Civ4 AI Survivor Season 5: Playoff Game Three Alternate Histories


Playoff Game Three 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. Playoff Game Three had a tense clash between Gilgamesh and Huayna Capac that narrowly went in the Sumerian leader's favor. 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 Seasons Three and Four of AI Survivor, I had gone back and investigated some of the previous 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 specific 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 5 seasons, we'll have a 60+ games 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. First I'll post the resulting data and then discuss some of the findings 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 Five Playoff Game Three

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.)

The match that we watched in Playoff Game Three turned out to be highly representative of the alternate history scenarios. Gilgamesh was the most common winner and Mehmed was the most frequent runner up. Speaking more broadly, there were three leaders that repeatedly had strong performances on this map and three leaders that were total failures. Gilgamesh, Mehmed, and Huayna Capac were the three leaders that survived to the finish in our game and these were the three leaders that contended for the victory again and again in the alternate histories. There were only two unusual aspects of the game that we watched separating it from the typical course of events on this map. The first and most obvious was the fact that Gandhi was not the first leader eliminated and lasted until Turn 184. This was highly unusual as Gandhi was First to Die in fully 80% of the alternate history scenarios and rarely made it much past Turn 100. The community was absolutely correct here and no, we weren't crazy for everyone tapping Gandhi as the overwhelming favorite for a quick exit. The other diverging point was Mehmed ending up much stronger in most of these scenarios, due to a combination of Gandhi dying faster and the Ottoman leader not having the bad luck to see four barbarian cities pop up on his borders. Nonetheless, the playoff game that we watched was broadly representative of the way that this setup usually unfolded and it was not in any way a fluke result.

The key to understanding this map was a combination of three factors: an unusually strong starting position for Gilgamesh, weak overall positions for the other two northern leaders in Hannibal/Peter, and a completely doomed Gandhi who had zero chance at survival due to his peace weight situation. Back when I was rolling this map, I noticed right away that one of the six starting positions was ludicrously strong with half a dozen floodplains tiles, a plains cow, and THREE gold resources. Due to random luck this spot ended up getting assigned to Gilgamesh:

This is the kind of setup that the people who compete in the CivFanatics Hall of Fame will re-roll maps a hundred times until they manage to find. It's at the absolute upper limit of what a starting position can contain without outright cheating by using the Worldbuilder. I did not edit this start in any way but the strength of the position was unmistakeable and it always allowed Gilgamesh to get off and rolling with a flying start. On Deity difficulty, the AI leaders can build settlers pretty fast and then tend to be most limited in the early game by a lack of commerce and a lack of happiness resources. Gilgamesh was able to solve both of these issues with his monster capital, with the gold resources pouring in 8 commerce apiece along with bonus happiness so that Sumerian cities could grow to larger sizes. This advantage was then compounded as Gilgamesh sent his extra Deity starting settler to the southwest to the corn/ivory cluster of resources. He sent his settler there in every single game, landing another strong location with a second early game happiness resource. Add in the Creative trait for free border pops and Gilgamesh was easily atop the scoreboard coming out of the landgrab phase in virtually every match.

The early strength of Sumeria then fed into the second factor: weak northern neighbors. I did not expect that Gilgamesh would always send his starting settler to the west in every game, a move which had the effect of squeezing Hannibal badly on land over and over again. We watched this take place in the actual playoff game and it was repeatedly in nearly all of the alternate histories. Hannibal did have a copper resource but he was stuck between the two strongest AI leaders on the map (Gilgamesh and Mehmed) which didn't leave much room to expand. As a result, Hannibal was eliminated in most games and achieved little outside of a ridiculously flukey victory in Game #6. Peter was even weaker although with less reason to blame the map. He was simply ineffective as an AI leader, directing his expansion into the tundra and picking fights with far stronger opponents. Peter scored exactly one kill in 20 matches and never managed a top two finish while being eliminated in 90% of these games. He wasn't exactly putting much pressure on Gilgamesh either.

The other major factor in how this game unfolded was Gandhi's dismally bad diplomatic situation. The poor fellow was the "worst enemy" of all five other leaders out of the gate in every game and this meant disaster without fail. In alternate universes where Mehmed was stronger, Gandhi was bowled over repeatedly by the Ottomans or the Sumerians or some combination of everyone at once. Not only was Gandhi First to Die in 80% of these games, fully 16 times out of 20, he was eliminated *SEVEN* different times before Turn 100! The Turn 78 elimination in Game #12 would be the fastest ever in AI Survivor history had it taken place in a real match. Gandhi never survived to the finish and he declared zero offensive wars against 59 invasions. Sheesh. So yeah, the picking contest was completely correct to make him an overwhelming favorite as First to Die. The fact that it took so long for Gandhi to be knocked out in the playoff game was a weird outcome that rarely repeated itself.

