Sullla's AI Survivor Season Eight - Championship Thread

[Eauxps I. Fourgott]

Sullla's AI Survivor Season 8 will conclude in two weeks, on September 27, on Sullla's Twitch Channel at noon EDT.

It all comes down to this! After a wild ride, we're at the Championship, this time a matchup between six different peaceniks - even the only low peaceweight, Qin Shi Huang, has more peaceful AI tendencies! The favorites are surely the Financial duo of Mansa Musa and Elizabeth, the former looking to become AI Survivor's first two-time champion, but Gandhi the ultimate peacenik is a contender as well. The less impressive Augustus Caesar and Churchill will have long odds, and doubtless little support. This match will also decide the fantasy contest winner: antisocialmunkey, j_mie6, and Kjotleik are the favorites, but an upset by Bellomorphe or Henrik is still theoretically a possibility.

Check out the video preview, written preview, and make your own prediction. (And view the current picks!) As always, all are welcome to discuss the conclusion to the season in this thread!

Finally, there's one more chance to take part in @Fippy's picking game!

 

[Noble Zarkon]

Sulla with his thumb on the scale giving Qin Skirmishers lol.

 

[lymond]

Balanced Map. I expect Mansa will be the overwhelming favorite. Qin FTD.

 

[Fippy]

Balanced Map. I expect Mansa will be the overwhelming favorite. Qin FTD.

Maybe but i think Gandhi has equal chances. It's his dream scenario (minus Mansa)

 

[Thrasybulos]

Maybe but i think Gandhi has equal chances. It's his dream scenario (minus Mansa)

I think Game 2 there is the one which has definitely removed Gandhi as a top leader (in specific circumstances) in my mind.
It was also a "dream scenario" for him, and he failed miserably.

I'm now of the opinion that Gandhi is a leader who always has a chance to win (barring extreme peaceweight circumstances), but can never be considered a favourite. He's just not good enough.
Elizabeth here for instance has basically the same game plan, but she's just better at it than him.

 

[Eauxps I. Fourgott]

One last seeding watch for the season!

Now that Churchill has displaced Zara Yacob in Pool Two, we're probably finally set as far as the seeded leaders go. But there's one more chance: If Augustus pulls off the win here (or possible a runner-up finish with at least two kills), he joins Pool Two as well! Depending on what Churchill does here, this would displace either Churchill or Charlemagne.

Churchill himself can move up quite a bit with a good performance here and make his new seeded status a bit more secure. If he repeats his Season 6 Championship performance, for instance, he'd move up to the #11 seed ahead of Darius. Pool One is out of reach, but if he does a three-kill Domination win or something like that, he can move up to the top of Pool Two.

Gandhi is probably staying put unless he wins, but if he does win, he'd move up into low Pool One (either the sixth or seventh overall seed depending on if he scores kills) and knock Stalin down into Pool Two.

As for the other three, Mansa is of course already the #1 ranked leader, but could pad that position here. Liz and Qin are far short of seeded status and even killing everybody else on the map here wouldn't get them there.

---

We have a new Championship map! I've been waiting for this for years as I never really liked the old wrapped one. And of course, a new map means new dynamics!

I happen to think that this particular map favors peaceful leaders. It's worth having that contrast when previous maps have favored the warmongers - I don't know if it's possible to get a symmetric map that doesn't favor one side or the other. And it's certainly fitting for the year of the peaceniks! Of course, saying that this map favors peaceful leaders is a bit less meaningful when EVERYBODY is a peaceful leader. But I think that it will especially favor the more economic leaders of this bunch; they'll be a bit more sheltered, which makes it that much less likely that they get knocked out before they can pull too far ahead. That would give this map two categories of leaders: the three contenders, and the three also-rans.

