AI Survivor - Alternate Histories AI Ranking

[Thrasybulos]

Quick answers, since I have a train to catch...

As it turns out, Mansa is a MUCH better military leader than we realize.

Oh yes, I know. Always cracks me up when Sullla describes him as a "peaceful leader". Now, calling him a warmonger would be a stretch, but he can certainly lay down some whack!
Another misperceived leader like that is Asoka: he's not "Gandhi-light".

Quite surprised at Justinian, but I guess Ragnar is a much tougher opponent especially with a Fishing start.

No, Ragnar didn't amount to much in these games, that start was bad. He had a few good games where he indeed managed to beat Justinian, but most of the times the best he could hope was kill Viccy before being smashed by Shaka and/or Justinian.
Justinian's problem was that Shaka would simply snowball harder than him.

Curious how Shaka and Cyrus were able to win games in this ultra economic setup. Also curious if this altered your opinion on the championship map.

Shaka's win was by a coupla turns: when the first batch of aiplay turns ended, I saw that Hannibal had launched and Shaka was at something like 58% land area. I thought "damn, so close, but still no cigar". I was pleasantly surprised that he managed to get to the Domination threshold in the few turns remaining.
Cyrus's wins were enabled by Shaka, who slowed down HC and Hannibal enough for Cyrus to catch up.
My opinion on the map hasn't changed.
It's damn fine-looking, but too big, too rich, too "clever" (the AI has no strategic analysis capability, it doesn't know it should compete for the middle ; many gamees feature one or two AIs cut off from the middle after the expansion phase, and thus out of the game already).

 

[Keler]

Mansa is not doing well... If you call that performance well for an equal pw game (no dogpile) with that god tier capital room for 5 more cities than any other leader on the map.. He hardly performed better than Asoka.
Mansa is total dissapointment here.

 

[Genghis the Barbarian]

Mansa is not doing well... If you call that performance well for an equal pw game (no dogpile) with that god tier capital room for 5 more cities than any other leader on the map.. He hardly performed better than Asoka.
Mansa is total dissapointment here.

Ehh, Mansa is pretty diplomatically isolated in this setup, both Asoka and Charlie are likely to have differing religions. If one of them were to get large Mansa is not safe from either, and that seemed to have happened a lot in these replays

 

[Thrasybulos]

Season 4, Opening Round

Back to the beginning, where the AH started.
It's also the season I'm probably the most familiar with, since I have my own set of AH for it, plus season 1 of Jumbled Rumble was based on AI Survivor Season 4.
So there shouldn't be any real surprise to expect...

Spoiler Game 1 :

Official AH

Pretty consistent with the previous results, with the exception of Elizabeth winning only once here.

And yet, in view of the results from Jumbled Rumble, this is actually rather surprising.
Cyrus's start was identified as the best on the map, by a mile. The next best spots were Isabella's and Elizabeth's. Catherine's was very poor, Roosevelt's offered no shot at victory but good survival odds, and Qin's was an absolute death trap.

So Cyrus winning only 8 games is actually a rather poor showing. Cathy, on the other hand, winning 6 games from that start is amazing.
Isabella could also be considered to have underperformed here, while Lizzy clearly did.
But Roosevelt and, especially, Qin actually winning a game: just incredible!

No one expects the Spanish Inquisition.
The Spanish spies, on the other hand...

Spoiler Game 2 :

Official AH

These results are consistent with the narrative from the official AH: Willem and Louis are the strongest on this map, Gandhi as usual manages to slip through a few victories.
This set makes that narrative stronger as no else was able to win here, contrary to the official AH where Mehmed, WK, and Pacal totalled 4 wins.

Spoiler Game 3 :

Official AH

This map was ridiculously unbalanced.
JC's starting position was ludicrous, and De Gaulle's was the other strong start on the map (but miles behind).
No wonder those two should come on top.

The real surprise here is Suleiman who scored a mere 4 points in the official AH, whose start was identified as a death trap, and who manages to pull off three victories!

Spoiler Game 4 :

Official AH

And the Peter conspiracy continues!
The official AH had him on top, with 11 wins to Justinian's 7 wins. Those results are reverted here.