The map therefore had all the elements needed for a Gilgamesh victory. He could snowball off Hannibal as we saw in the actual playoff game, or do the same thing against Peter, or take part in the mass beatdown of Gandhi. Mehmed typically became strong by conquering India and thus was a threat to win in nearly every game, although he was held back by his refusal to declare war at "Pleased" relations and also by Gilgamesh frequently doing the same snowball thing faster and better. Huayna Capac had a much harder path to victory since Gandhi crowded him badly on territory and he lacked the same easy opportunities to conquer land. Only in the rare situations where the Incans conquered Peter or got the lion's share of Gandhi's spoils did Huayna Capac become a runaway. Yet his teching and overall economic strength were so impressive that Huayna Capac still won four times anyway and came extremely close to several other wins despite having a far worse starting position than his rivals. Huayna Capac wasn't the top leader on this map but he's really, really good at the game. There's no way that Mehmed manages four wins from the Incan starting position if we were to swap their spots.

Now for a look at the individual leaders:

Leader Summaries

Gilgamesh of Sumeria
Wars Declared: 43
Wars Declared Upon: 22
Survival Percentage: 90%
Finishes: 11 Firsts, 4 Seconds (63 points)
Kills: 27
Overall Score: 90 points

Gilgamesh was unquestionably the top dog on this map with a victory in roughly half of the games played. The results tallied here likely overstate Gilgamesh slightly because he had a series of narrow-run victories where he pulled out improbable wins. He won by culture in Game #8 three turns before his capital was captured, he won again by culture in Game #15 with his capital under siege from two opposing armies, and he won by Domination in Game #17 (while not at war) two turns before Huayna Capac's cultural victory finished. But even with those outcomes in mind, there was no doubting that Gilgamesh had the best overall chance among any of these leaders to win. For the same reasons outlined above, he had a crazily strong capital city and the opportunity to expand in seemingly any direction. He could attack through Hannibal as we watched on Livestream, he could absorb Peter, or he could take a bite out of Gandhi. All of these scenarios played out multiple times in the alternate history matches. The only times that Gilgamesh failed to snowball ahead were the rare games where he was stuck in an unproductive war for long turns on end, or when he found himself the victim of a gangpile in the later stages. Neither of these happened very often and Gilgamesh was contending for the win pretty much every single time.

Perhaps the most impressive aspect of Gilgamesh's performance was his consistency. He hadn't been eliminated in a single game up until I ran Game #18, and in fact he scored at least 1 point in the first 17 consecutive games played. Even with two eliminations in the last three games, Gilgamesh still finished with a 90% survival rate. He was essentially tied with Mehmed for the most offensive wars initiated, although keep in mind that all of the warring stats are a bit wonky from this game since Gandhi was attacked 59 times without ever starting a conflict himself. Part of the reason why Gilgamesh won so many games was his proclivity to flip on the culture slider and head for an early ending. Eight of his eleven victories came via culture and most of them had extremely fast finishing dates: Turn 319, Turn 282, Turn 325, Turn 290, Turn 272 (!), Turn 312, and Turn 311. Even if someone else had more land area or a better economy, it was almost impossible to beat someone who could clock out of the game with a victory before Turn 300 even arrived. This wasn't the first game where we've seen Gilgamesh head for a cultural ending and it's clearly something that can work to his advantage sometimes. Anyway, it was obvious on repeated playthroughs that Gilgamesh was the best choice to win on this particular setup and he was well clear of the rest of the field in virtually every metric that one might choose.

Mehmed of the Ottomans
Wars Declared: 45
Wars Declared Upon: 18
Survival Percentage: 90%
Finishes: 4 Firsts, 9 Seconds (38 points)
Kills: 22
Overall Score: 60 points

Mehmed was a popular pick for second place in the playoff game and that turned out to be a great call on these repeated playthroughs. Mehmed was another leader who benefitted from a strong starting position that synergized well with his Ottoman techs. He had triple Agriculture resources and Agriculture as a starting tech to connect all of them quickly. One difference from the actual playoff game was the location of Mehmed's second city, which he often sent to the east rather than the northwest, unknowingly grabbing the closest copper resource and having it available as soon as Bronze Working was researched. This had the downstream effect of giving Mehmed stronger units at an earlier date and often causing him to attack Gandhi both sooner and more effectively. Mehmed was the leader most likely to attack Gandhi at an extremely early date and he was often able to catch the Indian leader with no resources connected. Those sub-Turn 100 elimination dates for Gandhi were mostly Mehmed's work, and that's the same reason why Mehmed almost tied Gilgamesh in kills despite being less dominant overall. I'd guess that about a dozen of Mehmed's kills came at the expense of Gandhi, at least half of the Ottoman total and far more than anyone else managed against the unfortunate Indians. This was Mehmed's path to victory and he executed it successfully many times.

