AI Jumbled Rumble : another set of AI Survivor Alternate Histories, with a twist

AI_League_S1_WCP1_G1_Summary.jpg


So Vicky and Darius move on to the Finals to try and get a wildcard, and we bid farewell to the others.
No definitive answer yet about Ragnar, though, as this game was stacked against him. Let's wait and see how he performs next season, but his results so far don't bode well.


(Note: I'll try and post Game 2 results tomorrow, but I may not have the time for it. I'll then be away for a long week-end, so the next update won't be before midweek).
 

Attachments

  • 09 - Wildcard - Pool 1 - Game 1.zip
    3 MB · Views: 17
Strange that Roosevelt proved inert after the low peace weight eliminations, as I recall him overall playing erratically in my tests. I believe he declares at pleased as well, not that it makes any difference in the rankings. Congrats to Victoria and Darius :)
 
You're right about Roosevelt able to spot at Pleased. I had missed that.
Well, he never had me check ("Wait, what? Oh, he can plot at Please!").

That said, "Can Plot at Pleased", especially in the case of the high peaceweights, does not equate "Likely to declare war at Pleased". I had a game where Vicky really should have launched an attack to secure a win. She did: 120 turns later. Too late.
 
After quite a few rerolls, I ended up selecting this map for WildCard League, Pool 1, Game 2:

AI_League_S1_WCP1_G2_Map.jpg


I nearly discarded it, as it boasts a few features that hinted at a very poor balance, but I decided to give it a shot nevertheless. It ended up being a good exercise at map reading.

Position A:
This was my main concern: this position has so much land available that I feared it would be completely OP.
Spoiler :

AI_League_S1_WCP1_G2_Map_A.jpg



Position B:
Spoiler :

AI_League_S1_WCP1_G2_Map_B.jpg



Position C:
This was my second concern: this position is locked behind position A. I had so far rerolled maps featuring such a situation, as they tend to be pretty unfair. But here, the "prison" seemed very roomy, and I thought AIs starting from here would be able to compete with position A for land.
Spoiler :

AI_League_S1_WCP1_G2_Map_C.jpg



Position D:
Spoiler :

AI_League_S1_WCP1_G2_Map_D.jpg



Position E:
Spoiler :

AI_League_S1_WCP1_G2_Map_E.jpg



Position F:
And then there was this spot: double Gems, Copper. If anything, Position A should be facing some competition.
Spoiler :

AI_League_S1_WCP1_G2_Map_F.jpg



(Game details in the next post).
 
As the previous one, this game featured four high peaceweights versus two low peaceweights, and we saw how badly that went for the low peaceweights.
Even though the AIs on the high PW team seemed inferior to the likes of Darius and Victoria, I expected this game to unfold along similar lines.


AI_League_S1_WCP1_G2_Results.jpg


... and it did, in large part.
The one difference was Louis managing to secure 3 wins, so "Team Evil" did a tad better here.

AI_League_S1_WCP1_G2_Leaders.jpg


Louis indeed felt like he was the best leader in that field, often teching well and fielding a large military. It just wasn't enough to overcome odds that were stacked against him.
To make matters worse, while Wang Kon and Monty usually founded the first two religions, Louis would often found the Monotheism religion. That meant there was actually no "Team Evil": Louis and Monty would most of the times be enemies, owing to their religious divide.
Still, it was a bad setup for Louis (apart from the obvious being the target of multiple dogpiles): he often tries to go for Culture, when here, owing to the diplomatic situation, Domination was the way to go. So his Culture tech and build emphasis would often prevent him from snowballing efficiently.
Montezuma, on the other hand, had the right plan. He's just bad at implementing it. He just made enemies of everyone (attacking religious allies for peaceweight reasons, and his one peaceweight ally for religious reasons) and paid the price for it game after game.

This game featured both German leaders (who haven't exactly done well in AI Survivor). So which of them is the worse?

