I appreciate the write-ups, but I'll likely skip ones more than a few sentences long due to the time to read them all compounding over 20 iterations.
|
AI Survivor - Season Four and Five Reruns
|
|
By all means, if you and others find all the details to be interesting, you can keep them around. I do want to know the general gist of how the results came to pass, but for these, I personally don't have a lot of interest in all the nitty-gritty details - what wars were declared and the city counts at those times, what who was researching when, ETAs when culture slide was turned on, etc. That's simply more detail than I really am interested in - the big war declarations and results, any particularly wacky happenings, and how the winner won would probably constitute the stuff that appeals to me.
But you certainly don't need to be bound by my preferences, especially if you're noting all the stuff down anyway. I can always skip extra details if I want to.
I'm adding in that I appreciate the writeups, they have some valuable information in them and were rather fun to read (though I probably prefer the tone of the normal Survivor writeups). I particularly like seeing who fights who, as I like to see how that that turns out (I really wanna see how often Louis and Pacal war when we get to that for example).
So I had to know: was Asoka winning last week's playoff game a fluke? The result was an early version of an alternate histories investigation for Playoff Game Two. Full details are at the following link but I'll toss out some teaser screenshots here: http://www.sullla.com/Civ4/civ4survivor5-11A.html
![]() ![]() (Note : "A" column tracks the number of war declarations initiated by the AI, "D" the number of times the AI is declared upon, "F" the points for finish ranking, and "K" the number of kills.) ![]() Thank you to Wyatan for providing the inspiration for this project.
So Asoka was in fact the strongest contender to die, but even that's just a coin flip.
The hilarious conclusion I'm drawing from this is that yes, Augustus really is that inert. He's just chilling off in the corner so much of the time, doing okay but never winning.
Very interesting results, Sullla. Thanks for doing these alternate games and posting about them.
It is nice to have some theories confirmed after the actual Playoff Game Two went in a very unexpected (for most of us, at least!) direction.
An other way to do this multiple simulations would be the usage of the Pitboss mode. After the game a replay file will be generated which can be viewed over the hall of fame. The war declarations etc. can also be found in the logs.
Problematic is just the space launch bug of vanilla BTS. Code for Bugfix:
I actually find those results pretty surprising.
The prediction I was the least confident with was Caesar winning. I evaluated his odds at about 40%. And apparently, that was spot on. I was a lot more confident about Suryavarman being the runner-up: I would have said about 50-60%. And that was just wrong. I was dead certain about Asoka FTD chance being 85%+. And the actual odds were much lower. Oh well. At least it provides one more very anecdotal piece of evidence for my growing theory that, very broadly speaking, leaders who can plot at pleased make the best candidates for the win (they need to be cutthroat and to be able to exploit all opportunities), while leaders who can't make the best candidates for second place (the ability to stay under the radar is particularly relevant in that case). I was away last week-end, and this week's been a tad more busy than usual, so I'll publish Game 3 results later than usual (end of the week-end at best). @Ram, Thanks for the suggestion, but I'm actually watching the games. I guess that if I were only interested in extracting statistics, I'd try and find a way to fully automatize the process.
Season 4, Game 3 (1/3)
This is a game which saw an ultra-dominant Julius Caesar win an early victory. Which is something he'd already accomplished quite a few times, as he's one of the best performing AIs in AI Survivor. But the way he did was extremely unusual: instead of going on a romp and conquering the world with his legions, he stayed at peace thoughout, building up to an overwhelming technological lead before blitzkrieging a rival in the late game, and achieving a super early spaceship victory. It basically felt like Caesar had hired Mansa Musa to oversee his civilization, while he retired to a country resort to play wargaming boardgames with a few friends. Was Caesar supposed to be that dominant on this map? Was there something in this game composition that caused him to renounce his warmongering ways, or was that a complete fluke? Let's see: (see the following posts for the game writeups) Now, before I comment on each leader's performance, I would like to start with a general observation. If you split these results into two groups, games 1-9 on the one hand, games 10-20 on the other hand, you get two different stories. The first tells us that: - Joao, Shaka and Frederick are tied for runner-up position. - De Gaulle wins once, but dies every single time aside from that. - Suleiman is the grand favourite for first to die (5 out of 9), who gets dogpiled right away without taking any active role in the game (he declares 3 wars while being declared upon 28 times, and scores a perfect 0 point). - Shaka has a better than