How random is Random AI civ selection?

[Badtz Maru]

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?

The person who created the thread was someone who saw that it did not appear very random, and was wanting to know how random it was. They got a bunch of people saying "It's totally random, trust us" and a bunch of other people saying "It seems awfully streaky to be totally random".

Then we get "We don't care if it's random, it's random enough for us (but trust us, it's totally random)".

There's a LOT of ways to generate random numbers, and not all methods are equal. I'm not going to assume that the method Civ4 uses and how it interacts with these other factors is going to be the absolutely best and most perfectly random system just because some people don't care.

See Benford's Law. Completely random numbers will statistically have certain digits show up more often in certain positions of the number. Let's suppose (which is all we can do as we don't know the details of how civilizations are chosen) that a totally random large number is chosen when choosing civilizations, and it's worked through digit by digit to choose which civilizations fill the slots - if the first digit is X, then civ Y is in slot 1, if the second digit is B, then Civ Z is in slot 2, but if it's X again, it's D. Now, that's not going to be exactly it, but this is just a hypothetical example. In that case, it's not going to be a totally random distribution, certain civs will show up more often. There's other possible methods that might make different civs show up more frequently when you use a certain civ for the player, for instance.

Now, the designers probably did not make the randomness of the determination of the civilizations in a game a very high priority. Why should they? Players can just choose who they want, if they want. As long as the random method is not so broken that it never chooses certain civilizations or always chooses the same ones, it's good enough. And honestly, that is good enough...but if it is a imperfect system then that is going to be noticeable to players, and some of them will have questions about how random it is. Saying that it is completely random (and it is almost certainly NOT 100% random as there are very expensive programs and technologies for getting randomness higher than what you can get from a RNG on a PC) is not a valid answer. We know it's not 100% random, the question is how random is it. "Random enough for me" is not an answer.

 

[Continental Op]

Same leader in 20+ games doesn't really mean anything (unless it comes to you - that would be very impropable indeed). I once got Pacal II as my randomly chosen leader three times in a row. The third time I regenerated, because I was starting to think that something was wrong. This time I got Justinian, again two times in a row. After that, I've never been given the same leader more than once in a row. It sometimes seems that some leaders appear more often than others, and it can be annoying at times, but it doesn't mean it isn't random.

 

[Badtz Maru]

Same leader in 20+ games doesn't really mean anything (unless it comes to you - that would be very impropable indeed). I once got Pacal II as my randomly chosen leader three times in a row. The third time I regenerated, because I was starting to think that something was wrong. This time I got Justinian, again two times in a row. After that, I've never been given the same leader more than once in a row. It sometimes seems that some leaders appear more often than others, and it can be annoying at times, but it doesn't mean it isn't random.

It's not 100% random, unless Firaxis has invented some new method of random number generation that can run on PCs that is better than the best RNGs science has created. What is questioned is the degree of randomness. I've explained several ways that a simple random number generator could create patterns of non-randomness in civ selection.

 

[foobarred]

OP made me think... I've played tons of games Standard map, random everything, and I don't recall ever, EVER meeting Winston Churchill.

 

[Kesshi]

The person who created the thread was someone who saw that it did not appear very random, and was wanting to know how random it was. They got a bunch of people saying "It's totally random, trust us" and a bunch of other people saying "It seems awfully streaky to be totally random".

Then we get "We don't care if it's random, it's random enough for us (but trust us, it's totally random)".

There's a LOT of ways to generate random numbers, and not all methods are equal. *snip*

F you, I already know about probability because I was in Mister Pennovich's class.

(for an explination, please see this post)


j/k on the "f you" part. But the fact is, I understad probability on a level most people don't (not you, Badtz Maru) as well as I've played more games than a lot of most people here (maybe not you). I have a few favourites when it comes to leaders, and I've noticed no such tendancy for the opposing leaders to be anything less than a reasonable* amount of random.

*Reasonable meaning that I cannot accuretly predict who my opponents will be.

 

[Grey Fox]

I always seem to get leaders who share one trait on the same continent...

 

[andreasb]

After all, who was the one who created the thread?

o_0 ?

 

[Racsoviale]

I think I read somewhere here that the random generator tries to at a minimum add one warmonger, builder, techer and REX'er. If this is true it is quite possible to meet Shaka and Monty fairly often. But in general I like it when I meet an aggressive civ, it spices things up.

 

[andreasb]

I think I read somewhere here that the random generator tries to at a minimum add one warmonger, builder, techer and REX'er. If this is true it is quite possible to meet Shaka and Monty fairly often. But in general I like it when I meet an aggressive civ, it spices things up.