As it turned out, Bismarck performed significantly worse than Frederick.
I would have expected the opposite to be true: the more militaristic Bismarck, who can plot at Pleased, should have been better able exploit this field than the more peaceful and passive Frederick. But that didn't happen.
Bismarck often got into trouble he couldn't manage, while Frederick had more success with his please-forget-I'm-here-and-let-me-tech-in-peace approach.
But all in all, they did little to dispel the notion that they're among the weakest Civ4 leaders.

That left the Koean and Portuguese leaders as the dominant AIs in this game, a position they seemingly achieved more by default than through intrinsic qualities.

Wang Kon was carried by his economic strengths: Financial trait + Shrine income. Although the latter meant he would draw quite a lot of aggression, particularly from Monty. And I suppose that's where Protective came into play.
He did win an impressive and uncharacteristic Culture attempt, with a pretty early date (T265), but apart from that, it was usually grow through benefitting from being on the right side of a dogpile, and let your economy carry the day.
Joao also played to his strengths: a good expansion phase granting him a position of strength both militarily and economically, which he was often able to exploit, if not for a win, at least for making it to the end.
He equalled Wang Kon with 6 wins, but was only eliminated 4 times, which makes him the best performing AI in this group.

All in all, it did feel like a weak field, with only Louis and Joao hinting at the capibility to be competitive against stronger AIs.


AI_League_S1_WCP1_G2_Positions.jpg

As I suspected, the map wasn't very balanced, but... not exactly in the way I'd thought.

Although it seems pretty obvious, I had missed Position B being an absolute death spot: AIs starting there would be hopelessly boxed-in. It had Iron, but no Copper: so no early break out attempt as a possibility, and any delay in researching Iron Working could prove fatal when sitting next to an AI with Copper at its capital (F).
So that made Position F, correctly identified as a strong spot, even stronger. AIs starting there would not only have a good early economy thanks to the Gems, they were also provided with an easy expansion prospect.
Almost as good (same number of wins, a bit more eliminations) was the other neighbour of position B, Position D. The capital site wasn't as good, but its Gold tile still made it pretty strong. What made this position so strong (in addition to its super weak neighbour) was that it allowed to compete for the western lands. What I had identified as Position A's land was also Position D's. Central positions tend to be all or nothing: they're either death spots or great spots. Here, it turned out to be the latter, and I had completely misread that.
The one weakness for that position was a lack of Copper, which could spell trouble when IW was delayed.

With positions D & F accouting for 16 wins, that meant Position A wasn't as strong as anticipated. It actually only saw 3 wins.
Now, the AI starting there was almost always an early and mid-game leader, so it was a strong position. But not enough. Part of the reason why is I believe that its huge backline became barb territory. That led to pillaging, and having to divert ressources to deal with them. It also meant that with those barb cities to conquer, the AI there would miss out on the early dogpiles. Or, if it did join in, it would then see another AI conquer those barb cities and deprive it of that backline.
Also, conflict with the tenant of Position C was pretty much unavoidable, and it would often come in the form of an inopportune backtab.

Albeit roomy, Position C proved a secure jail indeed.
Rougly half the time, the AI there was eliminated as a consequence of failing to escape through conquering the stronger AI in position A.
The remaining scenarios were a split between staying put and making it to the end as an irrelevant civ, or winning a long struggle, but too late to matter.
Position E seemed average, with decent land and expansion prospects, and that's what it proved to be, with only one win, but a 50% survival rate.
It suffered from two main issues. The first one was that it bordered four other civs, and thus was very conflict-prone. The second issue was that one of those recurring conflict was against the tenant of position F... usually a stronger civ.


AI_League_S1_WCP1_G2_Summary.jpg


So Joao and Wang Kon join Victoria and Darius to a finals which seem promised to a high peaceweight winner.
And from what I've seen so far, I don't see either of them having a chance against the Game 1 winners. It would seem that Victoria (or possibly Darius) has a lock for the wildcard.
But let's see first who the final two contestants will be.
 

