Civ4 AI Survivor Season 7: Game Five Alternate Histories


Game Five Alternate Histories Spreadsheet

One of the recurring features of past seasons of AI Survivor have been our "alternate histories", running additional iterations on the same maps to see if the same events would play out again. Game Five saw Kublai Khan stomp all over the rest of the field en route to an impressive domination victory. Was that something which would unfold in each game? This was a topic that called for more investigation with alternate history scenarios. Following the conclusion of previous seasons of AI Survivor, I had gone back and investigated some of the completed games and found that they tended to play out in the same patterns over and over again. While there was definitely some variation from game to game, and occasionally an unlikely outcome took place, for the most part the games were fairly predictable based on the personality of the AI leaders and the terrain of each particular map. Would we see the same patterns play out again and again on this particular map?

The original inspiration to run these alternate histories came from Wyatan. He decided to rerun the Season Four games 20 times each and publish the results. The objective in his words was twofold:

- See how random the prediction game actually is. There's a natural tendency when your predictions come true to go "See! Told you!", and on the contrary to dismiss the result as a mere fluke when things don't go the way you expected them to (pleading guilty there, Your Honour). Hopefully, with 20 iterations, we'll get a sense of how flukey the actual result was, and of how actually predictable each game was.

- Get a more accurate idea of each leader's performance. Over 6 seasons, we'll have a 75 game sample. That might seem a lot, but it's actually a very small sample, with each leader appearing 5-10 times only. With this much larger sample, we'll be able able to better gauge each leader's performance, in the specific context of each game. So if an AI is given a dud start, or really tough neighbours, it won't perform well. Which will only be an indication about the balance of that map, and not really about that AI's general performance. But conversely, by running the game 20 times, we'll get dumb luck out of the equation.

Wyatan did a fantastic job of putting together data for the Season Four games and I decided to use the same general format. This particular set of alternate histories were run by Amicalola - many thanks for spending so much time on this task! Amicalola posted the resulting data from the alternate histories and then discusses some of the findings below in more detail. Keep in mind that everything we discuss in these alternate histories is map-specific: it pertains to these leaders with these starting positions in this game. As Wyatan mentioned, an AI leader could be a powerful figure on this particular map while still being a weak leader in more general terms. Now on to the results:

Season Seven Game Five

Game One | Game Two | Game Three | Game Four | Game Five

Game Six | Game Seven | Game Eight | Game Nine | Game Ten

Game Eleven | Game Twelve | Game Thirteen | Game Fourteen | Game Fifteen

Game Sixteen | Game Seventeen | Game Eighteen | Game Nineteen | Game Twenty

(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.)

Amicalola: The game we saw was a semi-common outcome, and certainly not an outlier, but also wasn't the most common scenario that played out. Kublai Khan did win five more games, all dominantly, and there was clearly a plausible scenario where he crushed 1-3 rivals and ran away with the game. However, if KK failed to get that early conquest, he tended to fizzle out and be crushed by the two more dominant leaders, who turned out to be the surprising combination of Roosevelt and Churchill. To my great surprise, it turned out that Roosevelt had a massively weak game on livestream compared to what was expected. He had a beautiful, luscious start, and between his first two cities he would usually claim a wet corn, cows/pigs/sheep, horses, and two silvers. His third city would then also go on to claim gold, rice, and copper. Because he had such extreme resource clusters at these early cities, Roosevelt's tech path was near-identical in basically every game: he would start Animal Husbandry (while improving his wet corn), then tech Mining/Wheel in either order (while pasturing all over the place), and then up to Bronze Working while he mined and connected his gold and silvers. It was a disgusting opening that saw Roosevelt miles ahead of the pack in every demographic from T50-100, and his strength created the dynamic for this game: could anyone else snowball out of control before Roosevelt won by default? This was enough for Roosevelt to win seven games, and experience several near misses, such as Games 6 (where Roosevelt lost due to a silly Cyrus war declaration) and 18 (where Roosevelt was killing Churchill too slowly to stop the spaceship).

Roosevelt's opening was disgustingly strong compared to the other leaders.