Yeah, I don't mind it either. Was just curious how the selection worked. I find it much harder to play friendly civs only since they often tech/wonder whore much more than the warmongers. I recall having Pacal, Darius, Gilgamesh, Mansa Musa and Mehmed as opponents in one Emperor game. They all had some good land too and had impressive tech rates...none of the AI declared war at each other until 1500 AD or something like that..it was definitely one of the more challenging games I have ever played..guess I should have re-evaluated the decision not to rush early

 

[Öjevind Lång]

Sometimes a specific leader can turn up in one game after another, but that is no more remarkable than throwing dice and getting sixes five times in a row. It all evens out in time.

But I can also get the spooky feeling that some leaders are stalking me and others are avoiding me. I never seem to get Hattie any longer, and formerly, that nice little lady turned up a lot. (Ramses has also put in an apparance several times.) On the other hand, recently Shaka turned up in five games in a row, and every time on the same continent as I. That was even worse than the streak I had a couple of years ago when Montezuma seemed to have decided to always be my next-door neighbour.

 

[Badtz Maru]

I always seem to get leaders who share one trait on the same continent...

That's how it was in Civ3, and that factor was documented if I recall correctly. It also had the neat effect of having neighboring civilizations who were from the same geographic region in the real world, because all the European civs shared one trait, all the Asian civs shared one, etc.

Since it was documented in Civ3 and was not in Civ4, I kinda assumed they didn't do that this time, but I've never really paid attention enough to check. Of course, if this is true, that would definitely lead to some civilizations showing up more than others.

 

[Badtz Maru]

Sometimes a specific leader can turn up in one game after another, but that is no more remarkable than throwing dice and getting sixes five times in a row. It all evens out in time.

But I can also get the spooky feeling that some leaders are stalking me and others are avoiding me. I never seem to get Hattie any longer, and formerly, that nice little lady turned up a lot. (Ramses has also put in an apparance several times.) On the other hand, recently Shaka turned up in five games in a row, and every time on the same continent as I. That was even worse than the streak I had a couple of years ago when Montezuma seemed to have decided to always be my next-door neighbour.

Funny, I get Hattie all the time, but I NEVER get Shaka. Since getting BTS, I seem to get Charlemagne and the Ethiopian guy an awful lot, too.

If it's grouping them based on traits, you may get more of a certain leader if you prefer a certain trait yourself.

 

[Öjevind Lång]

Funny, I get Hattie all the time, but I NEVER get Shaka. Since getting BTS, I seem to get Charlemagne and the Ethiopian guy an awful lot, too.

If it's grouping them based on traits, you may get more of a certain leader if you prefer a certain trait yourself.

But is there really any evidence that they are grouped based on traits? I played one game where they were all peaceniks (including myself). I mean, if there was any such grouping, wouldn't the designers try to include leaders with different traits?

 

[Lone Wolf]

That's how it was in Civ3, and that factor was documented if I recall correctly. It also had the neat effect of having neighboring civilizations who were from the same geographic region in the real world, because all the European civs shared one trait, all the Asian civs shared one, etc.

That statement is wrong.

 

[thekaje]

I've never drawn Shaka or Egypt. I get Saladin every game, though.

 

[MrCynical]

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.

Because some civs have multiple leaders while some only have one, there will inevitably be a bias to the leaders that are the only option for their civ. The simple way to think about it is to consider vanilla Civ 4 and a huge map. 18 civs, but 26 possible leaders. If I created a game with 18 civs then, it was a certainty I'd see Montezuma, because all civs were present, and he was the only Aztec leader. However it was not a certainty I'd see Stalin. Russia would definitely be present, but could get Catherine as a leader rather than Stalin. Hence, a bias towards some leaders.

This effect was diluted by the addition of more civs and leaders with BtS, and is less pronounced on maps with fewer civs, but it is still noticeable. You've correctly spotted that Shaka and Toku (only options for their civs) show up more often than Peter or De Gaulle (one of three possibilities in each case).

There was a thread a while back where someone did some testing of the probability of each leader, but I haven't managed to find it. I vaguely recall it worked out something around 40% chance for a single leader vs. a 20% chance for a triple on an 18 civ map, but don't quote me on that.

There are no other biasing effects - nothing to do with traits, picking warmongers and so on. There's just nothing in the code to do that. Civ 4 didn't use the culturally linked start system of Civ 3 (which was broken anyway in most of the patches and meant you always started with the native american civs next to you).

 

[Patak]

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

THE BURDEN OF PROOF!!!!!! YOU ARE A LAWYER, ARENT YOU?

 

[Sidder]

Always when I play with one of the American leaders, I end up having SittingBull next to me ...

 

[Öjevind Lång]

Always when I play with one of the American leaders, I end up having SittingBull next to me ...

Once I played as Sitting Bull and found that Washington was my nearest neighbour. I invaded his territory and conquered his capital, Washington, exterminating the American civilization before it could spread. I then considered renamning my newly acquired city of Washington and calling it Sitting Bull instead, but I decided not to.

 

[thekaje]

You know, that is a really great story.