average chance of making it to the finish line (5 out of 9), and performs decently (33% runner-up finishes, average 1 kill / game). The second story tells us that: - De Gaulle is the grand favourite for runner-up position (7 out of 11), with Frederick a distant second (3 out of 11). - Joao is the favourite for first to die, albeit a weak one (4 out of 11). - Suleiman almost always dies (10 out of 11), but not first (1 out of 11 only), and not without putting up a fight (12 wars declared, 4 kills). - Shaka performs very poorly, with not a single placed finish, averaging 0.5 kill per game only, dying 10 times out of 11, and being the second most likely candidate for first to die (3 out of 11). I think this a reminder that 20 games remains a small sample, not exempt from streaky runs or outlier results. It's still a far cry better than a sample of one, though, and I believe it's enough to see the main scenarios emerge. But with each game accounting for 5% of the total results, this experiment only provides a coarse-grained picture. Julius Caesar He absolutely, completely, utterly dominated the game: 17 wins out of 20 games. And one of his only 3 losses was an absurd diplomatic loss, 5 turns before his ship would have landed. So, yeah. I don't expect to witness such a performance level in any other game. Certainly not in any Season 5 game. Maybe Huyna Capac's Season 4 opening round game? But then again, Huyna Capac does throw games at times... And with his fondness for a Cultural victory, I can't see him netting enough kills to beat Caesar's incredible 119 score. Here, in nearly every game, Caesar would quickly become the dominant economic and military leader, and eat up his opponents one by one until he reached Domination. So, how to explain that? For starters, Caesar has proven over the various Seasons of AI Survivor that he's a strong leader, belonging in the top tier. Conversely, he wasn't exactly facing a tough opposition in this game. For instance, there was no Financial, or simply strong economy leader to challenge him on that account. There was fearsome Shaka to contest his military supremacy... but the net result is that Shaka's lost quite a few credibility points. But that's not enough. The map, obviously, played a very large role. Starting with that monstrous capital for a Roman civ: three sea food ressources, two silver hills, when Rome's starting techs are Fishing and Mining? Disgusting. But that's not all. Caesar would settle his next two cities along the river flowing from the northeast. A very rich and fertile river valley that would net him Ivory and Gems, in addition to being perfect cottage terrain. Ivory, Silver, Gems: that's three early Happiness ressources, when the AI's early growth is usually stunted by lack of Happiness ressources. And he even had Furs and Gold in the tundra to his south/southwest! And then, there was the neighbourhood: ultra-pacifist and weak Frederick, Joao who was kept busy by the Zulu menace to his north, and Suleiman who was an aggression magnet and thus unlikely to cause trouble. So, a perfect capital, a rich core, an abundance of early ressources, and non-threatening neighbours? A perfect setup, which Caesar's AI was competent enough to fully exploit. He could build up in peace, and then strike whenever he deemed he was ready. Just look at the stats. He was only attacked 11 times, but he initiated 60 wars, with a deadly efficiency: more than half of these wars ended up in him delivering a killing blow. He scored by himself exactly as many kills as all the other AIs combined! Veni, Vidi, Vici indeed. Suleiman So, the "Silly man" as players call him, somehow ended up as a seeded leader in that game. Strong AI then? Well, not here, at any rate: if Caesar's achieved a high score that will be very hard to beat, Suleiman's low score is almost as impressive. He never finished placed. Not once. He died in 85% of the games, and the only three times he survived, it was with a coupla tundra cities about to fall when the game suddenly ended. On average, he's launched an attack once every two games, while suffering an attack three times per game. He's tied with Joao as the leader who was first-to-die the most often. Is he then just a bad leader who lucked himself into a few previous good games? Maybe. But mainly, I think that here he was set up for failure. His land was good. Nothing spectacular, but fertile terrain, a corn ressource at his capital to match his Agriculture starting tech. His ressources were mainly Calendar ressources, so came into play a bit late. His Copper was at risk of getting snagged by De Gaulle (which happened quite a few times), but he had Iron. The terrain doesn't explain his spectacular failure. His position does, though. He was in the middle of the map, bordering every single other AI. Although Imperialistic himself, he had two Imperialistic neighbours to his West (Joao, Caesar), meaning he got to settle very little of his western lands (Caesar invariably settling the gems southwest of Istanbul, a spot that should have been Ottoman), and expanded instead to the east, creating a large and tense border with France. No AI on this map started with Mysticism, which meant Suleiman got to found Islam nearly every time. And that turned out to be bad for him: not only did it slow his start, but it made him early enemies and unreliable friends. For instance, De