As a result, Mehmed was therefore typically much stronger in these alternate history scenarios than in the actual playoff game. He usually gained more from India's inevitable demise and he wasn't hamstrung by all of those barbarian cities in most games. Mehmed's problems tended to crop up down the road in the midgame or lategame, where he wasn't a great techer and he often ended up losing to better AI performances from Gilgamesh or Huayna Capac. He was also limited in his freedom of action by being unable to declare war at "Pleased" relations which almost certainly cost the Ottomans a victory or two. The result was an unsurprising glut of second place finishes, nine of them in all, although often with Mehmed being far stronger than the weakling that limped into the championship game. He was finding various ways to lose in almost comical fashion before picking up four wins in the last eight games which better captured the strength of Mehmed as a leader. The Ottomans equalized the Sumerians in sharing 90% odds to avoid elimination and were in the mix in virtually every game. Again and again there was a lategame trio of Gilgamesh, Mehmed, and Huayna Capac with two of the three of them scoring points for top finishes. Mehmed rarely could beat the other two to an economic win but still managed to get across the finish line with a Domination win and a pair of Diplomatic victories. He was a very strong pick on this map even if he wasn't in quite the same league as Gilgamesh.

Huayna Capac of the Incans
Wars Declared: 23
Wars Declared Upon: 28
Survival Percentage: 75%
Finishes: 4 Firsts, 7 Seconds (34 points)
Kills: 9
Overall Score: 43 points

The pregame favorite in the picking contest came close to equalizing Mehmed's performance despite notable weaker conditions. The biggest thing that the community missed about Huayna Capac was Gandhi's early game performance. Gandhi sent his starting Deity settler to the east in every single game, always grabbing the same floodplains region in between their two capitals and always seizing control of the iron resource three tiles away from Huayna's capital. This spot also became a holy city in every single match with the result that the Incans were stuck in a crowded corner in 100% of the matches. If Gandhi had sent his settler north some of the time or west some of the time, or in any other direction, then Huayna Capac would have been in a far superior position. We can't know where the Deity starting settlers will be headed and even the best guesses at this can sometimes be wildly off. Huayna Capac thus ended up with a tightly squeezed position and with no metal resources under his control. This largely kept him out of early wars and he rarely ended up doing much fighting against Gandhi. To add insult to injury, Huayna Capac usually picked up Gandhi's religion since it would be founded three tiles away from his capital, and once Gandhi was destroyed the Incans often found themselves practicing a religion that few or no other civilizations practiced. This led to several dogpiles against Huayna Capac that ruined otherwise promising games; he was the only leader other than Gandhi to face more defensive wars than offensive wars. It was quite common for Huayna Capac to be trailing far behind in population and territory while still being competitive for the win anyway because his AI personality was just that good.

Huayna Capac had two main routes to victory, neither of which were easy to achieve. The first of these was to gain territory through Gandhi's inevitable demise. A lot of the picking contest entries were backing Huayna for exactly this reason, since he seemed to be in prime position to capitalize on Gandhi's downfall, but it proved to be harder than expected in practice. Huayna Capac's lack of metal resources meant that he had no strong offensive units and therefore didn't have much success in capturing cities when he fought the Indians. More commonly Huyana would simply stay out of the early game fighting altogether, which resulted in Mehmed or Gilgamesh picking up juicy Indian cities along his border (and then causing diplomatic tensions afterwards due to the strong Incan culture). Only when Huayna grabbed a healthy share of the Indian spoils did he become truly powerful and that didn't happen much. The other alternative route to success came from snowballing over Peter and that also didn't happen very often. In part this was because Gilgamesh was usually the one conquering Peter, and it was also due in part to Huayna again not having metals connected in time for any early aggression. While there were a few cases where Peter was still hanging around in the lategame where the Incans were able to snap up those lands to the north, that was the exception and not the rule.