Qin Shi Huang is very dead meat and I'm sorry that we do have an obvious outcast in this group. I mean, technically, he could do something like snowball around the map counter-clockwise, while Augustus and Churchill don't do much of anything, but his only advantages are Industrious and starting with Agriculture, and he's the peaceweight outcast. Any successful outcome for him would be exceedingly unlikely (granted, so was the finish that got him here in the first place), and for him to somehow pick up his first career win in this, of all games, would be the most trollish outcome possible. He's more likely fodder to fuel early conflict. Augustus Caesar and Churchill will also have very long odds, as they simply don't have the economic chops of the others. I think they'd need to do significant conquering to stand a chance, and even then, they could easily be vulnerable to a cultural attempt, like what happened to Churchill in the Season 6 Championship. And it's quite likely that they become pleasedlocked with the rest of the map at some point. Successful snowballing by either of them seems to fall under the "possible, but quite unlikely" banner. Of the two, Augustus is probably better situated as he neighbors easy prey in Qin. Churchill has the most sheltered position on the map, bordering the two other leaders who don't declare at Pleased, but that could be a curse as much as a blessing.

That leaves our three economists. I was saying at the start of the playoff round that I think this season belongs to a Financial leader, and I do still think that's the most likely: Mansa Musa or Elizabeth. Liz will be able to focus a bit more on pure teching while Mansa has the chance to expand through Qin; I think their chances are roughly even. Then there's Gandhi: can his pure builder game and cultural bent overcome the advantage of Financial? I'm inclined to believe that in most circumstances, it can't, at least not on this map. I view Gandhi as the underdog, the most likely to come on top if the Financial pair fail, but with much worse chance than them.

What this game will really come down to is when and where conflicts break out. That's always the case, of course, but in other maps that's a smaller factor, with a "default" winner unless things go awry. Here I don't think there is a "default" winner; whichever of Mansa or Liz has diplomacy break his/her way is probably going to win. Or maybe Gandhi if things go south for both of them, or one of the other three if the game gets really crazy. Qin is obviously the massive favorite for First to Die, while second place is completely up in the air in this field. Looking forward to it!

(Also a note: This is really the season of the Mining starts! All six leaders start with that tech. Interesting.)

The other nice thing about this is that I think I'd be pretty happy with any outcome, especially since I no longer have a skin in the fantasy game. My order of preference, though, from greatest to least, is probably: Gandhi (I think it would be fun to see him scam a win here, and very appropriate for the season), Liz (she deserves it after such a rough rest of her career), Churchill (my boi), Qin (would be absolutely hilarious), Augustus (he's cool and I think underrated), Mansa (he's cool and would be a neat first two-time champ, but probably is the least interesting option here).

For my official prediction, I'm taking Gandhi to beat the odds. Augustus in second place, benefiting off of the first-to-die Qin.

 

[Fippy]

I think Game 2 there is the one which has definitely removed Gandhi as a top leader (in specific circumstances) in my mind.
It was also a "dream scenario" for him, and he failed miserably.

I'm now of the opinion that Gandhi is a leader who always has a chance to win (barring extreme peaceweight circumstances), but can never be considered a favourite. He's just not good enough.
Elizabeth here for instance has basically the same game plan, but she's just better at it than him.

None of them are
I played millions of deity games ( you know what i mean), and have seen them all fail badly.
Mansa, HC, Justi and so on.

Especially Mansa often makes an easy farm target, and there was one video from @Lain where he took his cities with Jumbos, Maces and (ofc most important cannons) against Artillery & Infantry.

 

[Thrasybulos]

Yes, sure, they can all fail. And I also believe that unfortunately leader "personnality" is less impactful than starting position or diplomatic context.
But this wasn't a single game (where anything can happen), it was 20 games played on the same map with varying starting position and an all high-peaceweight field. That's the scenario which Gandhi's results in AI Survivor have led us to believe would be his dream scenario, and where he'd dominate.
He didn't.
And while 20 games certainly can't yield an absolute truth, they do rule out a freak result.