Attachments

  • 10 - Wildcard - Pool 1 - Game 2.zip
    3.1 MB · Views: 15
Last edited:
Let's see if Gilgamesh dominates the next game? Other 3 evils Mehmed, De Gaulle and Mao Zedong really feels like they are identical, no one stands out.
 
As I've already mentionned, the map for Wildcard League Pool 1, Game 3 had me forgo my "no edits" rule. And once you start down that path... it's hard to stop! :D
So in the end, I made quite a few minor edits (moving ressources around, adding a bit of arctic food, altering rivers, ...).
Beyond the few major edits noted below, the other major alteration I did was ensuring every position had an easy access to metal (which most definitely wasn't the case for the original version of the map).

AI_League_S1_WCP1_G3_Map.jpg


Position A:
This starting position was horribly boxed-in in the unedited version of the map, with position C right on its doorstep to the east. I had thought there was room to expand to the south-east, but Mao closed that door immediately with his first settler, in a game I ending up discarding.
So at first, I removed some ice tiles from the antarctic regions, and added tiles to this position. That wasn't enough. In the end, I moved position C all the way to the southeastern coast of the continent. But I left the added tiles in.
Spoiler :

AI_League_S1_WCP1_G3_Map_A.jpg



Position B:
This position is what saved the horribly boxed-in Mehmed in the game I discarded: he launched an attack past Mao's territory and got his snowball started by an early conquest of Augustus... who completely lacked any source of metal!
So I swapped a nearby Coal resource for the closest Iron, but that wasn't enough: it was only 4 tiles from the capital (once tile south of where the Coal currently is), but Augustus kept missing it with his settlers!
I gave up and replaced the Gems at his capital (there were already two other happiness resources after all) with Iron.
In retrospect... I should have made that Copper, and left the Gems in. But more on that later.
Spoiler :

AI_League_S1_WCP1_G3_Map_B.jpg



Position C:
So this was the position which caused all sorts of woes, as it was ultra-dominant, basically having the run of the whole southern part of the continent once Position A was neutered.
I initially tried moving it half-way to the southeast from its initial spot just east of Position A, but that essentially changed nothing.
So I ended up moving it all the way. We'll see how that went.
Spoiler :

AI_League_S1_WCP1_G3_Map_C.jpg



Position D:
Alright, you'd think I'd learnt my lesson with the previous map: peninsulas locked behind another civ are bad.
But hey, it also borders Position B in spite of that tiny gulf, and it has so much room that it hardly qualifies as a peninsula. Right? Right?
Spoiler :

AI_League_S1_WCP1_G3_Map_D.jpg



Position E:
I was a bit worried about that position too, as it also seemed a good candidate for some early boxing in, but initial testing pointed at an okay expansion, so I left it largely untouched iirc, apart from moving the Fish to a less awkward spot.
Spoiler :

AI_League_S1_WCP1_G3_Map_E.jpg



Position F:
I left this one largely untouched, apart from improving its metal situation.
It benefited from my improvement of the tundra regions, but so did its neighbours.
Spoiler :

AI_League_S1_WCP1_G3_Map_F.jpg



(Game details in the next post)
 
Last edited:
Let's see if Gilgamesh dominates the next game? Other 3 evils Mehmed, De Gaulle and Mao Zedong really feels like they are identical, no one stands out.

Well, let's see about it...

AI_League_S1_WCP1_G3_Results.jpg


Okay... "dominates" is indeed one way of putting it. :lol:
Gilgamesh crushed the competition, setting a new record with 10 wins!

AI_League_S1_WCP1_G3_Leaders.jpg


Augustus and Washington were doomed, for the usual peaceweight reasons.
These are two leaders with pretty average stats which feel like middle-of-the-pack fare, only dropping to below average because of their peaceweight.
Out of the two, Augustus seems to enjoy more success, and this game was no exception: he was able to take advantage of the conflicts between the low peaceweights to pull off two wins, while the best Washington achieved was not dying. Sometimes.
Praetorians obviously spring to mind as an explanation, but I'm not so sure. Need even more games, I guess.

