Sullla's AI Survivor Season Eight - Championship Alternate Histories Thread

[Eauxps I. Fourgott]

Sullla has posted the latest set of alternate histories for the ongoing season, this time completed by yours truly for the Championship! The real game saw Churchill do nothing and Augustus conquer Qin far too slowly to be competitive, leading to a tense three-way space race that played out over a backdrop of Gandhi going nuclear and ended with Elizabeth winning by a mere handful of turns. But was this "supposed" to happen?

Writeup: https://www.sullla.com/Civ4/civ4survivor8-13A.html

Short version: This was a highly unusual setup, not least because it was extremely peaceful. The correct strategy turned out to be to tech up to a very early spaceship, and that resulted in only the three western economic leaders having a realistic shot at the trophy. Mansa also turned out to be far more likely to win than the others for reasons that I personally couldn't identify. That meant that Liz's victory was an unusual result, although the real game otherwise played out in a pretty typical fashion (outside of the nuclear fireworks, which were a special treat reserved for the live game alone).

Discuss below!

 

[Eauxps I. Fourgott]

I have plenty to say about this set... but I already said it in the linked writeup! Only limited extra comments here.

First, while the answers can all be found in the writeup, here are the short versions of answers to questions folks posted when I was beginning the set:

Whether Qin being comparatively so weak was a bug or a feature: when Sullla said "As expected, Qin is FTD", I thought "Hmm, no: him being FTD was indeed largely to be expected, but it didn't happen as expected at all! Everybody thought he's suffer a major dogpile, not be solo-conquered early!"

It was a feature. Qin was terribly weak even outside of war, and an easy target for whoever wanted to go after him.

Whether everyone but Gandhi ignoring religion for so long was a fluke or not.

It was an exaggerated version of the usual dynamic. Often it was 20-30 turns until the second religion was chased, not 30 (though once it didn't happen until the 60s!)

Whether 3 good techers + two Industrious leaders not among them + super rich map does mean that Space is much more frequent than Culture (I expected that, but was that a good read or just luck?).

It did indeed. Probably 75% of all games never had the slider flipped on.

How often does Ghandi go for space instead of culture?

Almost always. He made two or three cultural attempts in the entire set.

I'll also note that I've already started running the AHs for Game 8, the last one from this season that I've "claimed", so my writeup for that will be coming in just a couple more weeks.

I'll have a couple of other notes but don't have time to post them right now.

 

[Thrasybulos]

I haven't run games of my own on the Championship map yet, but your comments in the write-up did nothing to dispel the notion that while a fine-looking map, it's not well-suited as a Championship map.

As has been said earlier:

• Too big
• Too rich
• Too safe early
• Too impractical for warfare
• Too prone to "lock-ups"? (P1 & P2 don't have open borders, P1 dows P3, P2 dows P4, P1 conquers P3's central cities => P2 can no longer access P4's territory)

It feels different enough from the standard Pangea maps used in the rest of the tournament, that it's a bit as if the Championship game were played on an Archipelago map.

 

[Fippy]

I hope Sulla can be convinced about that, imagine Mansa reaches the CS again and everybody just picks him on the same map.

 

[Eauxps I. Fourgott]

The other notes I had:

It was theorized in the Discord that the big tipping point for Mansa's performance was shrine income. I didn't take specific note of this aspect during my replays, but he did pretty much always found at least one religion, and there were a lot of cities available to spread it to. Him emphasizing that more than Liz may have helped give him the edge over her, while Financial trait gave him the edge over the similarly shrine-y Gandhi. Food for thought, at least.

I'll also admit: this set got a bit tedious to run! The later turns take a lot longer to churn through on bigger maps like this, which combined with the largely static dynamics to make for slow lategames. I'm glad to be moving on to a more normal map!

I haven't run games of my own on the Championship map yet, but your comments in the write-up did nothing to dispel the notion that while a fine-looking map, it's not well-suited as a Championship map...

I don't know about the other aspects, but my opinion is still that this map is too big. Others like Sullla still aren't convinced that the map is a problem as opposed to these odd dynamics being a result of the field, and there's a general interest in more data about how it plays out with different fields of leaders.

I know you end up putting in a lot of replays - would you want to provide some data on this? I feel there's unlikely to be any changes unless somebody does it. I personally have no interest in spending more time replaying this map.

 

[0urBall]

Has anyone tried to run any of the other games in the season on that map? To see whether a field of lower PW/military Civs would change the number of wars?

 

[Thrasybulos]

I know you end up putting in a lot of replays - would you want to provide some data on this? I feel there's unlikely to be any changes unless somebody does it. I personally have no interest in spending more time replaying this map.

I won't be using that map for the League, but when I do my alternate AH for S8, that'll be 20 games I expect with a different set of leaders.

But Keler has already posted game results.

 

[danjuno]

My suggestion would've been a hexagonal map to provide some room but not too much.

 

[Thrasybulos]

Alright, you wanted some more games on this map, here are some (see attached file).

Differences with AI Survivor rules:

• No UN
• No barbarians
• Xenophobia: all AIs get an additional -1 to their relations with the other AIs in round 1 games, -2 in round 2, -3 in round 3, etc.

Out of 38 games, 3 ended with a Cultural victory, 2 with a Domination win, and 33 with Space.

 

[Keler]

You know map is unsuitable when a neighbour civ turns on culture slide %100 starting from physics not attacked for the rest of the game yet still can't beat space.. not to mention their neighbours not even culturally crushed to have their prod/tech speed ruined...