The diplomacy of the map also lent itself shockingly well to Roosevelt, as the 3vs3 dynamic that we'd been expecting peaceweight-wise did not play out so neatly. Roosevelt's most common attack target was Boudica, who was extremely weak. She tended to lose the religious race to Asoka, but would then acquire the other religious tech and Archery before getting any workers techs (usually Animal Husbandry first). This often left Boudica without Mining or The Wheel by T50, and frequently without BRONZE WORKING by T100. In Game 3, Boudica didn't have The Wheel or Bronze Working by T125, which might be the worst I've ever seen the AI crash its economy. If Roosevelt attacked before T90, which happened a decent amount thanks to the copper he always connected early, he could run over Boudica before she ever connected her iron resource (Boudica lacked copper); see Games 1 and 3 for the outcome of that. But because of his silver resources, Roosevelt also tended to found the third religion, which put him at odds with Asoka. They clashed repeatedly here, which was responsible for Roosevelt's worse games as it tended to weaken both parties (I suspect this is what went wrong for Roose in the livestream). This chaos played out all over the map though, as these leaders refused to play along the lines that peaceweight ought to have dictated. Churchill, for example, fought most often with Kublai Khan (as expected), but also attacked Asoka a lot of the time, and only fought early wars with Cyrus in 3-4 games, as they tended to share a religion. Kublai Khan also got stuck in wars with Boudica and Cyrus many times, which was responsible for his feast-or-famine outcomes - too often Kublai would just get dogpiled out of the game, even from initial positions of strength like in Game 4.

The leader who played the most impressive games here though, and I truly hate to say this, was actually Churchill. Other than a dry stretch during the first five games, Churchill was arguably the most consistent leader in Game Five, and was catching up to Roosevelt until the last two games locked in the overall American victory. Unlike Roosevelt (who was essentially gifted the 'default' victory by his disgusting start, and did nothing impressive in his actual play) and Kublai Khan (who could settle the most land and run over a neighbour if he wasn't dogpiled), Churchill had very weak land, forced to settle a jungle-bound second city, and not much of it, usually ending up with only 6-7 cities after the landgrab phase. There was absolutely nothing about his start that suggested he would be a major power, and indeed he was usually at the bottom of the scoreboard by T100. Churchill played his position beautifully though - he basically never attacked anyone before T150 - and Cyrus tended to rush past him into Asoka, or into Kublai Khan to the east. Churchill would usually rush Banking tech for his Stock Exchanges, and then move straight into Rifling for Redcoats. From that point, he could (and did) murder Asoka, Cyrus, Kublai, or Boudica with ease, as only Roosevelt seemed safe from English attack in the midgame (it probably wasn't a coincidence that these two were the strongest leaders). Even Roosevelt wasn't safe in the lategame, as once Churchill did have a strong army, he tended to push the advantage, and ended several American spaceship near-wins. Unlike we often see from Churchill, there were few backdoor victories or seconds, and his skillful use of diplomacy and ability to only war when conclusive was enough for him to emerge the second-strongest leader from probably the weakest start. Amazingly, he was catching up to Roosevelt until the end, and I half-suspect he would have actually been strongest in a larger sample.

As a final point, these games tended to have multiple competitors until very late, usually at least T300, and who won was usually decided by climactic wars, often to the death. It was rare for any of Churchill, Cyrus, Kublai Khan, and Roosevelt (the 'main' competitors) to ever be knocked out before T200 - this happened five times across all four leaders combined, three of which were Kublai - as they repeatedly feasted on the corpses of Boudica and Asoka before turning on each other like vultures. Because of this, it was a series of very slow games, with an average finishing date of around T360, which is the highest average I've seen so far (including Game One). This included Game 20, which ended at the astonishing date of T434 (the latest date ever recorded?) via a Roosevelt CULTURAL victory, while he and Churchill were both somehow still over a dozen turns from launching the spaceship. The games were so long because there were often lengthy wars of attrition between the strong leaders, and also because there were no strong techers in this group (it was a bad sign that Roosevelt or Churchill usually held the tech lead...). It's pretty remarkable that these leaders managed to research even slower than Shaka, Hammurabi, and Isabella from Game One.

With that said, here is a look at the individual leaders' performances:

Roosevelt of America
Wars Declared: 38
Wars Declared Upon: 37
Survival Percentage: 65%
Finishes: 7 Firsts, 5 Seconds (45 points)
Kills: 16
Overall Score: 61 points