Peter's start had everyting: next to the death spot civ (Sury), next to the dead man walking (Asoka), next to a rich starting position vacated by Sullla when editing the map.
And yet... Justinian beats him.

Spoiler Game 5 :

Official AH

Probably one of the craziest setups in AI Survivor history: both Mongols, Alex, and Toku!

Kublai on top, Toku second: consistent with the official AH result, even though KK did way better here.
The live game setup led a very unusual game dynamics for this map with those leaders: KK's central spot was actually a doomed start, GK and Alex should have been the best performers in this group.
But this particular configuration, where the two victims (Zara and Bismarck) were KK's nearest neighbours, changed everything.

Spoiler Game 6 :

Official AH

The official AH had Mao as a weak favourite with 7 wins, and basically everyone but SB tied after that.
Which, in view of my results from Jumbled Rumble, is pretty unexpected: the starting positions heavily favour the low peaceweights here.
In particular, Mao has the best start on the map by a good margin, while Hatty has a death spot.

So Mao doing slightly better here (9 wins) seems "correct".
In the same way, Hatty getting slaughtered game after game (a good portion of Mao's 29 kills are Hatty murders) certainly feels more expected than her somehow pulling 2 wins!
Darius getting 3 wins from the 2nd strongest position on the map feels about right.

But Mansa... 6 wins from that spot??
Also, look at the streak here: Mansa won 5 out of the first 9 games, with Mao getting only one win. And then, Mao wins 8 out of 9 games from 12 to 20!

My previous set was in-between, but also had Mansa winning a lot:

I guess that makes it an impressive showing by Mansa.

Almost as impressive as SB's survival strategy (from Game 1):

That's the whole Native American civilization, right there.

Spoiler Game 7 :

Official AH

HC in the best spot on the map = roflolstomp.
This is actually his worst showing: 80% win rate in the official AH, 85% in my previous set, only 70% here.

The only interesting aspect here: the official AH had Washington as the second best performer which led Myth (who ran those) to build a secondary scenario for this setup.
My previous AH set had Napoleon as the second best performer.
And this one has Augustus.
Whoops.
I guess that when a leader is ultra dominant, there is no secondary scenario. Only random noise.

Spoiler Game 8 :

Official AH

Last game... and the first genuine surprise?

The official AH had Stalin as the favourite for the win (8 wins), with Ramesses and Hannibal behind (5 and 4 wins respectively) but Hannibal almost equalling Stalin in score thanks to a lot of second places.

My previous set had the same 3, but added Pericles to the lot, with Hannibal that time barely ahead in number of wins:

This time, the same 4 leaders share the wins, but Ramesses is on top!

I'd say that this is probably the correct story: on this map, as far as winning is concerned, Stalin, Hannibal, Pericles, and Ramesses are roughly tied. Who comes on top in a given AH set seems a mere sample size issue.

 

[Eauxps I. Fourgott]

[IMG alt="S4_Opening_1_Spanish_Spies.jpg"]https://forums.civfanatics.com/attachments/s4_opening_1_spanish_spies-jpg.709995/[/IMG]
No one expects the Spanish Inquisition.
The Spanish spies, on the other hand...

Spoiler Game 6 :

Official AH

[IMG alt="S4_Opening_6.png"]https://forums.civfanatics.com/attachments/s4_opening_6-png.709998/[/IMG]

The official AH had Mao as a weak favourite with 7 wins, and basically everyone but SB tied after that.
Which, in view of my results from Jumbled Rumble, is pretty unexpected: the starting positions heavily favour the low peaceweights here.
In particular, Mao has the best start on the map by a good margin, while Hatty has a death spot.

So Mao doing slightly better here (9 wins) seems "correct".
In the same way, Hatty getting slaughtered game after game (a good portion of Mao's 29 kills are Hatty murders) certainly feels more expected than her somehow pulling 2 wins!
Darius getting 3 wins from the 2nd strongest position on the map feels about right.

But Mansa... 6 wins from that spot??
Also, look at the streak here: Mansa won 5 out of the first 9 games, with Mao getting only one win. And then, Mao wins 8 out of 9 games from 12 to 20!