As a sign of how Huayna Capac was operating with a high degree of difficulty in these games, he managed to win three times while *NOT* being the score leader. He pulled this off in Game #10, Game #12, and again in Game #19 with a photo finish as his cultural victory arrived a single turn before Mehmed's spaceship. It's pretty rare in AI Survivor for someone who's not in first place on the scoreboard to pull off the win and yet Huayna Capac managed it in a majority of his victories. His empire was heavy on tundra fishing villages in seemingly every game as the Incans continuously made something out of nothing. One of the tradeoffs for this low production setup was a lack of military success, with Huayna Capac managing about a third as many kills as Gilgamesh and Mehmed. The Incan leader tried to stay out of wars and generally did better when he could avoid conflict to tech away in peace. Huayna Capac did have an unfortunate tendency to avoid researching Rifling tech, including one game where he went all the way to Assembly Line without picking it up, and that spelled trouble many times just as we saw in the playoff game. Even so, I came away from these alternate history scenarios deeply impressed by the Incans. Huayna Capac is one heck of a leader and he managed a top two finish more than half the time despite a challenging setup with crowded lands and no metal resources.

Hannibal of Carthage
Wars Declared: 17
Wars Declared Upon: 16
Survival Percentage: 25%
Finishes: 1 First, 0 Seconds (5 points)
Kills: 3
Overall Score: 8 points

After the first three leaders, all of whom were strong competitors for the win in nearly every game, there was a massive dropoff to the remaining three leaders. The other three leaders were eliminated in seemingly every game and basically never posed any threat to take home a victory, outside of an extraordinarily unlikely win by Hannibal in Game #6. Hannibal was the best of this ignominious group thanks to that aforementioned result and a handful of kills in some random other matches. In his one victory, Hannibal managed to become relevant by conquering Peter (despite Gilgamesh separating the two halves of Carthaginian territory) while Mehmed and Gilgamesh fought a series of long and unproductive wars that saw the former eliminated and the latter marginalized. Huayna Capac was still en route to a likely win when Hannibal scored a Diplomatic victory thanks to the dying Gilgamesh casting his vote for the Carthaginian candidate. This was a bizarrely unlikely result which was not repeated in any other games and the odds of it taking place were likely even lower than the 5% that we saw here. I'd guess that the true odds of a Hannibal victory were probably more like 1-2% on this map.

This was a setup where Hannibal simply lacked options. He was bordering the two strongest leaders on the map and Gilgamesh always placed his (Creative) second city in a spot that badly crowded Hannibal. Game after game saw the Carthaginian civ stretching out in a snaky line along the western coast that had no cohesion and was strategically impossible to defend. Gilgamesh usually didn't attack Hannibal out of the gate as we saw in the actual playoff game, but when he did, it was pretty much a game over situation for Hannibal. There was no way for the smaller Carthaginian civ to defend against the more powerful Sumerians. When that didn't happen, Hannibal was still left with few options. Gilgamesh and Mehmed were both stronger and larger opponents; declaring war on them tended to be suicide. Attacking India didn't yield much in the way of spoils given how far away Gandhi was located. Generally speaking, Hannibal was left as a weak leader who would be snapped up by a stronger rival at some point in the midgame. It's not an exaggeration to say that Gilgamesh's Creative trait more or less single-handedly crippled Hannibal in most of these games. Even though Hannibal appears to be an above-average leader for AI Survivor purposes, he couldn't seem to figure out anything and died in 75% of these games without leaving much of an impact.

Peter of Russia
Wars Declared: 26
Wars Declared Upon: 11
Survival Percentage: 10%
Finishes: 0 Firsts, 0 Seconds (0 points)
Kills: 1
Overall Score: 1 point

Peter did not have the same excuses for his poor showing that Hannibal did and yet he still managed to perform even worse somehow. The Russian leader was not crowded to anywhere near the same degree as Hannibal and he did himself no favors by sending his starting settler up into the tundra in seemingly every game. The land between Peter and Gilgamesh was fairly low in quality and he didn't do a very good job of settling it. Making matters worse, Peter often neglected to research Mysticism tech which ensured that the whole border region would be dominated by Gilgamesh's free Creative culture. Many Russian cities were left crippled for lack of food resources which had been foolishly settled in the outer ring. Peter also had a tendency to cut off his expansion too quickly by starting wars that rarely led anywhere. He liked to attack Gandhi while almost never acquiring any territory of value from the Indians. We saw this in the actual playoff game as Peter fought for nearly 100 turns while gaining exactly one culturally crushed and useless city. Peter's other main target for early aggression was Gilgamesh and this tended to backfire even worse, as Gilgamesh inevitably crushed him when the two squared off. Peter was only First to Die in a single game but that's solely due to the extremely rapid demises of Gandhi elsewhere on the map. Peter was eliminated four times before Turn 167 without being First to Die - thanks Gandhi! The Indian leader kept Peter out of last place only because his peace weight situation was so utterly hopeless.