Gaulle, when he didn't found a religion himself, would more often than not pick up Islam. But the huge border tension between them cancelled any benefit from that, and they still ended up fighting one another. Furthermore, Caesar very often found his own religion, and as the neighbouring founder of a different creed, Suleiman ended up with a huge target painted on his back. Not even mentioning that bordering Shaka ensures an early aggression. Suleiman's setup was inviting dogpiles, and they sure kept coming. De Gaulle The French leader finishes with the second highest score... waaay below Caesar's. He narrowly beats Frederick as the runner-up favourite (see the game 20 report for how that came to pass) with 7 second places to Frederick's 6. He won one of the rare games Caesar didn't. No need to focus on his stats, though: they're not telling the whole story. And that story is that he was by far the best performer on this map, after Caesar of course. Game after game, De Gaulle would expand, peacefully and/or militarily, into a strong second position... only to run afoul of the outworldly Caesar. In the first half of my tests, that happened in the mid game, hence De Gaulle's abysmal survival rate. The later tests saw the confrontation between the game leaders move to the end game, and saw De Gaulle's final standings reflect more accurately his games. De Gaulle, whose AI designer's dubious Dubya views about France are made manifest, is meant to be a sneaky snaky cowardly villain: he's the AI with the highest chance for joining an ongoing war, and with the lowest capitulation threshold. And a peaceweight of zero. With Vassal States disabled, capitulation means nothing in AI Survivor. The propensity for fighting opportunistic wars, on the other hand, should ironically make De Gaulle a strong contender in the competition. And that's what happened in the one game he utterly dominated (Season 3, Game 8). But ultimately, he's not that good of an AI. Mid-tier? Hence his lackluster results so far, with the occasional moment of brilliance. Here there are two things to try and explain: his strong performance, and his even stronger enmity with Caesar. The map explains the first aspect. His terrain was good, with no lack of early Happiness ressources nor of metals, and he had room to expand, which he did. He almost consistently grew to the highest peaceful city count of all AIs. He was neighbouring Frederick, which he would often attack, but at a time of his choosing. And Suleiman, whose partitioning he would often partake into. Alright, that explains how he usually got strong. But he still died a lot, because Caesar hated him. And that's where the peaceweight comes in. Zero is bad. Not only in terms of "in-game moral alignment", but in terms in-game diplomatic position as well. Consider this: Caesar, the untouchable game leader, has a peaceweight of four. That's a 4-point differential with De Gaulle. Frederick, with a peaceweight of 8, is the only one matching that. So in that game, De Gaulle and Frederick were the two leaders Caesar was predisposed to dislike. But Frederick, being Caesar's neighbour, usually picked up his religion. Which improved relations to the point where they opened borders and started trading. Which improved relations even more. De Gaulle, on the other hand, being on the other side of the world, usually picked up a different religion. Which made relations worse. No open borders, no trades, no missionaries. And that's how these two came to be "worst enemies" in nearly every game, with dire consequences for De Gaulle's survival. Frederick There's... not much to say about him? Before the game started, he was the communauty's second favourite (behind the Zulu-bordering Portuguese) for first-to-die: with Caesar as a western neighbour, and peaceweight-zero De Gaulle to the north, it's understandable. But while De Gaulle indeed often showed ill-intent towards Frederick, a Roman aggression failed to materialize most of the time (there were only two instances of that in 20 iterations of the game). So Frederick was actually rarely the first-to-die, and on the contrary, he has the best survival rate of all the non-Caesar AIs. His was a sheltered land, with a short border area (and lots of peaks) limiting tensions. He also found himself protected at times from De Gaulle by Ottoman or Roman culture. Apart from the rare venture into far-away Zulu lands, he essentially stayed in his corner, waiting for one of the game leaders to come and claim his lands, or for the game to end. In the latter case, as the only unmauled civilization left standing beyond the leader (Caesar), he would often finish second, by default. And that's the thing with AI Survivor. The first rule is to survive. If you can't be the top dog, being an overlooked little mouse is a legit strategy. Joao Joao was the communauty's heavy favourite for first to die. And they were not wrong: he's tied with Suleiman for that dubious honour. He finished second three times, by dint of not being attacked by Caesar. He "won" once, by diplomacy. Although the game should have been Caesar's, that was a genuinely strong game on Joao's part. He got attacked twice as much as he attacked: standard Zulu neighbour fare. Funnily enough, people seemed to think he would be conquered by Caesar. When Shaka was the issue. Joao was almost never the target of Caesar's first