My previous set was in-between, but also had Mansa winning a lot:
[IMG alt="S4_Opening_6_alt.png"]https://forums.civfanatics.com/attachments/s4_opening_6_alt-png.710002/[/IMG]

I guess that makes it an impressive showing by Mansa.

Almost as impressive as SB's survival strategy (from Game 1):
[IMG alt="S4_Opening_6_SB_Survival.jpg"]https://forums.civfanatics.com/attachments/s4_opening_6_sb_survival-jpg.710003/[/IMG]
That's the whole Native American civilization, right there.

Mansa's quite the guy, isn't he?

 

[Thrasybulos]

Season 4, Final Rounds

Spoiler Wildcard Game :

Darius was doomed, both by his peaceweight and his starting position (central, no copper). That he actually managed to eke out two wins could be construed as him overperforming.
Pericles had a much safer start, but low on commerce. And well, his peaceweight situation wasn't any better.
Stalin had the best start on this map, so no surprise he did well, especially with a Persian snack next door.
Boudica and especially Mehmed were disappointing... although the latter's results might be explained by Gilgamesh's exceptional performance: Mehmed was often in the way...

Anyway, yet another result confirming Gilgamesh as one of the strongest leaders: it wasn't an all-star game, far from it, but he was able to dominate from an average starting position at best.

Spoiler Playoff 1 :

HC confirms my AH for the live game: although the game composition is somewhat different, he once again prevails, from a poor start.
Hannibal somewhat disappoints: although he has the best survival rate, he was never able to turn that into a win. Pleased-lock I suppose, plus the fact he kinda needed HC to be gone to stand a chance, and Toku was usually there first.
Louis was very lucky that both his neighbours can't plot at Pleased, as he got the metal-deprived start.
Justinian's start was bad, and he got Toku'd pretty often.
Cyrus was another disappointment, but I suppose "can't plot at Pleased" was a real handicap in this field: note that the three leaders who did well are the three who can plot at Pleased.

Spoiler Playoff 2 :

When there are high peaceweights around to draw attention away from his gambit, Willem can do well.
I expected Ramesses to do better than Mansa (I've already mentionned that: Ramesses + Mansa in a low peaceweight environment seems to benefit Ramesses more than Mansa ; make that the 3rd result set where that holds true). But that much better? Wow.
The thing is, this was another game where Mansa never pursued culture. There were games where he did well, but he was on a space or domination pursuit, and Ramesses or Willem's culture was just faster.

Spoiler Playoff 3 :

No real surprise here.
Julius and Gilgamesh were doomed: they had the death spots on this very unbalanced map.
Augustus was doomed by his peaceweight.
Mao and Cathy had the only starts allowing a real shot at victory, and that's basically all there was to that game.

Spoiler Championship Game :

We should be pretty used to HC winning by now, but such a dominant performance on a mirrored map?
Wasn't expecting that.
An explanation might be that he was bordering Willem who stinks in low peaceweight environments: he certainly conquered the Dutch a lot, and snowballed from there.
Something weird on this map with HC: it would seem he very rarely goes into culture mode there. In this case, the only time he did was a late game "switch to culture when at the end of the tech tree".

Mao finishing second feels a bit like a hold-up: Toku was usually stronger and doing better.

 

[Thrasybulos]

This is mainly a note for my future self, but if you guys have ideas, they're welcome.

Season 3 is almost wrapped up (currently doing Playoff 3), I expect to be posting the results on Wednesday.
Then only Season 2 will be left, so this project is nearing completion.

What will most likely happen next is that it'll stay dormant until AI Survivor 9 happens.

But should I feel the urge to revive it, I had mentionned the possibility of doing it again, while rotating the leaders on the maps one spot (Team 1 => Team 2, Team 2 => Team 3, etc.).
The idea wouldn't be to assess the maps (@Keler has done that, and my other projects are better suited for it), nor to assess the leaders (same: better projects for that). It would be to assess the impact of the maps on the AI Survivor format.
Now, it's occurred to me that the same result could be achieved by rotating the season rosters instead of the positions on the maps: play Season 1 on Season 2 maps, play Season 2 on season 3 maps, etc.
This would have two advantages: see different leaders on each map, instead of the same leaders from different starts, and it would solve the issue (extremely unlikely to happen as I doubt I'd even get to that point) that 6-player games are covered by 6 rotations while 7-player games need 7.

