Civ4 AI Survivor Season 3: Game Three Writeup

Game Three featured a more aggressive collection of AI leaders than our previous match. Nearly all of these leaders had low peace weights and a preference for engaging in war, with Genghis Khan Temujin leading the pack. The Mongol leader generally does nothing but build units and research military techs, making him a destructive wildcard in these games. The heavy favorite from the community was Mao Zedong, due to a combination of his excellent performance in past years and the large amount of open space near his starting position. Suryavarman was the other leader in this game with a past history of success, however his more cramped location and proximity to Temujin's insanity made him a less attractive choice. Hannibal and Peter both had good positions in this game, and the biggest thing holding them back was their history of failure in past years. Hannibal in particular had let me down twice in the first two years of the competition, and I didn't want to get fooled into picking him again. Finally, the one leader who seemed to have no chance to win was Sitting Bull. We've derisively referred to him as "Sitting Duck" in the past, as his AI personality seems determined to make a nuisance of himself without doing much to win. Sitting Bull makes tons of demands of other leaders and loves to poison enemy cities with spies. This causes everyone to hate him, and with his poor leader traits he doesn't have much to fall back on. Even worse, his high peace weight made him a pariah in this game where all of the other leaders had low peace weight scores. Almost 60% of the predictions had Sitting Bull as first to die, one of the highest margins we've ever seen. The community was gunning for him and rooting for his failure.

None of these leaders start with Mysticism tech, and that made it hard to predict who would found the first religions. Somewhat shockingly it was Mao who claimed the first one, establishing Judaism on Turn 19 by opening Mysticism into Meditation. Mao has the reputation of being the AI who cares the least about religion in the whole game, and it was a true surprise to see him go after his own faith. Suryavarman had been chasing his own religion and founded Buddhism a turn later (via Polytheism tech, heh). This group of AIs wasn't very interested in chasing after the third religion, and it didn't fall until Turn 47 when Hannibal established Confucianism (via Monotheism) as his own religion. This would provide the game with three major competing doctrines: Buddhism in the west, Confucianism in the north, and Judaism in the east.

Most of the initial AI settlements were placed in logical locations towards the center of the map. Temujin sent his initial extra settler southwest into the tundra though, which was a boneheaded move that at least one of the community predictions had called out ahead of time. Mao also chose to establish his second city on the coast to the northwest in a spot mostly surrounded by water. This ended up wasting a lot of the Holy City culture that he picked up from founding Judaism, and it was a questionable decision in the long run. Sitting Bull wound up putting his third city in a jungle location with so much unhealthiness that it couldn't grow at size 1. That was about what we were expecting from him.

Two leaders who were doing a good job at expanding were Hannibal and Peter. The Carthaginian leader put his initial cities in a diagonal line to the southeast, looking like a dagger stabbing towards the center of the map. With these blobby pangaea continents that we keep using for maps, expanding towards the center is inevitably a good thing. Peter was the first leader to five cities, planting all of them either north, east, or west of his capital and ignoring the tundra to his south for the moment. With an early Stonehenge build to complement them, Peter was putting himself in a strong position for the midgame. In the west, Suryavarman was also claiming as much land as possible, although he was starting to run up against the borders of his neighbors. Mao had the most room to expand, between his slightly more isolated starting position and Sitting Bull's incompetence to the north, but he was struggling out of the gate and failing to expand properly. It wasn't a great sign that Mao was last in score given that he was the most popular leader to win the game by a wide margin.

Barb cities began to pop up on the map, with a key one located directly between Hannibal and Sitting Bull. Whoever could claim that spot would most likely dominate the border region between them. Hannibal continued to settle border cities on top of everyone else, potentially starting feuds with half the leaders on the map. Soon we saw Russian forces moving up to the Carthaginian border, and it looked like Hannibal was about to be punished for his aggressive landgrabbing ways. Instead, the Russian units kept on going, with another target in mind:

Sitting Bull had drawn the short straw. This invasion didn't make a lot of sense from a strategic standpoint, but it can be explained easily enough via the AI programming. The AIs in Civ4 have a hidden peace weight score that dictates how they feel about one another (and the human player in Single Player games). The AI is predisposed to like leaders with a similar peace weight, and predisposed to dislike those with a different peace weight. In this game, all of the AI leaders had very low peace weight scores (either 0, 1, or 2) with the exception of Sitting Bull, who has a peace weight of 8. This would cause everyone else to hate him throughout the game, and Sitting Bull would only make that worse via his pointless spying actions. This was why such a larger number of the predictions had chosen the Native American leader as the first to die. Now it was causing Peter to launch the first war of the game against a civ that he didn't even border. The attack itself against Poverty Point was a miserable failure, with the Russian soliders unable to break through the heavy cultural defenses. However, this would tie up these two nations in warfare and drag them both down relative to the rest of the field. For Peter, this would end up wasting what had been a good opening. For Sitting Bull, this set him up to be devoured by other leaders down the road.