I've seen setups where Mansa would be dominant, others where Viccy would be (the one mentionned for instance), others where Lizzy was. None where Gandhi was, even when the field was apparently stacked in his favour.
Maybe I simply haven't seen them yet.
But until then, I'm off the Gandhi bandwagon.

 

[Fippy]

I always wondered if FIN makes such a big difference for AI games.
AC was maybe even more impressive than Mansa in his runaway playoffs win.
(his rate)

Maybe it's more about what they do in general, did we see Ragnar build a good economy yet?
I saw him being the tech leader in some of my games (aka with a human involved).
I know aggressive AI setting really hurts warmongers, they can be surprisingly opportunistic without it.
In Survivor games their personality and decisions spin out of control.

 

[Eauxps I. Fourgott]

I always wondered if FIN makes such a big difference for AI games.
AC was maybe even more impressive than Mansa in his runaway playoffs win.
(his rate)

Maybe it's more about what they do in general, did we see Ragnar build a good economy yet?
I saw him being the tech leader in some of my games (aka with a human involved).
I know aggressive AI setting really hurts warmongers, they can be surprisingly opportunistic without it.
In Survivor games their personality and decisions spin out of control.

I'm pretty confident that when I run the alternate histories for the Championship (as I hope to be the one doing those), they will pretty thoroughly show that Mansa and Liz are better techers than Augustus from equal land. Augustus's land in the playoffs was extremely good.

 

[Keler]

Religions are surprising and very random. We saw what Churchill did to Victoria. And we could have seen live but ended up seeing in alternatives what Joao was about to do to Gandhi this season. I don't know if Gandhi will be safe or not.
We also saw Brennus&Washington love fest until one of them converts to free religion. Would that happen if Augustus and Qin follows same religion?
Qin gets unlucky and misses rifling sometimes but since everybody writes him off before that era anyway, can he slow down Mansa Musa? He is not so easy to destroy either. What would Augustus do for instance, if he gets pleasedlocked with Qin for the earlier half of the game, such as first 200 turns?! Can he plot his first war on anyone else with different religion on the other side of the map?
This map has so much more coastline. Will the Great lighthouse and Collosus matter more? Will be there so many coastal cities?

Well my thoughts goes like this. I won't vote for Mansa Musa to win, that's a promise.

In my ranking currently, Qin and Mansa very close, the best two ais in this game. Around 15 out of 52
Elizabeth, Gandhi and Augustus are very close too. They are around 30th more or less. It keeps changing the more games I autoplay.
And Churchill has been quite bad. Just under 40th out of 52 or hangs around 40-42 mostly.

Who knows maybe he indeed snowballs over Gandhi and wins space or diplo. Gandhi ftd is in play maybe after all.

Maybe it is Elizabeths turn to win the game. I will vote Elizabeth/Augustus or Churchill somehow. Now I have to decide to select the ftd, beteeen Gandhi and Qin, maybe I repeat my past picks and click Mansa Musa ftd just like I did in his oppening round will I need a miracle again for that here?

 

[ManiaMuse89]

I always wondered if FIN makes such a big difference for AI games.
AC was maybe even more impressive than Mansa in his runaway playoffs win.
(his rate)

Maybe it's more about what they do in general, did we see Ragnar build a good economy yet?
I saw him being the tech leader in some of my games (aka with a human involved).
I know aggressive AI setting really hurts warmongers, they can be surprisingly opportunistic without it.
In Survivor games their personality and decisions spin out of control.

In my tests financial and spiritual are about equal best traits for AI games. There is a little jump down to Creative, Charismatic and Industrious and then all the other traits are quite a way below those.

However as you point out, there are exceptions including Ragnar. He is currently 2nd last in my rankings and rarely does anything of note and has one of the worst death rates! Elizabeth is bottom 5 as well and has even more of a struggle to survive. But it is made up for by the financial leaders who do have a lot of success (Huayna Capac, Willen, Pacal, Darius, Mansa, Hannibal).