Roosevelt graded out as the strongest leader here, although in a larger sample I'm honestly not sure if he or Churchill would have ended up on top. Roosevelt's strength was due almost entirely to a combination of having a disgustingly fertile start relative to the rest of the field; his demos were consistently disgusting coming out of the landgrab phase, and particularly around T50 when his initial three cities were up and running. Indeed, Roosevelt's other cities were usually terrible, but with these initial three he could get them out on the map faster than anyone else, and that was usually enough to be successful. Additionally, thanks to the copper at his capital (and Boudica's historically low power pre-T100), Roosevelt was also quite aggressive in these games, declaring more wars than he faced (the gap was bigger until the final 5 games, where Roosevelt suddenly found himself getting attacked more often). He was able to score the second-most kills at 16, and at least half of these came from rushing Boudica before she connected metals. However, Roosevelt was also capable of partitioning Asoka with Kublai or Churchill, or of eating Kublai Khan and Cyrus in the mid to late-game. Roosevelt is a very peaceful leader, and I think this really demonstrates the position of strength that he occupied here, as he averaged nearly two war declarations per game - comparable to Cyrus, and more than Boudica! On the other hand, things definitely had the potential to go wrong for Roosevelt too. Although he was only first to die once (at the bizarrely late date of T311, having been leading for most of the game), he did fully die in 7 games, and was close in a few others. Because he occupied the position as 'default' winner, it was rare for Roosevelt to fall apart without something major happening. Sometimes, he would get dogpiled by a few leaders at once, as in Games 7 or 13. At other times, he fell prey to the classic 'good leader, bad techer' conundrum we saw from Hammurabi in Game One - he would kill Boudica, and then sit on his hands while another leader snowballed into a victory faster than Roosevelt could. This was most common with Churchill, but also happened with Cyrus and Kublai Khan. Overall though, this was a very dominant performance. Roosevelt might have been semi-gifted these wins through his disgusting starting position, but he was solid enough to not mess around with religion or military techs before improving his resources, and sometimes that's all it takes.

Churchill of England
Wars Declared: 38
Wars Declared Upon: 21
Survival Percentage: 65%
Finishes: 4 Firsts, 5 Seconds (30 points)
Kills: 17
Overall Score: 47 points

Churchill played an extraordinarily solid series of games here - far more impressive than anything I've seen him do in the past - and it turned out his second place was far from an unlikely outlier result! This was all particularly impressive considering he had arguably the worst start on the map (competing with Cyrus, who also put up a strong performance), as Churchill probably played the best of these leaders over the course of the 20 games. Indeed, after a slow start he was catching up to Roosevelt quickly, with the overall victor only decided in the final two games. Because of his slow start, Churchill had to play a calculated, efficient game, and that's exactly what he did. He expanded terribly slowly, with his second and third cities being jungle-buried and his capital also being quite weak. He usually only got out to 6-7 cities, and had to bide his time. Almost always, the other five leaders would descend into chaos while Churchill peacefully teched up to Rifling and Banking, and even when he was attacked (usually by Kublai Khan), he was usually able to force a quick peace. Once Churchill did unlock Stock Exchanges and Redcoats, he was unleashed and could often wipe multiple opponents off the map. He scored the most kills, despite being involved in the second-least wars, as Churchill could not be stopped once he got started. Many of his wars were also won from tech parity, as Churchill put his CHA/PRO traits to use and created Super Redcoats (which, I have to admit, was pretty cool). It was the first time I've seen them become relevant, and they were consistent enough here to be far from a fluke. Churchill did have some bad games, usually when he got into a fight too early. In Game 20, Cyrus ground him down, while in Games 1 and 10 Kublai Khan did the same. Churchill's abhorrently weak status in the first 150 turns could be seen from many of his deaths only featuring 1-2 wars total - he just couldn't compete early on. Overall though, this was a shockingly consistent set of performances (with several of his second-place finishes also being near-firsts), and Churchill impressed me more than I'd have thought possible before this game. It is clear he isn't a top-tier leader, but he's also clearly a cut above the cannon fodder down below - I certainly won't underestimate Churchill after the last two years, that's for sure.

Cyrus of Persia
Wars Declared: 40
Wars Declared Upon: 16
Survival Percentage: 65%
Finishes: 3 Firsts, 5 Seconds (25 points)
Kills: 14
Overall Score: 39 points