In the west, Temujin and Peter had both adopted Suryavarman's Buddhist religion, and this was improving relations amongst all three civs. A shared religion was not enough to stop the great khan from declaring war though, and he launched an attack against Suryavarman on Turn 82. Genghis Khan didn't particularly care about his "brothers and sisters in the faith". The invasion was successful enough to capture one of the Khmer border cities, but not enough to hold it, as Sury was able to take it back again after a great deal of fighting. This ended up being a largely pointless war, slowing down both sides in their respective development. It seemed to hurt the Mongols a bit more though, as Suryavarman has better economic traits and a stronger AI personality.

Peter had been beating his head against Poverty Point for a dozen or so turns with no success. If Sitting Bull could have limited the danger to the Russians alone, he might have been able to sign peace and get out of this situation without much damage taken. Instead, Hannibal joined the war against the Native Americans on Turn 88, and claimed the first city capture on Turn 100:

We had specifically identified Chaco Canyon as a terrible city when Sitting Bull planted it, a location with zero food bonuses of any kind. Not only was the Native American leader in a terrible position diplomatically, his city placements had been mediocre to terrible as well in this game. Now it was under Hannibal's control, along with the disputed barbarian city to the north. Vandal was the city directly above Chaco Canyon, with the "Swordsman (21)" peeking out from under the interface. Hannibal had taken that location while Sitting Bull was tied up in warfare with Peter. The truly bad news for Sitting Bull was visible on the left side of that screenshot: Carthaginian catapults. Peter had been wasting his time trying to attack while unable to bring down the walls of Poverty Point. Hannibal would have no such issues, however, and that spelled potential doom for the Native Americans.

Indeed, things kept going from bad to worse for Sitting Bull. After all of the fruitless Russian attempts to capture Poverty Point, the city fell to Hannibal and his catapults on Turn 108. Then, perhaps sensing weakness, Suryavarman jumped into the fray too with a war declaration on Turn 116, followed by an attack as well from Mao on Turn 121. At this point, everyone in the game other than Temujin was at war with Sitting Bull:

The power of peace weight, ladies and gentlemen. Sitting Bull may be a terrible AI leader, but the deck was badly stacked against him in this game. He was the "worst enemy" of four out of the five AI leaders, and the only reason Temujin wasn't at war with him was due to the fact that the Mongols had already launched another strike against Peter. (The AI will not willingly declare war in Civ4 if it's already in a war.) In other words, the goose was well and truly cooked at this point for Sitting Duck. The main question now was who would claim the spoils of this war.

With Peter distracted by war with Temujin and Suryavarman too far away to make serious gains, the division of the Native American lands looked like it would fall between Hannibal and Mao. The Chinese leader had been quietly building his strength over the last dozen or two turns, staying out of destructive wars and continuing to plant more cities in the open spaces of the east. He was up to eleven cities now via purely peaceful means, more than anyone else in the game save Hannibal, and the score was close only because the Carthaginians were capturing cities from Sitting Bull. Mao had been spreading his religion around and building some useful wonders, including the Great Lighthouse, Colossus, and Statue of Zeus. Hannibal and Mao were starting to open up some separation between themselves and the rest of the field in score, and we looked to be heading towards a game with two clear leaders. The table was perfectly set for Mao to use his invasion to capture several cities from an exhausted Sitting Bull and reap equal rewards as Hannibal with a minimum of effort.

Instead, Mao made a halfhearted attack against the closest Native American border city, then signed peace three turns later. This was a huge missed opportunity for the Chinese leader, as he lost the opportunity to top the board alongside Hannibal. Instead it was the Carthaginians who captured every single one of Sitting Bull's settlements and delivered the death blow on Turn 134:

Hannibal would later peacefully cede the city of Snaketown to Mao due to the presence of surrounding Chinese culture. Nonetheless, Hannibal had successfully acquired every other city from the Native Americans, taking five cities in total to reach a grand total of thirteen. Mao was the only one in striking distance with eleven cities, and the rest of the AI leaders were significantly further behind with seven or eight cities apiece. Mao had been the favorite pick to win the game because many of us expected him to reap the biggest windfall from Sitting Bull's near-certain demise. Instead it was Hannibal who had been the big beneficiary of Sitting Bull's collapse. The game now entered a state with Hannibal in first place, Mao running a close second, and the other three AI leaders trailing a clear distance behind.

