolves. They don't need to be adjacent to an ocean coast to use crop rotation methods and fertilizer.

3. Most importantly, the rule lowers the quality of gameplay because it hinders the construction of better AI which might be better able to adapt to different maps. Look at the behaivior of the AI in ctp: it will do reasonably well along coastlines, often bulding many nets, but doesn't have a clue about how to place cities to bring water long distance to a mining district. In part this is a problem with the pw system since there is no wandering engineer, but if the rule is taken out each city could just decide for itself when to improve land, with no need to wait on a random-walking artificial idiot.

This irrigation rule is a dumb legacy that I really hope is scrapped in civ3; it's overkill, please take these excuses to drop it and help write a better game engine.

Henrik · February 6, 2001, 14:42

Still... It has been in there since Civ 1.
I think I agree however.

Ralf · February 6, 2001, 15:38

quote:

Originally posted by Sitting Bull on 02-06-2001 01:38 PM
3. Most importantly, the rule lowers the quality of gameplay because it hinders the construction of better AI which might be better able to adapt to different maps. Look at the behaivior of the AI in ctp: it will do reasonably well along coastlines, often bulding many nets, but doesn't have a clue about how to place cities to bring water long distance to a mining district. In part this is a problem with the pw system since there is no wandering engineer, but if the rule is taken out each city could just decide for itself when to improve land, with no need to wait on a random-walking artificial idiot.

I certainly agree about the big (but, often underestimated) importance of doing something about the problems you are describing. Imagine the following experiment:

Lets say that any AI-civ and any half-descent civ-player starts out on each exact clone-island. You can pretty much bet that by the time the islands is clogged with settlements/early cities, the human player have about 50-100% (or more) potential future shield/food/trade output-advantage, as the AI-civ has, comparing those two clone-islands with each other. And this because of two simple reasons:

- More efficient city-placement strategies.
- More effective city-area developments.

"There are however, much more effective ways to counter-act these AI-problems, then by allowing irrigation anywhere and everywhere. I have copied below from another post (my own, by the way):

QUOTE:
"I really hope, and I really keep my fingers crossed that a Civ-2 style "fully featured" AI-settler won't make it in Civ-III. Firaxis really must reconsider the absolut hopeless idea of moving around stupid AI-settlers in order to...

- Manage and develop AI-city areas.
- Expand & found new AI-cities on the map.

The former can be achieved by some kind of automatic AI-city area maturing-process instead. This maturing-process can be tweaked by the player for each AI-civ, if hes not happy with the tile-type emphasize and/or the speed of the maturing-process itself.

The latter can be achieved by (directly after the map has been freshly generated) letting the AI calculate, and then "perforate" the whole map, from top-left to bottom-right, with unlit red diodes (like unlit diodes in electrified public tourist-maps that one can see in big city central railway-stations).
These unlit red diodes (= potential, yet uninhabited AI-city placements, currently invisible for the human player) can be used by the AI-civs in order to expand & make use of the land much more efficiently, then Firaxis ever can hope to achieve by physically moving around AI-controlled settlers. When a new AI-city is founded; thats the equivalent of a previous unlit red diode, now being enlightened.

Scenario-creators could manually place those "unlit red diodes" (= potential, and for the time being; invisible AI-city placements), and by that get access to a whole new realm of scenario-tweaking possibilities, currently unavailable for him because of the built-in "Achilles heel" of the notoriously bad AI-settler pathfinding abilities.
The whole idea is much more elaborately explained in the Should the map-generator be scrapped thread.
(Never mind the title. I changed my mind about that pretty quickly. Read further down about the possibilities of pre-designated AI-city areas instead).

I regard above as one of the most important AI-strengthening suggestions I have ever made. And the scenario-creating possiblities really cannot be underestimated."
END QUOTE.

Any comments? Any thoughts about the "pre-designated red diode AI-cities" idea?

[This message has been edited by Ralf (edited February 06, 2001).]

Sitting Bull · February 6, 2001, 16:04

My only comment here, all the AI needed in that regard would be a point system to calculate for each tile, summarizing the food/shield/gold potential of a city placed on that tile. It could be done during exploration, and need not be pre-scripted. Just a simple point rating system, defined so that a giant glacier gets a low score.

EnochF · February 6, 2001, 16:42

In Civ2, the AI could irrigate a square whether it was next to a water source or not. Try playing a game for roughly ten turns, then turning on cheat mode and revealing the map. You'll see. So, the rule doesn't hinder AI.