Although Cyrus and Kublai Khan scored the same number of points, I've put Cyrus ahead of Kublai because he was more consistent overall throughout these games. Cyrus was first to die on livestream, and these games showed that to be an extremely unlikely occurrence. He only repeated it once in this set of 20 games, while advancing nearly half of the time. This was double impressive considering that, like Churchill, Cyrus had a terrible starting position, cramped on space and without any luxuries or rivers at all. Compared to Roosevelt's double-metal river valley, these leaders were nearly playing different games! Cyrus put his Charismatic/Imperialistic traits to excellent effect here, keeping up in expansion despite the crappy land, and in city size despite the absence of luxury resources. It was clear watching these games that while he couldn't use the traits to snowball to victory, they kept him in the ballpark, and a leader like Roosevelt would likely have been totally screwed in the Persian position. Bizarrely enough, it seemed that the key to a Persian victory was to hit Kublai Khan early, just as Cyrus did on livestream. This was a high risk, high reward move: sometimes, Cyrus could take Kublai's capital pre-T100 and run away with the game (Game 11), and sometimes it would backfire and Kublai Khan would stomp all over Cyrus' hopes and dreams (Game 2 and Real Game Five). However, the more average Cyrus game saw him be pretty peaceful, opting to keep up in tech rather than land. This was, in a way, also high risk/reward; Cyrus was able to score five second-places by doing this, but also died to stronger leaders in the lategame quite often, particularly to Roosevelt who he was quite fond of attacking in the later turns. In general, Cyrus was adept at making friends on this map; he was attacked only 16 times (less than once per game!), and this meant that he was relevant even when he could not claim additional land. Cyrus' bad games otherwise came when his farmer's gambits were called, as his production was terrible and he could not usually fight 1v1 wars at a technological parity. This is what happened in Real Game Five against Kublai Khan, and it also happened against Roosevelt and Churchill sometimes, such as Game 9. He really relied on his neutral-evil peaceweight to make friends, and when that failed, he was pretty much dead. Overall though, this was a series of games much akin to Churchill, in that it was a seriously impressive performance from a leader in arguably the worst starting position. Cyrus has been a bit shafted for several years in a row now, and though he has struggled recently, these performances suggest he should not be discounted when he is given a bit more to work with.

Kublai Khan of Mongolia
Wars Declared: 46
Wars Declared Upon: 43
Survival Percentage: 40%
Finishes: 5 Firsts, 1 Second (27 points)
Kills: 12
Overall Score: 39 points

Kublai Khan was the winner of Real Game Five, and he backed that up with five more victories across these Alternate History games - it not often you can say that about the equal third-placed leader! I've listed Kublai Khan below Cyrus in this list, because even though they scored the same amount (and Kublai won nearly twice as much), the Khan was a lot less consistent in his performances. Kublai had a smattering of dominant games where he crushed the competition, such as Games 2 and 16. These games tended to see Kublai running over Asoka, Cyrus, or Churchill first, and then inexorably consuming the rest of the map as he reached a critical mass of cities. In particular, Roosevelt would often maintain a score and tech lead/parity with Kublai during these games, but eventually die to weight of numbers as Kublai would have twice or thrice as many production queues. However, what was far more common was for Kublai to perform well in the expansion phase, as we saw on livestream, but run into multiple opponents at once and exhaust himself. Kublai fought A LOT in these games, at 89 total wars (Roosevelt was next-highest at 75, and everyone else was barely above 60). This was most apparent during Game 4, where Kublai was a juggernaut but was repeatedly attacked by literally everyone except Asoka, until he was eventually brought down. It happened in other games too though, as Kublai found himself the surprising diplomatic pariah over and over again; this was evident in Kublai's offensive and defensive wars, where much like Boudica he was roughly equal in both, despite being a 'warmonger' leader. Churchill was the most likely leader to attack him, which happened in most games at some point, and was usually the downfall of the Khan. Cyrus and Boudica also had a decent chance to attack Kublai though, particularly Cyrus who sabotaged Kublai's chances far more successfully in these games than his failed attempt on livestream. Asoka was the only leader unlikely to join the dogpile, and that didn't matter because the warring usually kicked off with Kublai attacking him! Even when Kublai did find success in his constant warring, he often fell behind in tech, and this doomed him several times as well, such as in Games 10 and 20 (both to Roosevelt). Overall, Kublai was a strong leader with a good chance to win, but he also did very badly in a lot of games, with very little room in the middle. A bit surprising for a leader often thought of as the King of Second Place, but quite fitting for the central position he occupied.

Boudica of the Celts
Wars Declared: 31
Wars Declared Upon: 32
Survival Percentage: 25%
Finishes: 0 Firsts, 3 Seconds (6 points)
Kills: 4
Overall Score: 10 points