While the Native Americans had been falling before the Carthaginian onslaught, the other war in the west had been grinding on between Peter and Temujin. The Mongols successfully managed to capture the iceball tundra city of Yekaterinburg while the Russian armies were engaged off in the east, and they were ultimately able to hold onto that city after a great deal of bloodshed on each side. A treaty was finally signed on Turn 143, bringing the whole world back to a peaceful state for the first time in many centuries. That lasted for exactly half a dozen turns before Suryavarman declared war on Temujin, sniping the Mongol border city of Ning-Hsia in the process. That was the only significant action of this short war, which was concluded with a new peace treaty after eight turns. The main effect of all this warring had been to cripple the long term situation of the Mongols. Temujin had attacked both of his neighbors by now, and this ruled out any long term cooperation with either Suryavarman or Peter. Both of their nations hated his guts at this point. Furthermore, the Mongols had been at war so many times that their economic situation was simply disastrous. Later on, we checked the science rate of the respective civs on Turn 200, and found that Temujin was making 80 beakers/turn at a point where the game leaders were up near 1000 beakers/turn. Needless to say, this boded poorly for the future survival of Temujin.

We were watching the diplomatic situation closely during this interval of peace to search for where the next major war would begin. One noteworthy development had been Peter founding Christianity and then swapping to his own native religion, breaking the bonds of shared Jewish faith that he had enjoyed with Mao. We noticed that Mao was building up his units for some kind of an attack, and Peter seemed to be the most likely target. If Mao could hit Peter and take over his territory, that would put the Chinese favorite into a position of clear first place. Instead, Mao sent his units further afield:

His target was... Suryavarman?! That made no strategic sense at all, and even going by diplomatic relations, Peter would have been just as likely of a choice as Suryavarman. This was one of those places where the arbitrary nature of the AI programming reared its head, with Mao losing his mind and embarking on a truly ridiculous war. His attack against Angkor Wat was a failure, with the whole stack sitting around waiting for the two catapults to reduce the defenses at 4% each turn, then nothing happening until Mao signed peace ten turns later. I think the city was still sitting at 70% defenses or something like that. What a waste of time. This foolish move poisoned the well diplomatically with the Khmer and ensured that Suryavarman would never work with the Chinese again. This was the second time that Mao had blundered badly when it came to a war declaration, and even though he was in clear second place, this had been a rather poor showing of missed opportunities.

More consequentially, Suryavarman had declared war on Temujin on Turn 172, followed by Peter also declaring war on Temujin six turns later. This was their respective wars of revenge against the Mongol oppressor, and the khan's backwards forces did not handle it well. Temujin was pinned between the blue forces of Suryavarman in the north and the red forces of Peter in the east, two hinges of a trap swinging shut together. The Khmer took two core cities in the north while the Russians captured four low quality tundra iceball settlements in the deep south. This likely would have been the end of Mongolia altogether, except that the two leaders signed peace simultaneously on Turn 193. It looked like a forced peace from the Apostolic Palace, and therefore Temujin lived to fight another day. With only four cities remaining, however, he was finished as a force in this game.

Hannibal had been quietly teching along while all this was going on. As the other four civs squabbled amongst themselves with largely fruitless wars, he was developing his economy with his Financial trait and building several key Renaissance era wonders. Hannibal landed the trifecta of Liberalism, Taj Mahal, and the Statue of Liberty, while also researching to Rifling tech and setting up pictures like this:

That many cavalry were downright terrifying. Hannibal had enough power to wipe out anyone else on the map, and the big question was whether he would choose to unleash those cavs in a storm of flame and fury. Unlike most of this game's aggressive leaders, Hannibal will not declare war at "Pleased" relations, and that was keeping most of the other nations safe for the moment. Then suddenly there was a brief moment where one of the other leaders had their relations drop to only "Cautious" and Hannibal struck:

Against Mao. The Chinese leader was caught completely unprepared for this move and was in no position to resist. Mao had researched all the way to Democracy and Scientific Method techs, yet had somehow failed to research Rifling. That meant it was Carthaginian rifles, cavs, and cannons against a medieval army on the part of China, and long experience with Civ4 has proven that this kind of discrepancy in military tech always turns into a one sided rout. Making matters worse, Peter piled into the war on the side of Hannibal two turns later, and the feeding frenzy was on. Over the course of the next two dozen turns, China completely collapsed before the attack of these two rivals. Hannibal was gobbling up most of the territory, but Peter claimed three cities in the south that made this war more than worth his time. Mao fell from the clear second place civ down to the bottom of the scoreboard faster than we could blink. It was an ignominious showing from a leader who had been in such a good position only a short time earlier.