war declaration (Game 4 being an exception). It may be a season 5 thing, but Joao's issue should have been very apparent: he's bordering Shaka... and he has no Copper ressource! And his single Iron is near the border, pillaging impi bait (not to mention that Joao was often the last civ to research Iron Working, which is very unusual for an AI lacking copper and having jungle to cut down). That was Joao's story, really: when Shaka came calling early, Joao would die. But if Shaka exhausted himself first against Suleiman for instance, and let Joao become stronger, then Joao would often end up conquering the Zulu instead. Another part of his story is that he can't plot at pleased, which led most of his rare military ventures to be ill-fated, across-the-map expeditions. Joao's land was good (apart from the critical lack of metal), but it was in short supply. As an Imperialistic civilization, he would fill it up fast. In my game reports, I've often mentionned his expansion patterns, because they would somewhat shape the early game: - His third city would either go up 5N 1W, on the coast, and make Shaka mad; or it would go 5E 2N, securing him horses and giving him a chance at not being Shaka's initial target. - Depending on his settling pattern, Rome would get anywhere between 0 to 3 (4 in one particular case) cities on their border. So exactly how strong Rome came out of the early game largely depended on Joao. Shaka Shaka's a warmonger. The warmonger. With Montezuma, he's a favourite amongst the AI Survivor crowds, because mayhem is his middle name. Trouble with warmongers, they tend to do poorly: they exhaust themselves into early unfruitful wars, and drop into irrelevancy. But when the conditions are right, they do spectacularly well (cf. Season 5's Alexander opening round). So where the conditions right here? The results seem to speak for themselves: Shaka's managed a successful snowball exactly once. Out of twenty games. Apart from that, he's managed a few second place finishes, but his survival rate is the second worst, after Suleiman's: he's made it alive out of 6 games only. He's declared 54 wars for 13 kills only: Caesar's been twice as effective a killer. I think that on average peace periods for Shaka were 5 turns long. Even when he was dead last in power, he would never stop to consolidate, he'd just jump back into battle. Waaagh! He's not a Zulu, he's a Space Ork. So, bad setup for him? Alright, his starting land wasn't great: his capital didn't have ressources matching his techs, and the surrounding terrain was food- and commerce-poor. But he had all the strategic ressources in his immediate vicinity: Copper, Iron, Horses, Ivory, which is all he needed, right? He did have the perfect neighbour in Joao, though: no copper, a single iron connected very late and eminently pillageable. Shaka was actually set up for a textbook warmonger snowball: conquer a militarily weak but spoils-rich Joao, then move on to Suleiman, then keep going. And he failed spectacularly. In Game 15 for instance, when Shaka attacked, all Joao had were warriors: he couldn't access his Horses, he hadn't researched Iron Working nor even Archery. Sure, Shaka conquered him in the end, but that was a painstaking conquest which took 60 turns. When watching that, my thoughts were "This guy's no Alex". I dont know why he was so bad. None of the forward-situated cities, including Lisbon, were on a hill either. I know there's a stat determining how cautious a leader is with his units, with Napoleon for instance being an extreme for not caring about the odds. Maybe that's also the issue for Shaka? He kept throwing units away when he simply didn't have the numbers (for instance attacking with 4 units when there were 5 defenders), so maybe that's part of the issue for Shaka too? At a glance, the game ran pretty much according to expectations: - Caesar won. - Joao was one the two tied leaders for first-to-die. - The number of wars (10) fell right in the middle of the two most common results (9 and 11). (Although remember that without the AP in those reruns, this is the least reliable statistics.) - The victory was an early one, as expected. - Shaka as the runner-up was more unusual, but you can't get them all, now, can you? There was one complete outlier result, though: the victory type, which was overwhelmingly likely (85%) to be Domination. And that's a strong hint that the game was actually a complete fluke. The closest I had to a scenario similar to the actual competition game was game 20: Caesar tech'd in peace for 200 turns... but then the hammer fell and he killed two opponents before fighting a last war for a Domination victory. Mansa Musa's shadow was simply never felt. So that game was an oddity: its outcome lined up with expectations, but its scenario was a complete one-off event. (A side note if you want to try and run this game yourself: there was a peak at the southwestern tip of the land which isolated a tile (with Horses). I forgot to remove that peak. I noticed it only in game 17 when I saw a barb city there. From then on, I replace the peak tile with a hill tile. It had little impact on the games: if anything it hurt Caesar a bit, since without the peak, he had one more city site available. He didn't seem to care. That said, having a reachable barb city spawn there ended up having a strong impact on the way game 19 unfolded...)
