In the light of PTW I remembered my old pet peevee with the 'strange' maps the generator calls pangaea and archipelago. Since it might be a good idea to know what kind of map you start your PTW games on I did some tests with the map generator, using Civ3edit to give me 60 huge 'Pangaea', 'Continents' and 'Archipelago' maps each. Settings where 60% water, everything else standard.

I categorized these maps into suitable groups and found that my suspicions are true:
the map generator will usually not give you the type of map you expect.


here are the results for 'Pangaea' maps:

settings:
http://www.civfanatics.net/uploads2/pangaeatestsetup.jpg

I varied only the seed number.

categories:

'1': 'true' Pangaea: all strating positions are one the same landmass

'1+1' / '1+2': in addition to the huge landmass there is/are 1/2 island/s with room for 1 or 2 civilizations on it

'2': there are two major landmasses

'2+1' / '2+2': in addition to the 2 huge landmasses there is/are 1/2 smaller islands with room for 1 or 2 civs each

'3': 3 large landmasses, sometimes with a few smaller islands

'>3' 4 or more more or less equal landmasses


results for 2 times 30 tries:

... '1' .. '1+1' . '1+2' . '2' . '2+1' . '2+2' . '3' . '>3'

.... 9 .... 2 ..... 0 ..... 8 .... 1 ..... 1 .... 3 ... 6
.... 9 .... 5 ..... 2 ..... 9 .... 1 ..... 0 .... 2 ... 2


The idea of 'Pangeae' is (for me) that (almost) all civs start on the same large landmass. So for 16 players, '1', '1+1' and maybe '1+2' are OK, for 8 civs '1+2' might mean only 4 civs on the main continent - not acceptable.

So, for 16 players this gives 11 / 16 'Pangaeas' on the first / second 30 tries respectively, giving 27 out of 60 - less than 50%! :eek:

Pangaeas with all civs on the same continent - 'true Pangaeas' make up only 18 out of 60 maps - less than a third!!!!!! :eek:

all others rather fullfill the 'continents' to Archipelago definition!


one of the 'less ecpected' mpas:
http://www.civfanatics.net/uploads2/pangaea12.jpg

placeholder for continent maps results

First round of archipelago maps done - 30 times these settings:

huge
60% water

categories:

'1': all starting pos on 1 landmass

'1+': large majority of starts on 1 landmass, some isolated civs

'2': all starting pos on 2 landmasses which have land area and number of civs roughly equal to each other

'2+': large majority (i.e. at least 12) starting positions on 2 major landmasses, some isolated civs

'3': all starting pos on 3 landmasses which have land area and number of civs roughly equal to each other

'3+': as '3', but some isolated civs

'4': all starting pos on 4 landmasses which have land area and number of civs roughly equal to each other

results:

... '1' ... '1+' ... '2' ... '2+' ... '3' ... '3+' ... '4'

.... 1 ..... 18 .... 1 ..... 6 ....... 1 .... 2 ..... .1


in most cases, the exact opposite of the desired effect (isolated civs / groups of 2 to 3 civs) happened.

Only 1!!! map resulted in 4 groups of 3 or 4 civs - not a single map gave more groups.



Since this may be the result of using 60% water I will now try 80% water.



2nd round: I used 80% water this time.

grouped the maps in these categories:

'non-archipelago': too many civs start on the same continent/island

non-5 --> 5 civs on one island
non-6 --> 6 civs on the smae island
and so on

non-4 is borderline IMO - I still counted them as non-archipelago

all other maps (i.e. maps with at most 3 civs on one island) are grouped by number of islands with starting positions on them. the lowest possible number is 6 - 5*3 and 1*1 civ = 16 in all.

a-6 = 6 islands
q-7 = 7 islands and so on

results:

non-4 = 7
non-5 = 9
non-6 = 2
non-7 = 4
non-8 = 3
non-11 = 1

a-9 = 2
a-10 = 1
a-12 = 1

all in all only 4 out of 30 maps are 'true' archipelagoes :mad:

Wow, Killer....great analysis. :goodjob: :goodjob:

This has always been one of my pet peeves as well...but more so with Island maps....they always seem more like Continents than I think they should.