The truly crazy thing is that it was Mao's foolish attack against Suryavarman that had opened up this strike in the first place. Hannibal had picked up a -2 modifier with Mao as a result of "you have declared war on our friends" from that little conflict, and that was enough to push their relations down to the "Cautious" level. Mao could have gone after Peter instead, which would have put him on the positive side of a two-front war when combined together with the Mongol/Russia war raging at the time. Instead, Mao found himself cornered and his cities rapidly taken over. His capital was gone by Turn 233, and Hannibal landed the finishing blow on Turn 241:

Hannibal had now wiped out two competing civs and looked certain to be the victor in this game. He had essentially done what we thought Mao would do: take out Sitting Bull and then snowball from that position into taking over the rest of the continent. Hannibal had an absolutely massive lead in technology at this point, more than a full era ahead of Peter and Sury, and he seemed certain to win a Spaceship or Domination victory from this point. As for Peter, he had also done very well for himself and was now the clear runner-up. Playing Robin to Hannibal's Batman had been highly successful in this war. He was about 800 points ahead of Suryavarman and that gap seemed too wide for the Khmer leader to make up.

Suryavarman was doing his best to try though. Temujin had failed his latest sanity check and redeclared war against the Khmer a little bit earlier on Turn 216. That was going about as well as expected, with the cuirassiers of Suryavaryman fighting against the entirely medieval army of Temujin. The Mongols didn't even have Engineering tech for castles at the same time that the Khmer were about to finish their research into Rifling! The Mongol capital of Karakorum fell on Turn 227, and after an eternity spent waiting for trebuchets to bombard the defenses out of the last few remaining Mongol cities, Suryavarman sealed the deal on Turn 260:

Across the three years of AI Survivor, Temujin has now been eliminated in three out of the four games where he's appeared. The khan is simply too aggressive for his own good, picking fights he can't win with anyone and everyone. I'm sure that he'll win one of these games eventually if we keep running them - heck, Shaka and Tokugawa finished #1 and #2 in one of the games last year - but he's definitely not one of the stronger AI personalities. As for the rest of the field, it appeared as though we were unlikely to have any more major wars. Hannibal was "Pleased" with the remaining two leaders and won't declare war at that level of relations, while the other two would have to be suicidal to start any kind of trouble with the Carthaginians. It was theoretically possible for Peter or Suryavarman to start some trouble brewing, particularly with one another, but neither seemed to be plotting a war by changing their city builds over to all units. With Peter now a thousand points ahead in score, this one looked to be finished.

Then again, the AIs can do some weird things sometimes. Note the research time remaining for Hannibal's investigation into Plastics: 47 turns. He had stopped researching new technologies and decided to pursue a Cultural victory of all things. That made absolutely no sense since Hannibal already had a lock on the Spaceship victory condition, and he didn't have any cities even remotely close to the 50k culture required. Nevertheless, that was Hannibal's decision and he set the culture slider to 100% and kept it locked there for dozens of turns on end. Over time, this allowed the other two civs to come close to matching Hannibal in technology, although when Peter was on the verge of reaching tech parity, Hannibal stopped running the culture slider for a few turns and zoomed up to Robotics tech (and mechanized infantry) for safety, then went back to cranking out the culture again. It also had a curious effect on the scores: Suryavarman began slowly closing the gap with Peter. The Russian cities were being crushed under Carthaginian culture, losing tiles and then losing population as they were swallowed up by Hannibal's borders. Meanwhile the Khmer conquests in former Mongolia were growing in population and popping borders, grabbing more score points. The gap between the two fell to 600 points, the 300 points, then close to 100 points. Was Suryavarman actually going to make this work?

In the end, the Khmer leader came up just short:

Hannibal took home the victory and Peter scored the runner-up spot. But look at how close the scores had become in that race for second place! There were only 119 points separating Peter from Suryavarman, after there had been a gap of more than a thousand points at the conclusion of the Mongol conquest. If Peter had picked up one fewer city from Mao, then it would have been Sury moving on to the playoffs and Peter going to the Wildcard game. This was one of the few times where the bar graph tracking score was significant enough to be worth posting:

For a game with no fighting over the last sixty turns, the ending wound up being a lot more exciting than we expected. That was all she wrote for Game Three, with the biggest favorite to win of any game in the competition thus far getting eliminated in humiliating fashion. It was proof that past history is no guarantee of future performance, and when an AI consistently makes poor strategic decisions, they're likely to be sent packing. As for Hannibal, he finally lived up to the expectations that he had failed to meet in past seasons. He'll have a chance to keep the ball rolling in an upcoming playoff game in the next round. Until then, thanks as always for reading and following the competition.