Season 4, Game 3, continued (2/3)
Game 1 Shaka's attack on Portugal backfires horribly as Germany and Rome side with Portugal. Caesar annexes most of the former Zulu lands. France rises to prominence by carving up the Ottomans before falling into oblivion because Caesar said so. Caesar honours his vanquished foe by taking up their Ottoman-carving tradition and achieves Domination. Game 2 Shaka initiates the dogpile on Suleiman, Frederick and Caesar join in. De Gaulle attacks Germany while its armies are away. Shaka and Caesar like the idea and attack De Gaulle. Caesar performs a culling at Germany's expense. Shaka and Joao fight it out for second place, Shaka ends on top as Caesar gets ready to conquer Alpha Centauri. (Note: Caesar was at 62% land area, which would have secured him a Domination win under Season 4 rules.) Game 3 Shaka and Joao fight into a stalemate. De Gaulle and Suleiman fight into a stalemate. They decide to swap fights, just in case, and it works better indeed: France conquers Portugal and Shaka cheers along as Caesar conquers Suleiman. Then the top dogs have at it while the underdogs fight for the scraps. Top dog winner: Caesar Underdog winner: Frederick Game 4 De Gaulle launches an early conquest of the Ottomans while Shaka's foray into Portugal also becomes a frank success when Caesar joins in. The Roman Church is just that: Roman. No one else likes it. Tired of the sermons, Shaka and De Gaulle head for Rome to reform it. Then they also reform Frederick. Finally, just for fun, they run an experiment by pitting Zulu hordes of riflemen and cavalries against outnumbered French tanks and infantries. Quality is a quality of its own: the French win. Game 5 Caesar has second thoughts about Deity's free settler. Barbarians provide him with peace of mind by deleting his 3rd settler. Caesar's mishap propels Suleiman into a position he's not comfortable with: game leader. His discomfort shows as he gets dismantled by the Zulus, the French, and the Romans. Especially the Romans: they've hired Joao to distract Shaka and Frederick to distract De Gaulle so that they get the Ottoman lands for themselves. De Gaulle and Caesar are best buds now. Until Caesar decides he wants De Gaulle dead, that is. And what Caesar wants, Caesar gets. Shaka kills Joao while being killed by Frederick. Caesar watches in amusement as his culture closes in on all those gaps left open. Closer. Closer. Closer. Now. Game 6 Caesar is a practiced Muslim-slayer. This time around, he decides to get to know them and their creed better, and to befriend Suleiman. Shaka launches an early attack on Portugal... and succeeds. De Gaulle launches an early attack on the Ottomans... and dies as Caesar comes to the rescue. Emboldened, Shaka attacks Suleiman, and meets the same fate as De Gaulle, under the same circumstances. Suleiman rumoured to have uttered "Et tu, Julie?" as Roman legions turn around and slaugter his troops. Frederick sets up a good-luck charms business as the world bows to Roman domination. Game 7 Caesar conquers the Ottomans while Shaka cheers him up and Joao pokes him. Caesar's tired of the poking and conquers Portugal. De Gaulle is in the process of conquering Germany when Shaka attacks him. De Gaulle is in the process of conquering Zululand when Caesar conquers him. Game 8 Shaka exhausts himself in pointless war, with nothing but a perfect kill snipe to show for it. Then he goes for one war too many and Portugal gets a sizable territorial upgrade. Meanwhile, Caesar (mainly) and De Gaulle (a little) conquer the Ottomans. Then De Gaulle is off to Germany and is offed by Caesar and Joao. The two of them stay best buds but Caesar's bored: he builds a spaceship to find lands to conquer. Game 9 Shaka takes everyone by surprise when he invades Germany. De Gaulle thinks it's a jolly good idea and the two of them carve up Germany. Joao has squeezed Rome pretty hard and Caesar needs room. Suleiman's will do. All that cosying up has established a solid relationship between Caesar and Joao, so when Shaka comes knocking at Portugal, Romans knock him down. De Gaulle has failed to take heed, so when he has at Joao, Caesar beats him up too. Every relationship has its ups and down, like when Caesar "borrows"" a few Portuguese cities for his "Domination" project. Game 10 Shaka, Caesar, and De Gaulle keep Suleiman entertained. Joao wants to visit France, the Zulus and French come visiting instead. It gets really confusing as the Germans and Ottomans visit Zululand while Caesar goes sight-seeing in Germany. Things get simpler as Joao, Shaka, and Frederick stomp off. Suleiman takes a liking to the invading stuff and heads to France. Caesar teaches him a bitter lesson about how it's done. |

![[Image: survivor5-11-22.jpg]](http://www.sullla.com/Civ4/Survivor/S5/survivor5-11-22.jpg)
![[Image: survivor5-11-23.jpg]](http://www.sullla.com/Civ4/Survivor/S5/survivor5-11-23.jpg)
![[Image: survivor5-11-24.jpg]](http://www.sullla.com/Civ4/Survivor/S5/survivor5-11-24.jpg)
.