These are amazing results. Has Firaxis ever confirmed that the map generator in the editor is EXACTLY like the one in the game? I'm curious if the in-game generator gives the same results.

SirJ: I feel (!) they are indeed giving the same proportions of continents, pangaeas and islands. The editor DOES gove better resource spreadings though.

Just wait until i have the archipelago results finished later tonight or tomorrow......

I just want you to know that I think it's messed up that you've figured that much of it out. Not in a bad way, but in a "Wow! You sure like Civ 3." way.

Kudos on the info, though.

RobOz: I actually did it for archipelago - haven't gotten a 'true' archipelago yet :( and I'd love to play one NOT made myself in the editor.....

I've only gotten 'true achipelago' maps in tiny worlds... Even then, there's 1 or 2 large islands.

I've only been able to get a good island map once, and I had to edit it to get what I wanted. I created an initial map in the editor (it took more than 30 tries). After that, I had to separate some of the larger landmasses into multiples.

After a couple hours of work, I finished a huge map, with 16 Civs and no more than 2 Civs on any one landmass (some were single Civs). That means I had at least 8 separate islands. There are actually quite a few "small" islands (room enough for 3-5 cities) that may be fun to add even more Civ's to.

It was a very fun map to play...but was not easy at all. It would make an interesting GOTM. If anyone is interested in the file, let me know.

http://www.civfanatics.net/uploads2/arch_pic.jpg

Someone at 'poly opsted a screenshot of a standard size archipelago map with all 8 civs starting on the same landmass. As it turned out, increasing the value "Average distance between civs" somewhere in the editor helped. Loading the same map, the civs were now scattered all over the map. You might want to try it.

And it is an excellent analysis.:goodjob:

Zero-Tau: but that would mean moding :( I'm not against it, but it makes matters more difficult with PTW since then everyone would have to make sure he know what to expect from the modded version......

Yeah, Killer, I've noticed similar results. It seems the game favors continents over all other maps. I'm fine with a bi-continent pangea though, but any more is too odd for me.

CG

Too many times I hunted that last civ all over a pangea only to find they had moved to an island and I would have to build a boat to catch them!

Too many times I started England in an archipelago to take "advantage" of the headstart toward mapmaking and sucky UU only to find myself in a land war with Germany and France while a cluster of other civs moved on into medeival times on another big landmass.

Nice job, Lt. Killer - it seems like you have too much time on your hands but this is coming from the guy who wrote a "best of the 1000 clues" article...

Killer I remember you had a simliar thread on this topic some time ago. Great analysis.

Somewhat off-topic: but does anyone notice that the "culturally link starting locations" option doesn't really do much? I have it on everytime but in an example I'd still see the Romans to my left, the Japanese to my right and the Russians above, and finally the Americans below which is the only one related to my Iroquis. It makes me wonder if the Americans just got there by coincidence or not.

Well, the Iriquois have the least to choose from in their cultural group, so that might not be too weird. Try a 16 player setup and see how it looks.

At least you didn't write a utility to check pathfinding between starting locations... for a scoring formula. :)

From what I understand of the map generator, the only real difference between landform types is the number of "continents" it builds, the type of each continent (ie. land or water) and the number of tiles in each continent. For pangaea maps, there are fewer continents of larger size, and archipelago has more continents of smaller size. I'm not sure, but it doesn't seem like there is any other difference based on landform settings.

Having spent some time coding various types of map generators, it isn't all that easy to get things coming out right. Keep in mind that right to the programmer isn't always going to be right to each customer.. It's difficult. The map generator in Civ III is one of the best I've seen for a tile based game. If you really want to know what the map you're playing with will look like, there is always the editor.

Originally posted by Lt. 'Killer' M.
The editor DOES gove better resource spreadings though.


Hi Killer.