As far as the scoring system goes, I'm not happy the PWR kill points (I knew it was the hardest part to translate).
The system I've used here has the advantage of keeping roughly the same number of points as Sullla's, so it allows for a direct comparison between Sullla's ranking and this one.

That constraint would be gone for hypothetical future iterations: the ranking comparison would be between the iterations, and not with Sullla's.
So what I think would be best in that case is to limit the PWR kill points to one per game.
So we'd have:
Most wins = winner = 5 points
Highest score (winner excluded) = runner-up = 2 points
Most kills (wins then survivals as tie-breaker) = "golden spear" = 1 point

 

[Thrasybulos]

Season 3, Opening Round

Some of these games have been recently discussed, with Atesh publishing his AH here.
I have also played those maps "jumbled style", results here.

Spoiler Game 1 :

Official AH

All in all, pretty consistant with Amicalola's set, the main difference being Stalin performing better in mine, Willem better in his. Stalin had a much better spot, Willem is a better AI.
Quite a few of these games followed a standard scenario: Ramesses eats up Ragnar while Stalin and Kublai carve up Augustus. Then Ramesses and Kublai kill off Willem. And finally, Ramesses faces a 1v2 against Stalin and Kublai... and wins the confrontation!

Spoiler Game 2 :

Official AH

When I played this game jumbled-style, Viccy dominated. And she also dominated Amicalola's set. So I thought this one was a no-brainer...
... and then this happened.

Now, I can explain my results: this was a fairly balanced map by AI Survivor's standards, but some starts were stronger and others weaker. And it turns out the live game draw had Viccy in the weakest spot, and Gandhi and Asoka in the strongest.
So it kinda makes sense.
But then, how do we explain Amicalola's results?
As he noticed, Gandhi's diplomatic situation was far from ideal, so it wouldn't take a lot to have a bad roll for him. But in my games, when Gandhi faltered, Asoka would be the beneficiary.
So while I can see how a bad peaceweight roll could knock off Gandhi, I'm not sure I can explain how both of them underperformed in Amicalola's game.

Spoiler Game 3 :

Official AH
Atesh's AH

My results are closer to Amicalola's than Atesh's: Hannibal dominated, and Peter was a dud.
Now the most remarkable aspect of this set are the victory conditions: he went for Culture almost every time, which he usually never does (unless to throw an otherwise certain win at the end of the tech tree).
I think the reason was Holy Cities: Hannibal would found some religions on account of being the tech leader, and the only AIs who'd pursue religion here were SB and Sury... both of which Hannibal would conquer very often.

Spoiler Game 4 :

Official AH
Atesh's AH

There as well, my results are closer to Amicalola's (a Justinian roflolstomp), although the overall shape remains the same in all three sets: Justinian first, Saladin second.

Is my game broken or what? Here as well, check the victory conditions: Justinian, another leader whom I've almost never seen go for Culture, made it his main game plan here!

Too bad I was nearly done with my set when Atesh published his writeup: I would have tracked Justinian's placement of his second city. All I can say, is that in all four of the last games in my set he sent his settler to the "godly" spot.

Spoiler Game 5 :

Atesh's AH

Confirming Atesh's results, with Lincoln being seriously underwhelming from an OP start.
Suleiman was a tad unlucky in my set I believe. There were several games he was clearly winning... and didn't.
In particular Game 18 where he was a full era ahead of Gilgamesh and had a lock on the spaceship victory... except the Sumerian spies disagreed:

Spoiler Game 6 :

Official AH

Elizabeth didn't do as well in my games as in Amicalola's, but her starting position was one of the most OP in AI Survivor's history, so that was enough.
Probably a direct consequence: Qin did better in my games.

Spoiler Game 7 :

Official AH

While HC got the most wins in Amicalola's set, he couldn't replicate that here.
The main story remains the same, though, with the "southern" leaders dominating.