Mao, De Gaulle, and Mehmed performed similarly, with Mao disappointing once again and being worse than the other two: he won one fewer game, and got eliminated more often.
So Mao confirmed as a fraud? Maybe, but let's wait a bit: his first game was utterly dominated by strong high peaceweights, and this one by Gilgamesh. So we can say for sure that he doesn't belong in the same tier as Gilgamesh, but it seems premature to try and assess how far below he stands.
Mehmed ends up performing slightly better than De Gaulle (same number of wins, just survived two more games), but somehow, he felt a lot stronger. Winning by himself, while De Gaulle lucked out on dogpiles.
But was that luck? After all, De Gaulle is, I believe, the AI most likely to join a dogpile. So that's actually a strength?
Let's put it another way: after watching those 20 games, Mehmed moving on, not De Gaulle, "feels right".

But anyway, it's a bit hard to discuss the performance from those leaders, because there was only one story to that game, and it was Gilgamesh's dominance.
He confirmed that he is one of the more successful warmongers out there.
He expands well, Creative always helping the AI make up for dubious city placements (resources in the 2nd ring). Vultures certainly help with the early wars. And while he's no economical powerhouse, he tends to do better in that regard than most other warmongers, often managing to actually turn his conquests into a tech advantage (when other warmongers just turn them into moaar units).


AI_League_S1_WCP1_G3_Positions.jpg

If this were an unedited map, I would say that balance was fine overall.
But as it was modified... could have done a better job, I guess.

The big one: Position B was an absolute death trap. It wasn't a central spot, it was the central spot on this map. As such, I suspect it would have been impossible to properly balance: if it were improved, it would stay bad, until the improving reached a tipping point, and that spot would become godly and dominate the map.
But here... let's say that tipping point was a long way off.
Position C, in spite of its exile, remained the best position. It was even more striking in-game, as its tenant would almost invariably have a very strong early and mid-game.
Position A, being so awful in the original map, had received the most "care", and it kinda shows here...
Position D... was a peninsula-type position. As such, it has the highest survival rate, but the fewer wins.
Positions E and F were overall subpar (highest elimination rate after Position B), but could be made to work, with an average number of wins (3-4 is the expected number for a balanced map, and that's what we got there).


AI_League_S1_WCP1_G3_Summary.jpg


So a very impressive Gilgamesh (with Mehmed tagging along) joins the finals for this pool.
He'd be the favourite in my opinion... except that's going to be a 4 high peaceweights vs 2 low peaceweights game, and we've kinda seen how those turn out.
Oh well.

But in the meantime, let's see how things went in the other group (pool).
 

Attachments

  • 11 - Wildcard - Pool 1 - Game 3.zip
    3.1 MB · Views: 12
Last edited:
I don't have much to add to the map setup and game analysis, but I really appreciate the effort you put in, @Thrasybulos! It's always interesting to see AI v AI action and how it turns out.
 
Finally good to see you back, I really enjoy reading your posts and also good that I was right with Gilgamesh. :lol:
I didn't quite catch which of the Sulla's map you have edited, but I assume position B could win quite a few too if this wasn't aggressive ai.

As for the next game, pool 2 game 1 between Zara Yaqob,Elizabeth and Pericles, I would guess Zara based on score/survival and Elizabeth in terms of victory count to come as top 2 for such a safe group for her. Other three leaders are just background characters.
 
@NickGabben Thanks. :)

Finally good to see you back
Oh, I'm not going anywhere. Just a bad combination of being tired and busy.
The "tired" part should improve once the current heatwave ends.
The "busy" part should... actually get worse. :( September is a bad month at work.
I won't be able to sustain the one update per day pace I'd started with, but I should manage better than once per week. ;)

I didn't quite catch which of the Sulla's map you have edited
That one wasn't a Sullla map. It was one I had rolled. But after it appeared it was completely unbalanced, I decided to keep it and edit it rather to roll a new one.
Sullla's edited maps start with the next game.
 