"Better" than what? I always have to mod to get decent resource distribution when I have less than the max number of civs. For example, on a huge map with 8 civs, I have to double the appearance ratio on everything.

In my opininion, the resource distribution does not work very well unmodded under those conditions.

When you say it doesn't work well do you mean that not all civs start out with an even amount? Becuase if thats the case I disagree. I think its good that there are clusters in one place and not in others. Makes trade realistic.

Originally posted by Yzman
When you say it doesn't work well do you mean that not all civs start out with an even amount? Becuase if thats the case I disagree. I think its good that there are clusters in one place and not in others. Makes trade realistic.

I agree! Isnt that one of the main points of having luxuries/resources? Go to war to get a resource another Civ has, etc?

on the resources: I hate it when everyone has one of everything in their core empire of the strat res, and when there yre NO lux whatsoever on islands without starting locations :( This happenes less often when i use the editor.

nice one killer!

Originally posted by Yzman
When you say it doesn't work well do you mean that not all civs start out with an even amount? Becuase if thats the case I disagree. I think its good that there are clusters in one place and not in others. Makes trade realistic.

No, no, no. The distribution is random, and that is how it should be.

This is the problem: If you have 8 civs on a huge map, the map puts 50% of the resources that would have appeared on the same land area if there had been 16 civs. I think resources should be distributed randomly, but why should the resources be more scarce when there are fewer civs? You may think it's fun to have to build a colony on an iron resource 50 tiles away from your cities, but I don't. In my opinion, the map should distribute resources according to the appearance ratio in the editor on the land, but it does not work that way.

I admit that I can get the correct resource distribution by editing the appearance ratio in the editor sometimes, but I don't think it comes out right even with the max appearance ratios for 4 civs on a huge map. This is a pretty tedious way to start a game, editing the resource appearance ratios, but it's better than playing for several days and then discovering that your map is ****ed.

This also changes trading drastically (unless the editor's max appearance ratios are big enough to make the distribution right).

In 1.29f the map generator seems to be better in one way though: On 1.21f sometimes 2 of the luxuries would be completely absent on a huge map. I haven't seen this problem in 1.29f (yet).

Good job Killer. You and I have both described problems with map generation. One way to solve this is to have an impartial human, who will not play in the game, to create a playable map.

It's frustrating when you play for several days and then realize that the map wasn't what you intended at all.

:goodjob:

I suppose a "happy place" would be a combination of the fantastic civ 2 archipelago editor and the almost (but not quite)perfect continent editor of civ 3 plus a hypothetical magic pangaea editor on civ4. ;)

Originally posted by sumthinelse
Good job Killer. You and I have both described problems with map generation. One way to solve this is to have an impartial human, who will not play in the game, to create a playable map.

thanx. Yes, that would work - and it does as far as user created maps are feasible - it takes an awefull amount of time to get one 'right'.

It's frustrating when you play for several days and then realize that the map wasn't what you intended at all.

exactly! I once had a game where I delayed naval units since it was obviously pngagaea - only to find all 6 AIs had 1 city each on a tiny island with exactly 6 sources of uranium - all there was in the game :(



continents are gving me trouble grouping them atm - the results will take a while and a far larger data base.

Don't bother. The continents are lovely!

(Though they could be a bit morre raggedy(?) at times.

There was one time I chose Archipelago with 60% and it gave me this huge, thick band of land, all connected that spanned the entire world, with only like a north and south sea, and a large bay with one tiny island off a peninsula. I expected a bunch of larger islands or a helluva lot of nice sized islands.

Originally posted by The Yankee
There was one time I chose Archipelago with 60% and it gave me this huge, thick band of land, all connected that spanned the entire world, with only like a north and south sea, and a large bay with one tiny island off a peninsula. I expected a bunch of larger islands or a helluva lot of nice sized islands.

The '1's and '1+' are all of that kind :(

Hey, Killer. Do you have PTW? If not it would be very interesting if someone who does replicated your work on the new version of the map generator.

:( not yet - I hope sabo10 will help me out on PTw.....

but I WILL do the PTW part ASAP