One thing I hadn't paid attention to when I played this jumbled style: HC may not have drawn the worst start ever as far as the surrounded land is concerned (for instance, on this map, Toku's was arguably worse), but I believe he had the worst BFC ever. This was clearly a Sullla edit, as that start wasn't "legal": the normalizer ensures at least 4 ressources in the BFC (2 food, 1 luxury, 1 strategic is standard), and will add forests to the starts barely meeting that standard. Here, HC had a grass Cow and a Silk. And that's it.

Spoiler Game 8 :

Atesh's AH

So, Pacal did way better in my set than in Atesh's.
But... look at that streak: he won 2/12 in the first part, and then 6/8 at the end!
So it might be a sample size issue, and he might have been overperforming in my set?

 

[Genghis the Barbarian]

Season 3, Opening Round

Some of these games have been recently discussed, with Atesh publishing his AH here.
I have also played those maps "jumbled style", results here.

Spoiler Game 1 :

Official AH

View attachment 711238

All in all, pretty consistant with Amicalola's set, the main difference being Stalin performing better in mine, Willem better in his. Stalin had a much better spot, Willem is a better AI.
Quite a few of these games followed a standard scenario: Ramesses eats up Ragnar while Stalin and Kublai carve up Augustus. Then Ramesses and Kublai kill off Willem. And finally, Ramesses faces a 1v2 against Stalin and Kublai... and wins the confrontation!

Think in many of Amica's games, Stalin got shafted by really bad barb spawns. Looks like we can say he was not a total fraud from this set

Spoiler Game 3 :

Official AH
Atesh's AH

View attachment 711240

My results are closer to Amicalola's than Atesh's: Hannibal dominated, and Peter was a dud.
Now the most remarkable aspect of this set are the victory conditions: he went for Culture almost every time, which he usually never does (unless to throw an otherwise certain win at the end of the tech tree).
I think the reason was Holy Cities: Hannibal would found some religions on account of being the tech leader, and the only AIs who'd pursue religion here were SB and Sury... both of which Hannibal would conquer very often.

Yup, Hannibal almost always went for Culture in my games. I really thought GK would do better, but I guess he is GK...

Spoiler Game 4 :

Official AH
Atesh's AH

View attachment 711241

There as well, my results are closer to Amicalola's (a Justinian roflolstomp), although the overall shape remains the same in all three sets: Justinian first, Saladin second.

Is my game broken or what? Here as well, check the victory conditions: Justinian, another leader whom I've almost never seen go for Culture, made it his main game plan here!

Too bad I was nearly done with my set when Atesh published his writeup: I would have tracked Justinian's placement of his second city. All I can say, is that in all four of the last games in my set he sent his settler to the "godly" spot.

I have a feeling that my set may have seen the fanatics high-roll on PW which helped FDR and Darius immensely.

Spoiler Game 5 :

Atesh's AH

View attachment 711245

Confirming Atesh's results, with Lincoln being seriously underwhelming from an OP start.
Suleiman was a tad unlucky in my set I believe. There were several games he was clearly winning... and didn't.
In particular Game 18 where he was a full era ahead of Gilgamesh and had a lock on the spaceship victory... except the Sumerian spies disagreed:
View attachment 711246

Ha, this is pretty much identical to my results except Suleiman was definitely better.

Spoiler Game 8 :

Atesh's AH

View attachment 711249

So, Pacal did way better in my set than in Atesh's.
But... look at that streak: he won 2/12 in the first part, and then 6/8 at the end!
So it might be a sample size issue, and he might have been overperforming in my set?

I'm wondering if 1) my three Boudicca wins should have been Pacal wins (I think she ran over Pacal in all three of them due to Pacal's Rifling allergy) and 2) if Pacal high rolled on PW to his detriment

I knew Shaka's blistering start to my games was a fluke.

 

[Thrasybulos]

Season 3, Final rounds

With Playoff 1 featuring four high peaceweight AIs, and Playoff 2 at least three (and those being strong AIs), was Season 3 the actual season for the high peaceweights ?