 

[Keler]

In my tests financial and spiritual are about equal best traits for AI games. There is a little jump down to Creative, Charismatic and Industrious and then all the other traits are quite a way below those.

Will you share your results some day?
I believe you keep your record. I want to see. I will share what I got right after season ends.

Ranking traits is tricky, especially for the AI having them, instead of human player. We probably need all AI programmed in same parameters, with tech flavors, unit/wonder build ratings etc, except maybe for peace weight so wars can happen, aggressive,protective traits could be in play and also symmetric custom maps should have been used such as old championship map. But since we have it more complex

This is what I got in my tests.
Traits for AI having them is this:

A Tier: Creative, Financial
B Tier: Industrious, Imperialistic, Spiritual
C Tier: Organized, Expansive, Charismatic, Protective
D Tier: Philosophical
E Tier: Aggressive

Of course this tier list can be outrageous for some
This might be quite random too.

I also found out that Wang Kon is a terrible terrible leader with financial trait... need someone to examine on that. Ragnar is finer.

 

[ManiaMuse89]

Will you share your results some day?
I believe you keep your record. I want to see. I will share what I got right after season ends.

Ranking traits is tricky, especially for the AI having them, instead of human player. We probably need all AI programmed in same parameters, with tech flavors, unit/wonder build ratings etc, except maybe for peace weight so wars can happen, aggressive,protective traits could be in play and also symmetric custom maps should have been used such as old championship map. But since we have it more complex

This is what I got in my tests.
Traits for AI having them is this:

A Tier: Creative, Financial
B Tier: Industrious, Imperialistic, Spiritual
C Tier: Organized, Expansive, Charismatic, Protective
D Tier: Philosophical
E Tier: Aggressive

Of course this tier list can be outrageous for some
This might be quite random too.

I also found out that Wang Kon is a terrible terrible leader with financial trait... need someone to examine on that. Ragnar is finer.

Creative is #3 trait at the moment for me, quite close to spiritual and financial. Aggressive is also the worst trait in my results sample. Ditto Wang Kon is a bottom 5-10 financial leader for me as well. Dies a lot without doing very much and very rarely wins.

Yeah I will share my results at some point although I want to wait until I have a ludicrously large sample to try to reduce some of the randomness.

Basically my set up is Sulla's settings, AP removed, correct starting techs and units but then archery removed (otherwise the AI won't move their starting settler straight away if they just start with warriors), 6/7 AI games on Pangea medium/low sea level maps. I don't care about making the maps/starts fair other than making sure that there are no islands. I don't worry about barb galley ponds in the ice caps because the navy spamming is relatively rare and one day the AI will build submaries. I am running it with a round system so each AI has played the same number of games although AI combinations are completely random in each round. As a result some AIs have played each other more often than others and some have had more 6/7 AI games but it should average out over time though.

There is a lot of noise in the results over a smaller number of rounds, especially in the midtable, but over time I am seeing groups of AIs pull ahead from others. So I do believe that some AIs are genuinely better than others and that it isn't completely random.

I am using a fairly arbitrary scoring system with a win score based on the starting number of AIs in the game (winning a 7 AI game gives one more point than a 6 AI game) and a survival score based on the starting number of AIs and the number of AIs that survive. Failing to survive scores 0 points. I don't care about rankings in games other than win/survive/die because a lot of 2nd place finish places are pretty random depending on who the big dog decides to pick on at the end of the game. My theory is that 'you have to be in it to win it' so if an AI is good at surviving to the end of the game then it means that they had at least a chance of winning. I'm not being that obsessive with recording things like number of kills or wars but I have been recording victory conditions and finish date which is giving quite interesting results.

I use the autoplay mod because I seem to have issues with using the console command. It does affect the game slightly because the AI controlling the human player will sometimes sign defensive pacts, be bribed into a war and will vote in the UN but I don't think it influences the results too much. The AI will end a war eventually if they can't access their enemy.