Peter never accomplished anything in these games so I can't detail what a path to victory would look like for him. He was usually eliminated in the midgame at roughly the same point in time as Hannibal, with both of them having lots of removals in the Turn 150 to Turn 250 range. Gilgamesh was the leader who did the most snacking on these two and thus was the leader who pulled out the most victories. If Peter has a claim to fame in this game, it was being the leader with the most suicidal war declarations. He repeatedly started conflicts with far stronger opponents only to be steamrolled almost immediately. It's impressive in a weird way to start 26 wars (more than Huayna Capac or Hannibal) and have them result in exactly one kill and a 90% self-elimination rate. This guy was a total doofus on this map. Gandhi had no shot but Peter did and managed to fall on his face anyway.

Gandhi of India
Wars Declared: 0
Wars Declared Upon: 59
Survival Percentage: 0%
Finishes: 0 Firsts, 0 Seconds (0 points)
Kills: 0
Overall Score: 0 point

Poor, poor Gandhi. This may be the most hopeless situation we've ever seen in AI Survivor and the repeated playthroughs of this map fully lived up to the expectations of the picking contest. Gandhi was beaten over and over again like a birthday pinata and there was basically nothing he could do about it. He was First to Die fully 80% of the time, 16/20 matches, with those eliminations coming at ridiculously early dates. Gandhi was eliminated on Turn 84, Turn 83, Turn 90, Turn 86, Turn 94, Turn 99, and an amazing Turn 78. He only made it to Turn 150 four times and he only made it to Turn 200 once... only to be eliminated two turns later on Turn 202. Gandhi was absurdly unpopular in these games thanks to his high peace weight, perhaps highlighted best by Game #19 when he was invaded by all five of the other leaders before getting knocked out of the game on Turn 99. All told, Gandhi started zero offensive wars and was attacked 59 times. That's a stupefying amount considering that he was eliminated before the halfway mark of almost every game! Gandhi never even made it to the lategame so there was no possibility of him pulling out a win. If we ran this setup a hundred times, would he survive to the end even one time? While I think there's a chance, I would probably bet "no" if I had to make a wager. Gandhi was never close to surviving in any of the twenty matches that I watched.

To his credit, Gandhi did an excellent job of building his cities in the very limited time that he was given. He tended to found multiple religions and he was usually more or less tied with Gilgamesh for the score lead during the initial 50 turns of the game. Unfortunately, everyone hated his guts and therefore it couldn't last once the war declarations started rolling in. Gandhi's defense was made significantly worse by the Ottomans typically grabbing the only nearby source of copper along with Gandhi's own refusal to research military techs. He had iron thanks to swiping it from Huayna Capac but he wouldn't research Iron Working until there were no other options. This was the reason why Mehmed and Gilgamesh could swarm over India's cities so quickly as their axes and swords found a ready target lacking metals. It was rare to see Gandhi put up the same kind of stout defense that we saw in the actual playoff game where he held off the invaders for long turns on end. He pulled it off in a few games but it was more common to see a rapid collapse. (There were multiple times where I initiated the " play" debug command and walked away to get a drink, only to find that Gandhi was already dead by the time that I came back two minutes later.) No matter how brave or spirited the defense, it was never enough and Gandhi was left as worm food by one conquerer or another.


This was another playoff game where the outcome that we watched largely went according to form. The most fair outcome would have been some way for all three of Gilgamesh, Mehmed, and Huayna Capac to advance but of course only two of the three could make it to the championship. This game was unusual for having such a vast gulf between the top three and bottom three leaders in terms of performance. It was also strange for having 11 cultural victories and 3 diplomatic victories - there were only 2 spaceship finishes out of the 20 games that I watched! That was probably an outlier result but still weird to note. These games were also unusual for having some extremely fast finish dates with an average finish of Turn 309. There were eight different games that finished before Turn 300, including a blazingly fast Diplomatic victory on Turn 266 in Game #18 and a Domination ending on Turn 236 in Game #2 which would be the fastest-ever by a wide margin. This is likely part of the reason why the number of wars was so low; after we saw only 7 wars in the actual playoff game, that number was reinforced by having 90% of the alternate history scenarios end up in single digits. The other reason is that the field narrowed rapidly in each game, with three weak leaders typically seeing fast exits, and that cut down on how many wars could take place. If there are only three leaders left standing in the midgame, there can't be very many war declarations taking place.

It's nice to have confirmation that this was another mainstream result that didn't spiral off into crazyland. As always, thanks for reading along and following Civ4 AI Survivor.