So, as I've already mentionned, for Pool 2 games I ended up re-using the 6-player Opening Round maps, but after editing them in an attempt at balancing them.
Pool 2 Game 1 will thus be played on a modified version of the map for Opening Round Game 1.

AI_League_S1_WCP2_G1_Map.jpg


Position A:
This was Cyrus's position in the original game, and it ended up being by far the strongest position on the map.
Which is why... I did very little about it.
I fixed an "illegal" resource cluster Sulla had left after moving a starting position, which weakens it very slightly. But it also benefits a little from my improving of the tundra and ice area.
I didn't want to actively nerf it, because I thought it would mechanically be made weaker when I improved the surrounding positions. It is a central position, and it would be easy to mess up and over-correct.
Spoiler :

AI_League_S1_WCP2_G1_Map_A.jpg



Position B:
This had been identified as one of the weaker positions on the map: no Copper, little room to expand, jungle covering a lot of the expansion room.
I removed a few tiles from the antarctic regions and added them here. I added Gems at the Capital and removed some of the jungle (in particular on the Rice tile which is usually where the first settler is sent).
I did not add Copper as I didn't want to just give Copper to every civ. There's Iron at the capital, the extra Gems should enable a timely IW research. If the AI doesn't get it in time... play better next time!
Spoiler :

AI_League_S1_WCP2_G1_Map_B.jpg



Position C:
That position was largely okay. It just suffered from being next to the strongest position.
I did improve it a bit, mainly the southern, icy part. Also, there was always a barb city popping to the east on the initial map: I moved some of the resources there closer to the capital so they wouldn't be claimed by the barbs every time.
Spoiler :

AI_League_S1_WCP2_G1_Map_C.jpg



Position D:
This was the second strongest position on the map. Apart from having Gold at its capital, I believe the main reason why was that it essential had a lock on the northern part of the map, thus squeezing both its neighbours for room, and getting stronger through having two weak neighbours in a kind of feedback loop.
So what I did was move it on the coast, to the north-west of its original position: the North should thus be a more contested area, which should balance out the 3 northern starts.
As a compensation, I made the tundra area more exploitable.
Spoiler :

AI_League_S1_WCP2_G1_Map_D.jpg



Position E:
This was Qin's position in the initial game, and the worst start on the map: boxed-in with room for 4 cities, 5 in a good game, jungle-choked... yuck.
I cleared the jungle (moving the jungle towards the center of the map), moved/added resources. And fiddled with the immediate southeast until the AI would stop sending its initial settler there (seriously, what's up with that?).
The rest of the balancing would hopefully be achieved through Position D's moving.
Spoiler :

AI_League_S1_WCP2_G1_Map_E.jpg



Position F:
This one was really bad, too. It offered basically no shot at victory, achieving a decent survival rate mainly by virtue of being out-of-the-way.
So what I did was to put it back into the game by moving it inland, to the east.
I also improved the surrounding area.
In retrospect... I moved it too far inland. I should have stopped midway. Oh well.
Spoiler :

AI_League_S1_WCP2_G1_Map_F.jpg



(Game details in the next post)
 
As for the next game, pool 2 game 1 between Zara Yaqob,Elizabeth and Pericles, I would guess Zara based on score/survival and Elizabeth in terms of victory count to come as top 2 for such a safe group for her. Other three leaders are just background characters.
You sure about that? Let's see...