Without spoiling everything, I think I have seen enough to draw some conclusions:

- The AIs which are single-mindedly obsessed with pursuing cultural victories are probably the most consistent at winning. In that group I would put Huayna Capac, Willem, Mansa, Gandhi, Hatshepsut, Ramesses and Louis. The leaders which commit to an early cultural victory attempt often manage to get it done before the other AIs have a chance to take them out. That in itself is probably explains why spiritual is so high up in my trait ranking because 4 of those leaders are spiritual (which increases the chance of going for culture victory from early on in the game). 3 of them are also financial which can power up a culture victory attempt if they have nice land.
- I am convinced that Napoleon is the best warmonger. He seems to be the most balanced of the warmongers and went on a mega winning streak at one point. Fun to watch as well because he often declares cross-continent wars like a madman.
- In fact, all the French leaders have done surprisingly well.
- The biggest outlier compared to Sulla's power ranking is probably Suryvarman who in my tests is a consistently bad leader, bottom 10 for sure.
- The leaders at the top and the bottom of my rankings both begin with H. I will let you guess which ones they are

 

[Fippy]

Hammurabi top HC bottom?

 

[ManiaMuse89]

Hammurabi top HC bottom?

Yup, Hammy best leader of all time, HC a complete fraud

 

[urbestfriend]

Question since I'm pretty terrible at judging this sort of thing - how large is the championship map in terms of tiles, compared to a randomly generated pangea map usually used? It seems to my eye it's larger in terms of tiles but I might just be imagining it. Either way the capitals are at least much further away, which makes peaceful victory even more likely, as if it needed any help with these leaders.

Since this may be a little less... explosive of a championship than usual, could we get a bonus round after where we drop 6 Montys in the center and then replay the map

 

[Thrasybulos]

I know aggressive AI setting really hurts warmongers, they can be surprisingly opportunistic without it.
In Survivor games their personality and decisions spin out of control.

Which reminded of a test I'd been meaning to do for quite some time now: running AH with Aggressive AI turned OFF.

So today, I started with Season 7, game 1.
First game was a turn 319 Isabella Domination victory with 5 wars total.
Second game was an Hammurabi turn 390 Spaceship victory where Shaka, after absorbing Pericles and Gandhi, called off his snowball to turn on the culture Slider instead, even though he controlled a mere 4 Holy Cities and obviously was in no position whatsoever to pull off a Cultural victory.

I don't think I need to pursue that testing line.
Sullla was in fact 100% correct to run this with Aggressive AI.

 

[ManiaMuse89]

Question since I'm pretty terrible at judging this sort of thing - how large is the championship map in terms of tiles, compared to a randomly generated pangea map usually used? It seems to my eye it's larger in terms of tiles but I might just be imagining it. Either way the capitals are at least much further away, which makes peaceful victory even more likely, as if it needed any help with these leaders.

Since this may be a little less... explosive of a championship than usual, could we get a bonus round after where we drop 6 Montys in the center and then replay the map

Sulla mentioned in his preview video that he thought the first version of the map was too small so he asked for it to be revised with more land around the starting areas.

 

[a pen-dragon]

Also consider that this map has a significantly longer coastline than a pangea. This means that even with the same number of tiles there is room for more cities, due to coastal cities (obviously) occupying less land tiles. Also lake tiles matter, since they spread out the land more, allowing for more cities.

Counting tiles, each arm of this map has approximately 130 tiles. The center is somewhat bigger, maybe 220 tiles. Also there are 62 lake tiles. In total 1000 land tiles and 62 lake tiles. The total coastline is maybe 300 tiles.

The map for Playoff 3 has approximately 800 land tiles and 11 lake tiles. The total coastline is maybe 160 tiles.

With those numbers I estimate, that this map has room for maybe 30-50% more cities. (Or 2, maybe even 3 civs.)