How random is Random AI civ selection?

[andreasb]

I always play with Random in the AI civ fields in Custom Game. However, there are some leaders that are almost always present in my games, and some that I rarely see. Is the algorithm for this truly random, or is it based on which Civ I choose and other factors?

I probably roll Shaka and/or Toku in 75% of my games, and I don't think I have ever seen Peter (Russia) or De Gaulle (France) in 20+ games. There are some others that I see often too, and some I rarely see, but the four cases above really stand out.

I usually play Elizabeth, Darius or Inca if that matters. Oh, and I mostly play standard maps (6 AI opponents?).

 

[Kesshi]

Is the algorithm for this truly random

You're potentially going to open a can of worms with that question.

But to answer your original question, the enemies used are NOT based on any leader you choose and are as random as expectable.

 

[Gorey]

i always wondered this too. if its just purely random, or if it weights things.

 

[Badtz Maru]

You're potentially going to open a can of worms with that question.

But to answer your original question, the enemies used are NOT based on any leader you choose and are as random as expectable.

I remember in Civ3 the AIs were grouped by their traits, which also corresponded with which region of the world they were from in most cases, so European civs were usually close to each other. Is there anything like that in Civ4?

Another thing I noticed is starting points seem related to the civilization you chose. It seems like a disproportionate amount of time you will start with elephants if you are Indian, and silk if Chinese, etc.

 

[TheMeInTeam]

It "weights" leaders that are the only option for their civ.

You have an equal chance of meeting any civilization. However, for civs like america, the leader is again random but one of three possibilities. So while you will see zulu and america equally often, you will see shaka a LOT more than lincoln, because you could meet washington or roosevelt instead of lincoln to represent america.

Other than that it is as random as the RNG gets.

 

[FrozenTitan]

As with many random things, the bigger the field is, the more random it actually is. if you only play 20 games, it could seem as though you get one of the same leaders a lot. but expand that field to 200, and you might not see him again. it can seem flawed at times, but it's pretty random in the grand scheme of things.

 

[NKVD]

i also find Shaka, Alexander ,Toku and Mehmed in all my games...I dont know what i have with Shaka...he have been in all my last 5 games i think

 

[Patak]

Shaka, Gengis Kan, Charlemagne, Mehmed, Hatepsut are almost always there. Never played against DeGaule, Churchil, Mao Tse or Washington.

 

[TheMeInTeam]

The above 2 posts do not answer the OP's question and are misleading .

Nobody has a higher chance to roll any civ.

 

[Badtz Maru]

The above 2 posts do not answer the OP's question and are misleading .

Nobody has a higher chance to roll any civ.

We should test this. Civ games have had hidden mechanics before, it's not unreasonable to suppose there may be some unknown factor weighting which civs are selected. In the previous Civ game, there were known and documented factors that weighted which civs were picked. And even if it is supposed to be totally random, it's known that sometimes RNGs have flaws that cause them to behave in non-random ways under certain conditions.

It should be easy enough to do. Someone with a fast computer (I may do it myself if I find the time) should repeatedly create maps with 5 or 7 civs and record which ones appear, for a hundred games or so. Then graph it and see if there's any spikes.

I'm pretty sure mine would show a big spike for Hattie and Tokugawa. They seem to be in every one of my games.

 

[TheMeInTeam]

You can't use leaders. You have to use civs.

Obviously there *will* be bias towards toku, shaka, monty, and all other leaders who are the only leaders for their civs, because it's random based on civ rather than leader. If you measure it by civ there's unlikely to be any discrepancy.

I only have 512 MB of RAM and a 128 MB video card ----> not exactly good specs for running 100 test games . But, I'm pretty confident in the RNG here anyway having literally played 100's of games start to finish over my time on the forum.

I'd be interested to see bias based on civ someone chose, I know colonies have bias in this regard but IIRC the code divers said that for random civ selection for game start that doesn't occur.

 

[Badtz Maru]

You can't use leaders. You have to use civs.

Obviously there *will* be bias towards toku, shaka, monty, and all other leaders who are the only leaders for their civs, because it's random based on civ rather than leader. If you measure it by civ there's unlikely to be any discrepancy.

I only have 512 MB of RAM and a 128 MB video card ----> not exactly good specs for running 100 test games . But, I'm pretty confident in the RNG here anyway having literally played 100's of games start to finish over my time on the forum.

I'd be interested to see bias based on civ someone chose, I know colonies have bias in this regard but IIRC the code divers said that for random civ selection for game start that doesn't occur.

There's only one leader for the Khmer, but I have only had him appear in one of my games. Hattie has been in EVERY game I've played since I got BtS.

I think that if Japan and Egypt (and always the same leader for Egypt, I NEVER see Ramses) show up in practically every game (and I usually only play with 5 civs), that indicates some kind of non-randomness. Some civs seem to be picked more often, and some leaders seem to be preferred for some of these civs.

 

[Lansky]

I cannot remember the last game I played where the AI was an Egyptian leader. I have not gone against Shaka in a few weeks at least. My last 5 or 6 offline games have featured me starting in close proximity to Mansa Musa. Seems like a standard streaky randomizer to me.

 

[TheMeInTeam]

There's only one leader for the Khmer, but I have only had him appear in one of my games. Hattie has been in EVERY game I've played since I got BtS.

I think that if Japan and Egypt (and always the same leader for Egypt, I NEVER see Ramses) show up in practically every game (and I usually only play with 5 civs), that indicates some kind of non-randomness. Some civs seem to be picked more often, and some leaders seem to be preferred for some of these civs.

Could easily be observer bias, you know. In fact, logically you should tend toward that way of thinking and assume it so, rather than the other way around. People are wrong on this matter far, far more often than they are right.

Just as rolling egypt 5 times in a row is a somewhat rare occurrence, people lose 99% battles. You can't use your limited number of observations (which may or may not have been accurately recorded, but let's assume they were) to make any valid conclusions.

 

[Badtz Maru]

I'm aware of observer bias, but we can't assume that it's totally random without testing it.

 

[TheMeInTeam]

The burden of proof is on those claiming it isn't random, not the other way around.

 

[Badtz Maru]

The burden of proof is on those claiming it isn't random, not the other way around.

Why is that?

 

[TheMeInTeam]

Why is that?

Because the game (and its programmers in charge of the RNG) imply it is random. Directly. Doubts are expressed here from time to time, but it has been quite long since anybody has actually proved the RNG isn't accurate (last I knew of was very early on when the odds displayed for combat didn't factor first strikes properly).

 

[Kesshi]

Why is that?

Because the game imply it is random. Directly.

Or it could be just because those of us claiming it's random don't care if it isn't "truly" random, because it's random enough for us. After all, who was the one who created the thread?

 

[Badtz Maru]

Because the game (and its programmers in charge of the RNG) imply it is random. Directly. Doubts are expressed here from time to time, but it has been quite long since anybody has actually proved the RNG isn't accurate (last I knew of was very early on when the odds displayed for combat didn't factor first strikes properly).

Imply? Directly?

How do you imply directly?

What is the exact mechanism for choosing rulers from the RNG? Is one number generated and an algorithm used to determine the rulers from that one number, or is a new number generated from the seed for each slot to roll? Is a reroll done when a civ that was already selected comes up or does it move down the list to the next rolled leader? Always, or in just certain conditions?

There are too many unknown factors to assume that it is 100% random.