AI_League_S1_WCP2_G1_Results.jpg


Well... that was spot-on! :thumbsup:

AI_League_S1_WCP2_G1_Leaders.jpg


Boudica kept on with the ongoing trend of the Celtic leaders performing extremely poorly. She did manage to get eliminated one fewer time than Cathy, but that's about it.
Now, this game was stacked against her, so it would be premature to dismiss her off at this point. We'll see how she fares in the next seasons.
Catherine was in the same situation, and performed accordingly poorly. But her more aggressive nature (she can plot at pleased, so even if sharing religion she would have no dearth of targets) allowed her to overcome those odds on two occasions... and led to her elimination most of the rest of the times.

Sitting Bull benefitted from being on the winning side... and that's about it. He's widely considered as one of the worst leaders, and let's say this game did nothing to dispel that notion, in spite of his two wins.
Pericles was, again, a disappointment. He was only eliminated 6 times and got two wins towards the end, but he never really felt like he was in the run. His Cultural attempts were for the most part fated to fail, and it took rare circumstances for him to achieve a dominant position or a tech lead. We'll see in the next seasons, but so far his performances strongly hint at an overrated leader, which rather belongs near the bottom of the pack.

Elisabeth got nine wins, which is an excellent performance, hard to argue with... and yet I will.
She had two very good games: the first game, where she turned her tech lead into a strong military edge and faced a 1v2 situation head-on, conquering both her aggressors and eliminating them both within a few turns of one another. An impressive Domination win. And then, the very last game, where she turned the slider after having secured her position. She was in control the whole time.
Apart from those two games (and possible game 3)... her victories felt "miraculous"? It seemed that at any point another AI would call her bluff and her flimsiness would be revealed. And it did happen: out of the high peaceweight leaders, only Sitting Bull died more often.
One of Elisabeth's issues is that she's just bad at expanding. She would do well in positions with a lot of backline, but get badly squeezed when expansion was a race.
In her game, Victoria often achieved dominant positions and you'd wonder how she was going to scuttle herself. Here, Elisabeth achieved the same number of wins, but you'd keep wondering how the other AIs could have let her get away with it.

Zara Yacob, on the other hand, felt a lot stronger even though he only achieved 5 wins. He expands well, and usually ends up with a strong economy and decent military.
So how come he wasn't more successful?
He can't plot at pleased, and gives a massive +8/+9 to relations for shared religion, with only -3 for a different religion. And his favourite civic, Theocracy, will often be run by civs at war, making him that much more unlikely to join an ongoing war. So he'd stay pleased with the other high peaceweights even with a different religion, and also be pleased or even friendly with low peaceweights who shared his religion.
Zara often seems too passive... but that's because his hands are tied because of those diplomacy modifiers!


AI_League_S1_WCP2_G1_Positions.jpg

So, how did my balancing go?
Not exactly stellar. :(

I was correct in being wary about nerfing Position A: it did become the worst spot on the map, simply by virtue of having its neighbours improved.
Moving Position D out was the right call I believe... but I clearly over-improved the surrounding land as it ends up as the best position for winning, and the second-best for survival.
Position C suffered: it ended up slightly worse off than in the original map version. Guess I should have really improved it, after all.
Positions B and E did get better, with an expected number of wins. But elimination rate remained high.
Finally, Position F was over-improved. As I've already said, I believe I shouldn't have moved it that much inland. That would have allowed a fairer competition for positions A & B.

AI_League_S1_WCP2_G1_Summary.jpg


Zara and Elisabeth fight on, we bid goodbye to the others.
The next two games in Pool 2 are balanced 3 high PW vs 3 low PW affairs, which actually tend to go the low PW way, so the finals might be a tough field for Zara and Elisabeth.
We'll see...
 

Attachments

  • 12 - Wildcard - Pool 2 - Game 1.zip
    3 MB · Views: 15
Last edited:
Just a quick update: as feared, September is off to a busy start.
In addition to that, I've discovered that a not insignificant amount of money has been siphoned off my bank account, and dealing with that issue is taking up a lot of my free time and energy.

So I'll try and resume posting ASAP, but "soon" and "possible" may be at odds here.
 
Top Bottom