Spoiler Wildcard Game :

If there's one thing which has surprised me when running these AH, it's that those big games are no less predictable than the smaller games.
The three high peaceweights were dead meat here, but it was Charlie, not Washington, who was able to grow strong at times. And Charlie's good games were usually at HC's expense.
JC would have been a major disappointment if I'd expected anything from him (the Stalin fraud or not discussion has been reignited, but as far I'm concerned, JC's case is settled). Funny thing is that most of his deaths were courtesy of... Brennus. That made little sense.

Apart from that, the story of this game was pretty simple: if HC wasn't Charlie'd out, he put a timer on the map.
... a timer that Napoleon was more than willing to meet! Apparently his starting position was that good that it enabled him to race one of the best Financial civs to Space!
Shaka was usually the Eastern winner, and the strongest civ on the map... but pulling a Domination win on those large maps is hard, and time usually ran short for him.

Spoiler Playoff 1 :

A 4v2 setup is usually very bad news for the low peaceweights, especially when one of them has drawn a tundra start!
Ramesses' central start was the strongest position on the map, with the AI in that spot usually off to the best early game. It was also the most exposed, as experienced by Asoka in the live game.
Here though, De Gaulle would most often attack Asoka instead, leaving only a weak Justinian as a potential early threat. It turned out that Bismarck was actually Ramesses' main early-to-mid game danger.
Lincoln, except for his one domination win, played a completely passive game, teching in his corner and not getting involved in the world's affairs. And while that failed as you'd expect, and allowed De Gaulle to exceed expectations, it suddenly started to work in the middle of these runs, with Lincoln chaining wins. I'm actually happy that streak was cut short as Lincoln making it to the Championship through playing non-games would have felt very unsatisfying.
Bismarck was a disappointment of sorts: not that his AI is conducive to great expectations, but his start was decent and safe enough. I guess his lack of results can be explained by the fact he usually targeted a strong Ramesses instead of a weak Lincoln.

Spoiler Playoff 2 :

When I played Jumbled Rumble Season 2, I picked this map for the Championship as it seemed the most balanced among AI Survivor Season 3 6-player maps. The top left position was significantly weaker, but all other starts seemed to offer a decent shot at victory, albeit for different reasons. The bottom positions were stronger, with more room to expand, and the middle one was the strongest, but very exposed to invasion. The top right start was rich, with Stone + Marble allowing successful tall play, and it could claim the Eastern peninsula if bottom right was involved in early warfare. Top center allowed for opportunistic dogpiles, and had the weak top left as a natural conquest prospect.

The setup for this game, 3 of the strongest high peaceweight vs 1 strong low peaceweight and 2 mediocre low peaceweights, seemed to heavily favour the high peaceweights. But Zara had the weakest start, while Pacal had the strongest. Then again, Pacal was pincered between Gandhi and Mansa... All in all, I thought it would make for a balanced and undecisive game.
Well...
Guess I read that wrong!
Atesh commented on another game than Zara was no ally of Gandhi's, and that certainly proved true here, as Gandhi would often die early to a combined assault from Zara and Mao, with Mao as the main beneficiary. Napoleon would fight Mansa nearly every game... which left Pacal to grow into a juggernaut game after game.
Mansa and Gandhi did pull off the odd wins, but the setup turned out to be heavily in favour of the low peaceweights!

Spoiler Playoff 3 :

HC landed the best start on the map, next to the doomed AI: no surprise that he should end up as the heavy favourite here.
Saladin's start had lots of available land, with no commerce whatsoever: when I played Jumbled Rumble, it was often the crash-your-economy-hard spot.
Mehmed's start was pretty poor (squeezed, jungle).
I was surprised that Hannibal, from another poor start, would end up as HC's runner-up: but he was ideally placed to benefit from Elizatbeth's obliteration, and Kublai seemed to prefer targetting the tough Ottomans.

Spoiler Championship Game :

Not sure I got the real story here: streeeaky again
It started as a complete Hannibal roflolstomp, then HC seemed to be on his way to remind everyone who's boss, until De Gaulle decided to stake his claim!
In the end, Hannibal's early momentum carries him through.