But I'd be glad to see the rule go, anyway. I mean, a terrain "tile" in Civ represents hundreds of square miles. Surely there must be a river or stream or underground water source somewhere.

Ralf · February 6, 2001, 17:08

quote:

Originally posted by Sitting Bull on 02-06-2001 03:04 PM
My only comment here, all the AI needed in that regard would be a point system to calculate for each tile, summarizing the food/shield/gold potential of a city placed on that tile. It could be done during exploration, and need not be pre-scripted. Just a simple point rating system, defined so that a giant glacier gets a low score.

The biggest problem is that most civers just dont understand the built-in AI-pathfindings and data-quantity problems, in games with the complexity-level of Civ-3.
Also, most civers just read the word "pre-", in "pre-designated", and they almost instintively shy away from it, without really bother to read through and understand the enormous hidden potentials with this "pre-designated unlit/lit red diode AI-cities" idea.

I can only hope that the Firaxis-team understands this better. I think (and I hope) they do.

Grumbold · February 8, 2001, 11:11

Predesignated city sites could be good news for scenarios. I don't see how they benefit a random map game unless the AI is allowed to cheat and head for "diode" spots without having to explore the map to reveal the world. If they do have to explore then calculating the relative values on the fly would seem to be perfectly acceptable unless you are actually going to calculate best placement for whole series of cities rather than each on its own merit.

Stuff2 · February 9, 2001, 07:42

I think irrigation as a land improvement are only neccessary when the land is dry. Up here in northern europe we hardly have any irrigation at all in the agriculture, the weather is wet enough anyway.

I would like to see water be implemented as a resource. The farming abilities on every square would be close connected with how much water there is. Also, a city would be able to farm desert squares if they can provide water from another square.
The irrigation improvement would simply fetch water from the citys surplus of water.

Zealot · February 9, 2001, 11:22

I agree with you Stuff2. Yet, I would like to remember you guys that Colonization had something called "Plow". Remember? But I confess: I don't like those boring settlers, but I hate even more Public Works! Because I keep forgeting to improve my cities... =)

Grumbold · February 9, 2001, 12:31

Irrigation seems to represent the organised use of land for farming so I agree it should not need to be connected to a major river on grassland. I am less sure of that for plains, scrub or desert terrain where there really isn't any surface water for miles sometimes. It isn't pratical to irrigate thousands of square miles of scrub and desert even with modern methods. I could cope with different rules about how easily you can irrigate/cultivate different terrain types but the current system is probably better just on KISS principles.

Ralf · February 9, 2001, 14:56

quote:

Originally posted by Grumbold on 02-08-2001 10:11 AM
[quote]I don't see how they benefit a random map game unless the AI is allowed to cheat and head for "diode" spots without having to explore the map to reveal the world. If they do have to explore then calculating the relative values on the fly would seem to be perfectly acceptable unless you are actually going to calculate best placement for whole series of cities rather than each on its own merit.

It is much easier to program AI-scouts just to uncover yet unexplored black map-tiles, then it is to program AI-settlers to found ideally effective AI-city placements. It really is a big difference between the two tasks, in terms of AI-workload.

The HUGE cheat-advantage any Civ-player have over the AI-civ programmers at Firaxis, is that the player have the ability to overview the map (= experience it), and then within a blink of an eye, sort out huge parts of less-promising potential city-placements, thus allowing him to concentrate directly (and only) on those potential city-placements that actually DO seems promising.
The AI-programmer simply CANNOT immitate that ability on fly, on random (= unknown) huge-scale Civ-maps. He is instead forced to program the software to calculate any AI-settler pathfindings without that "overview-ability".
From the AI-programmers point of view, the look of the uncovered map (and the game situation, in general) might as well be a 2-dimensionally and completely "flat" world - a world without that vital living-being-unique third experience-dimension.

And now to the second problem:

Lets say that the AI-controlled settler find its way to a good fertile spot on the map. Can the programmer design the AI to calculate the surrounding 20 potential city-square "on fly"? Yes, sure - but the problem is that...
if the programmer have the ambition to place those AI-cities just as efficient as any civ veteran-player (using the traditional "move around AI-city founding AI-settlers" method); he must also let the software/hardware meticulously calculate any other potential AI-city placements surrounding him, within a radius of at least 5 squares away. Also, both the existing (potential) AI-city-placement and the surrounding 5 squares away potential city-placements, have to be compared with, both the already existing cities/city-areas and the irregular coastline as well. And he MUST calculate & evaluate above before the AI-settler actually decides to found an AI-city on that spot he is currently standing on (otherwise it can be too late to make further placement-adjustments, of course).