Boudica was honestly very lucky to score more points than Asoka, and clearly played the worst out of these leaders. Her economic prowess was beyond bad, to the point where sometimes it seemed like Boudica had suffered from a catastrophic bug in her programming. Boudica starts with Hunting/Mysticism techs, which improved zero of her entirely Animal Husbandry-based food resources. Boudica always opened with a religion tech... but so did Asoka, and he always won the tiebreaker. So it wasn't uncommon for Boudica to end up researching two religion techs out of the gate, unable to improve any resources! But things got much, much worse during Boudica's descent into the Bankruptcy Hall of Fame. While she usually managed to get Animal Husbandry out of the way next, Boudica always slotted in Archery (military flavour's curse!) and a surprising Masonry (for her Unique Building presumably, since she had no quarries) too. All of this usually took Boudica until about T60-75, by which time she still lacked Agriculture, The Wheel, and Mining (let alone Bronze Working or Pottery!) techs, and her research rate started to crawl. It turned out that Boudica crashing her economy was actually the norm in these games, as she struggled to connect her iron until around T100-120, and was usually crippled either way. The main consequence of this was that it allowed Roosevelt to kill Boudica multiple times basically for free, which was responsible for his wins in Games 1, 3, 10, and 12 (as well as many strong seconds). Even when Boudica escaped her early deaths, she failed to impress. She attacked basically everyone, and was incapable of winning nearly any of those wars. Her only success came from three very distant seconds, where she tagged along behind another low peaceweight leader, and had the fortune not to be attacked by them. Perhaps the biggest sign of her weakness was that despite being one of the most aggressive leaders in the game, and in a context where her main enemy was Roosevelt(!), Boudica somehow had more defensive than aggressive wars. This was arguably the worst economic performance AI Survivor has ever seen, and it is worth noting that Boudica might be the warmonger most hampered by a commerce-poor start; her preferences are too screwed up by religion, military, and the Celtic Dun to ever truly recover.

Asoka of India
Wars Declared: 10
Wars Declared Upon: 54
Survival Percentage: 20%
Finishes: 1 First, 1 Second (7 points)
Kills: 1
Overall Score: 8 points

Asoka was very weak in these games, but it was because of being diplomatically screwed, rather than necessarily playing his start poorly, which was more understandable than whatever the hell Boudica was doing. Asoka typically found himself expanding competently and winning a religious race out of the gate, landing around the top or middle of the scoreboard in the early game. However, he suffered from a lack of strong land to snowball his early advantage from, and he also didn't have the immediate core of uber-broken cities that Roosevelt benefited from. This meant that Asoka could sometimes stalemate 1v1 wars that came in the early game, but couldn't fight them in the late game, and also could be dogpiled and eliminated right out of the gate. This happened a lot, with Roosevelt, Churchill, Cyrus, and Kublai Khan all at least semi-likely to war with Boudica (especially Kublai, who attacked in nearly every game). It was extremely rare for Asoka not to have at least two enemies coming out of the landgrab, and three or four were also highly plausible. This left the Indian leader in a no-win situation, as his main role was to spread a religion around (that someone else inevitably conquered) and dictate the diplomacy from the grave. Asoka did have two successful games, although both were outlier results. In Game 9, Asoka came first after an unlikely combined rush of Boudica with Roosevelt, while no one else attacked him. Then in Game 10, he stayed alive while Roosevelt killed literally everyone else. Otherwise though, Asoka was a total failure on this map, who couldn't compete with the bid dogs economically or militarily. It was a deserved loss here, and there isn't much more to say about it.


On the one hand, I would have loved to run even more sets of Game Five, to see if Roosevelt or Churchill was truly the strongest leader on this map; it might not look too close, but that really felt like a coincidence of the final two games. On the other hand, this set of games showed pretty clearly that the result we experienced on livestream was quite plausible, with Kublai being the leader second-most likely to win the game (if not second-strongest overall). Churchill advancing in second was also far from an outlier result, as the English leader consistently managed to exceed expectations. However, Real Game Five drastically underestimated the strength of Roosevelt, who advanced more than half the time and was a dominant force in nearly every game, and it was almost a shame we couldn't get another crazy result in a season of upsets. For Fantasy purposes, Henrik seems to have bid very well on this game, getting Churchill for a bargain, and Boudica (who was quite likely to be first to die). Myth was truly robbed with Roosevelt though, who realistically should have scored at least a few points. Similarly, Jmie was a bit unlucky for Cyrus not to advance, although at least he scored some first-to-die points. The real result was still relatively reflective of these games overall though (as compared to some of the other bizarre outcomes we've experienced!), and it was nice to see a semi-normal result at last, even if it was a shame Roosevelt couldn't finally get it done. He'll have to prove the haters wrong next year, I guess.

Cheers ~ Amicalola

Postscript: Slashin' ran an additional 20 alternate histories after Amicalola finished with the following results:

Clearly the finishing order changed around a bit but the broad trends of the game appear to have remained the same. Slashin' seemed to get fewer of the results where Kublai managed to snowball to a victory and, as Amicalola noticed, whenever that failed to happen Kublai perfored very poorly. Asoka similarly seems to have had better diplomatic luck in this series of games and was therefore able to translate more of them into successful victories. Thanks for providing us with some additional data to consider for this scenario!