When I saw Hannibal being dominant early on, I remembered Saladin had won Season 1's Championship from that start (although it made sense there, as he was bordering the doomed AI).
But it got me to wonder whether there was some subtle unbalance on the map. Those doubts were somewhat alleviated by the latter half of the runs, but I still played 5 games after that while inverting Hannibal and HC's positions. That yielded 2 Pacal wins and 3 HC wins. Doesn't prove anything, but food for thought?

 

[Thrasybulos]

Season 2, Opening Round

That season was played with the Arid Climate setting.
As Eauxps has observed elsewhere, it led to the most distinctive maps. But contrary to him, I'd be very hesitant to suggest revisiting that setting: it also led to the most unbalanced maps, and while there were a few upsets, for the most part, the game outcomes were largely dictated by the map, far more so than with more standard settings.

Spoiler Game 1 :

That first game was absolutely shocking to me.
We've have that discussion in the pregame to S8 Wildcard 2 where Cathy was isolated in a high peaceweight field and I expressed the view that making her a favourite was pure lunacy: a lone low peaceweight in a high peaceweight field is as much very dead as a lone high peaceweight in a low peaceweight field.

And yet... here we are, with Mao actually ending up as the winner.
Then I checked Keler's data, and it suddenly made sense: his starting position was one of the most unbalanced in the whole dataset!
There is a difference between the lone "baddie" vs the lone "goodie" scenarios: the inevitable dogpile may be delayed in the former case. Which doesn't matter: high peaceweights tend to be better techers, so even if they skip Rifling for some time, that dogpile will eventually involve superior military tech and lead to the expected conclusion.

Except here... Mao's starting position was so OP that it put a timer on the game. If left alone, he'd not only grow larger and stronger than the others, but also outtech them, even the Financial English queens and Zara!
The dogpiles did come: 60% survival rate only, and some of those survivals included stuff like

(Mao is in yellow, and he starts at the opposing end of the map...).
But they often came too late (funnily, tiny Viccy was his bane, with most of her kills being Mao kills!). And since no other leader truly emerged as a dominant force, 6 wins were enough to carry the day.

Spoiler Game 2 :

Not much to be said about this one: HC was extremely dominant, DG slips into second position on account of being the most successful warmonger.
Probably the most remarkable aspect was that HC never went for the Culture plan here (his only Cultural win being a late "accidental" win).

Spoiler Game 3 :

My first 100% win rate!
It was simply the dream setup for Gandhi: lots of room to expand into, two meatshields, a favourable diplomatic situation.
Even though Saladin and Gilgamesh are no slackers, the odds were stacked against them here: they did sometimes manage to emerge as the dominant AIs, but could never beat the tight timer Gandhi put on the game.

Spoiler Game 4 :

Once Charlie and Bismarck were dispatched, Justinian's "cannot-plot-at-pleased" proved his Achilles heel in this setup.
Cathy had no such qualms and could keep the snowball rolling (also helped that according to Keler's data, she had the best start).
Poor Charlie was triple-doomed: by his peaceweight, by being squeezed between the two strongest AIs on the map, and by his awful start!

Spoiler Game 5 :

Sullla had played a few AH for this game which had shown Shaka and Toku as the dominant AIs on this map iirc.
Well, scratch Toku: Shaka was the only dominant AI here (Toku is indeed the second best-performer, but waaaay behind). And indeed, Freddy as an available early snack is largely responsible for that.

Spoiler Game 6 :

The high peaceweights dominated this field, and Ramesses was the stronger here (and funnily enough, he never went for Culture, laying the smack instead before heading off for the stars).
Hatty was the most exposed and died often in the early runs, but the streak was inverted at the end and she gets the best finish, with more wins than Mansa in the final tally, but he still beats her with a better overall performance.

Spoiler Game 7 :

SB and JC had the best starts by far here... and while SB completely failing to achieve anything should come as no surprise, I bet JC still has some fans.
Well, he certainly did way better than SB (especially in the department), but he ought to have been dominant here... and wasn't. Only the last four games where he won 3/4 redeemed him somewhat.

You could certainly argue that Kublai was even more disappointing... but his start was bad.
Qin, on the other hand, did very well from a starting position which Keler has at < 5% win rate and < 30% survival.