Now, add to above that the AI is suppose to handle sometimes upto 5-10 AI-city founding settlers simultaneously, and add the fact that the AI have many other, equally important (and more important) tasks to attend to, within that same game-turn - and within that same limited game-turn waiting time-frame.

You see the potential AI bog-down calculation problems now?

This is where the idea of calculating ALL above potential AI-city placements AT ONCE in conjunction with the freshly generated map instead, before the game actually starts. Remember that...

ANY AI-civ can start randomly from ANY pre-designated spot, and expand randomly in ANY direction. Theres NO pre-designated AI-capitol city starting-points whatsoever, and not necessarily any pre-designated non-tweakable AI expanding directions/strategies either.
Also, the AI-civs can only expand to those pre-calculated AI-city spots that have been uncovered by their own AI-units during the actual game. Theres is NO AI-cheating involved; The AI-civs still have to produce the equivalent of an settler, and they still have to wait so-and-so many turns (simulating the AI-settler journey) before that new AI-city can be founded on that pre-designated spot. Finally, if a pre-designated AI-city spot currently is within any forreign borders, the game-AI adjust to that dynamically - of course.

quote:

Predesignated city sites could be good news for scenarios.

I certanly agree.

Read below and imagine.

A: These pre-designated AI city-placements can be as many (or as few), as the map-maker wants them to be.
B: They can be manually placed by the map/scenario-creator, and they are invisible for the human player.
C: They can also be calculated and distributed by the map-generator in the beginning of the game, and...
D: they can apply to AI-civs only, or both AI-civs and the human player. Its the players own choice.

That player can choose to play on either hand-edited, computer-generated or scripted scenario-maps, there...

1: the scenario-creator can choose to pre-designate AI-cities only, or both AI-cities and human-player cities.
2: the scenario-creator can choose how many global city-placements is allowed, and where they shall appear.
3: the scenario-creator can script where each AI-Civ (or HP-Civ) shall found any new cities.
4: the scenario-editor can script when, each AI-civ shall found new cities. HP always decides for himself, of course.

Complete scenario-making freedom!

Its like adding a second time-dimension added to the previous old place- dimension. More explanation on this clickable link.

[This message has been edited by Ralf (edited February 09, 2001).]

Sitting Bull · February 10, 2001, 02:11

Say, send the settlers one at a time to the highest scoring tile within x number of tiles from any existing city. Recalculate point scores accounting for new overlap (quick) and repeat. Should be good enough. And, I'm not so sure that I want the extra work in scenario design. Just let the AI have it's own way, a lot easier. Incidentally, how does this help the AI learn how to take water to a mining district?

February 10, 2001, 03:22

Call To Power 2 has change the way to improve a tile. You do not need to be next to water any longer. I'm sure Firaxis has look at their method and will come up with something close. I'm not saying PW or a settler, just something close. Personally I liked the settler or SMAC method because you could put him anywhere on the map and do something, build roads or other improvements. With CTP 1 or 2 method you can only build in your own territory.

------------------

Ralf · February 10, 2001, 05:12

quote:

Originally posted by Sitting Bull on 02-10-2001 01:11 AM
Say, send the settlers one at a time to the highest scoring tile within x number of tiles from any existing city. Recalculate point scores accounting for new overlap (quick) and repeat. Should be good enough.

It wouldnt be near enough to imitate the efficient city-placements of any half-descent civ-veteran player. What you basically saying is that its OK that the civ-player have about 50-100% (or more) potential future shield/food/trade output-advantage, comparing with AI-civs on any similar-terrain/similar-sized land-areas. Well, I for one DONT think its OK - and I hope Firaxis deals with AI-city placements and AI-city area improvements without using weak AI-settlers, thus bypassing the problem.

quote:

And, I'm not so sure that I want the extra work in scenario design.