Fun, the avg end game date here is 296 just like in my 24 runs.

I don't understand what the tournament structure of this thunderdome is but that's another topic. It quickly selected some of the better AIs in just 38 runs.

 

[Thrasybulos]

I don't understand what the tournament structure of this thunderdome is but that's another topic. It quickly selected some of the better AIs in just 38 runs.

Spoiler :

(Spoilered because it is indeed off-topic)

I discovered by accident the other day while having a look at the wiki that something I had wanted to try for a looong time was actually possible: having the AIs compete in an Always War setting.
So I designed a tournament format for it, and started testing it.
Obviously, I was worried that the AIs wouldn't be able to cope: we know how poorly they deal with a two-front war, it would be much worse there.
Well, I was right to worry, but for the wrong reason.
Every single game unfolded in the same way: one AI would start getting an edge, and it would snowball and kill everyone. No competition, not even kill steals.
So, I dropped the AW idea: it doesn't work.

But I decided to have a go at testing the tournament format, just in case.

"Credits":

• All AIs start with the same number of "credits".
  I went for 2 so that each AI would play in at least two games, without dragging things for too long.
  (The name "credit" doesn't really fit now: I came up with it for a previous version of the rules.)
• There are two ways for an AI to "steal" credits from another AI:
  • by killing it
  • by winning the game (the winner gets credits from the survivors)
• The number of credits "stolen" equals the round number, but cannot exceed the remaining credits of its victim at the start of the game
  Say this is a round 3 game, Shaka starts the game with 2 credits, Cyrus with 7.
  Shaka kills Cyrus and survives, but Pacal wins the game.
  Shaka gets 3 credits from Cyrus (who leaves the game with 4 credits remaining) but only loses 2 to Pacal.
  • This means the total number of "credits" in play is constant (here 52 x 2 = 104).
• If an AI finishes a game with 0 credits, it is eliminated from the tournament.
• If an AI finishes a game with more than 50% of the total credits (here: 53 or more), it wins the tournament.

"Rounds":

• At the start of the tournament, the list of AIs is randomized and put in the "active queue".
• The top 6 AIs of the active queue are the participants of the next game.
• The winner of a game qualifies for the next round: it gets placed in the "Next Round" column, and sits out the rest of the current round.
• The other participants of the game, if they're not eliminated from the tournament (credits > 0), get placed in the "on hold" queue for the current round.
• When the "active queue" is empty, the "on hold" queue is randomized and placed in the active queue.
• And so on until there are less than 6 AIs left in the active + on hold queues for the current round.
• Those remaining AIs get then added to the "next round" AIs, the list is randomized, and becomes the new "active queue" for the new round.
• The whole process goes on until one AI wins by majority, or there are less than 6 AIs left: the winner is then the AI with the most credits.

It probably sounds very complicated when explained that way, but really, it's not.

 

[Kaitzilla]

Sullla has posted the latest set of alternate histories for the ongoing season, this time completed by yours truly for the Championship! The real game saw Churchill do nothing and Augustus conquer Qin far too slowly to be competitive, leading to a tense three-way space race that played out over a backdrop of Gandhi going nuclear and ended with Elizabeth winning by a mere handful of turns. But was this "supposed" to happen?

Writeup: https://www.sullla.com/Civ4/civ4survivor8-13A.html

View attachment 706585

Short version: This was a highly unusual setup, not least because it was extremely peaceful. The correct strategy turned out to be to tech up to a very early spaceship, and that resulted in only the three western economic leaders having a realistic shot at the trophy. Mansa also turned out to be far more likely to win than the others for reasons that I personally couldn't identify. That meant that Liz's victory was an unusual result, although the real game otherwise played out in a pretty typical fashion (outside of the nuclear fireworks, which were a special treat reserved for the live game alone).

Discuss below!

Jeez, Mansa smoked the field in Alternate Histories.
Thanks Eauxps!

 

[Keler]

It probably sounds very complicated when explained that way, but really, it's not.

I get it now I think

Did you also keep track of who started where?

 

[Thrasybulos]

Did you also keep track of who started where?

The players for each game are listed in the "Teams" order in the wb file.
So the information is there. Just add a column with 1,2,3,4,5,6,1,2,3,4,5,6,... if you want to exploit it.

 

[Keler]

The players for each game are listed in the "Teams" order in the wb file.
So the information is there. Just add a column with 1,2,3,4,5,6,1,2,3,4,5,6,... if you want to exploit it.

I just did and of 38 games I found

Capital 1 (Mansa Musa) with 11 victories
Capital 2 (Elizabeth) with 5 victories
Capital 3 (Augustus Caesar) with 9 victories
Capital 4 (Gandhi) with 2 victories
Capital 5 (Churchill) with 10 victories
Capital 6 (Qin Shi Huang) with 1 victory

Parallel and actually even more extreme than my results

edit: my results from ai survivor switchhero of 24 games

Spoiler my results :

Capital 1 (Mansa Musa) with 4 victories
Capital 2 (Elizabeth) with 4 victories
Capital 3 (Augustus Caesar) with 6 victories
Capital 4 (Gandhi) with 1 victories
Capital 5 (Churchill) with 6 victories
Capital 6 (Qin Shi Huang) with 3 victory

everything else aside, the map is not even fair enough?

 

[Thrasybulos]

4 & 6 seem oddly weaker indeed, with 2 a bit suspicious?
1, 3, & 5 have registered the same number of victories (15, 15, 16).