Spoiler Game 8 :

Suleiman, Pacal, Cyrus had the best starts (in that order).
Suleiman and Pacal delivered, Cyrus... not so much.

The worst starts were: Sury, Washington, Genghis.
So with the exception of Napoleon very slightly outperforming Cyrus, the results here seem completely dictated by geography.

 

[Eauxps I. Fourgott]

Quite the set of results. I nailed some guesses - but was completely off on more. (How in the world did Joao get the most wins from that spot???)

If geography was indeed that big of a factor, I wonder if "normal" alternate histories with Deity starting techs will be more different than with most seasons, by helping to smooth the worse starts somewhat.

 

[Thrasybulos]

If geography was indeed that big of a factor, I wonder if "normal" alternate histories with Deity starting techs will be more different than with most seasons, by helping to smooth the worse starts somewhat.

Or it might on the contrary make matters even worse? Enable any AI to immediately make good use of its better land?
Only one way to know...

 

[Thrasybulos]

Season 2, Final Rounds

Spoiler Wildcard Game :

Hatty was doomed in this setup: peaceweight, poor start, squeezed between enemies.
Louis ended up being the monster AI in most games, with the Qin vs Napoleon rivalry often determining which of the two had a shot at competing for the win (with Napoleon usually getting the upper hand).
Only Willem, when left alone in his corner, seemed to also have a chance at winning... but then again, he won 3 games in a row then died most of the time. Outlier streak?

Spoiler Playoff 1 :

That game hasn't left a strong impression on me, so I've essentially already forgotten about it.
Looking at Keler's data, Joao had by far the strongest position... but was doomed by his peaceweight, which changed the dynamics of the map. Cathy and De Gaulle were the best placed to take advantage, but DG's start was a real stinker, so Cathy ended up being the main beneficiary.
In a favourable diplomatic environment, Pacal was able to score quite a few wins, but Mao beats him for second place thanks a better overall showing.

Spoiler Playoff 2 :

When I set up the map, I thought that Mansa's start was OP: lots of room, three early happiness ressources, and I even made it stronger by turning the illegal BFC desert tiles into plains (which meant a PH start for him).
So I wasn't surprised when he started dominating.
Then I had a look at Keler's data... and apparently HC's starting position was the strongest!
So what happened?
First, let's mention, once again, the streakiness of the results: Mansa dominated the first 13 games, then it was an HC stomp. So... more games would have been needed to get a reliable picture imo.
That aside, the two weakest starts were Shaka's and Napoleon's: Mansa's neighbours, who were also the most likely to start an ill-advised war against Mansa. That meant that game after game, Mansa would get to conquer one of those early, and thus get to critical size early.
Plus, Roosevelt served him well as a dogpile prevention measure.
So all in all, those results seem to make sense.

Spoiler Playoff 3 :

Another very unbalanced map, and Justinian lucked into the right spot. The fact he's a pretty damn good AI to boot sealed the deal.
Gandhi was the sacrificial lamb here: Justinian usually started by conquering him, then snowballed from there. But in the rare cases he didn't, Ramesses conquered India instead, and that usually propelled him into a dominant position.
Julius and Louis were almost as likely to fight one another than to team up to tear apart Pericles, which meant that when the dust settled in their western corner of the map, it was too late to stop the eastern juggernaut.

Spoiler Championship Game :

And... another HC roflol stomp. *yawn*
Funny thing, he went for culture every single time here, when with similar setups on the S4-7 Championship, he almost never did. That's weird, and would propably deserve investigating to get a better insight into the AI's inner workings?
I thought Mansa + Ramesses would get a shot at times, but that almost never happened. Ramesses won one, and there was a game (game 9?) when HC, Rammy, and Mansa all went for Culture, and were very very close to one another... but HC still won that one!
Maybe if they had shared a border?

 

[Thrasybulos]

I don't have time today to write conclusions about that experiment, it'll have to wait for a coupla weeks unfortunately.

But here is anyway the final ranking that it yielded (left is Sullla's ranking, right is mine):

PWR totals are very similar: 1073 vs 1093.