Firstly: its partly optional. These pre-designated AI city-placements can be as many (or as few), as the map-maker wants them to be. Secondly: please lets not exaggerate the "extra work", shall we. Already today any map-creator spends time on fine-tuning each island/continent by re-arranging terrain-types and coastlines. What hinders him from easily placing (point-and-click) those pre-designated spots while hes doing above task?

quote:

Just let the AI have it's own way, a lot easier.

I am constantly amazed how some civers are willing to more or less cram a whole world of parameters into the Civ-3 game (in the name of making the game "more realistic") - yet, when it comes to ideas that really could help the AI-programmers to achieve a much stronger AI-food/resource/trade production-capacity from an given land-area (without cheating) - and, by that achieve much better strategical/logistical game-positions for the AI-civs: The response from (some) civ-gamers is, at best lukewarm and halfhearted.
How low can the expectations be from (some) civ-players on how one can make the AI-civs significantly more competitive - at least in all-important strategical/logistical areas of the game?

Another problem with the old model, by the way, was that many times in late end-games, this or that AI-civ had only populated a part of a big island, despite the fact that the rest of the island still was uninhabited and full of potential resources. This is another reason why stronger and more efficient AI-city placements are absolutely vital and necessary.

quote:

Incidentally, how does this help the AI learn how to take water to a mining district?

What are you talking about?

Are you talking about why mined hills/mountains currently dont have to be near water? This is because the game is a... well, only a game. Much can be implicit and every real-life aspect doesnt necessarily have to be translated to the game-screen.

A more appropriate question would be: How does AI-civs expanding beyond oceans?

Well, these invisible pre-designated AI-city placements, could easily be conditional. In order for any AI-civ to expand over the oceans, they GOT to have the necessary techs and available ships.
As long as any AI settler-journey is simulated across any ocean, those ships used are faded out; Not available for any other AI-use.

PS: before critisizing the "pre-designated AI-city placement" idea any further; be sure to read through this clickable link first. It can be quite tiresome having to re-formulate answers to objections that already have been explained previously in that thread. Constructive criticism from people that have read and understand above link, and who not necessarily see things instinctively in a "cup half-emty" kind of way - is however more then wellcome.

PPS: I think that the "irrigation only near water" condition is a good and sensible idea. It shouldnt be too easy (= less challenging) to improve the food-outputs on land-tiles. Keep this condition the way it was in Civ-2.

[This message has been edited by Ralf (edited February 10, 2001).]

Sitting Bull · February 10, 2001, 08:38

What am I talking about? I'm talking about the behaivior of the AI in ctp, as I said when I started the thread, before you hijacked it, it doesn't have a clue how to bring water long distance overland. The cities won't make enough surplus to feed a lot of miners and still grow. You would force cities to be next to water to use fertilizer, crop rotation, improved plows and tractors? I can do it, but the AI can't. So, it doesn't make the game more challenging.

Ralf · February 10, 2001, 10:31

quote:

Originally posted by Sitting Bull on 02-10-2001 07:38 AM
before you hijacked it,

Then I start a topic of my own I do that in order to promote an idea, yes - but anyone who stays within that subject is completely free to join, agree or criticize it, and instead suggesting alternative methods in how to deal with the described topic-problem. This way one can debate different viewpoints within that thread. As a result of that the forum becomes more vibrant and worthwhile.

You obviously dont like that, and; OK - I stay away from Sitting Bull topic-starter after this post then.

quote:

I can do it, but the AI can't.

Yes, but the suggestions you lay forward in order to rectify the problem is totally inadequate - at least seen from my viewpoint. I thought you wanted to discuss it and be challenged with alternative methods dealing with it. Obviously I was wrong.

[This message has been edited by Ralf (edited February 10, 2001).]

Straybow · February 11, 2001, 07:13

Somehow I doubt that computational power or time are the limitation for AI of this sort. The problem is creating an algorithm to weigh the possibilities and make a decision. What are the relative valuables of trade, shields, and food?

Food requirements are easily forecasted (2*citizens) so the AI has to decide what size is manageable. But for trade and shields the decision is more complex. Their values aren't linear. You need a minimum level of each, but above that increasing one at the expense of the other is hard to evaluate.

Then there are the less tangible factors: proximity to hostile or friendly neighbors, suitable transportation routes, strategic defense points… these are the things living players can evaluate visually without fully realizing the process going on in our noggins.

I'm afraid that pre-plotted city locations would only work for scenarios. Shifting the AI burden from the game engine to the map building engine doesn't make the algorithm any easier to write.