Civ4 AI Survivor Season 7: Championship Game Alternate Histories


Championship Game 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. The Season Seven Championship saw a fairly straightforward Louis snowball, as he and Mehmed teamed up to roll over the rest of map. 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 Championship Game

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: I think it would have been pretty easy, after watching Louis unequivocally crush the competition on Livestream, to assume that a fairly typical outcome had taken place; indeed, that's what I had assumed. But instead, it quickly became clear that, as per usual for these Championship games, we did not witness a typical game on Livestream this year at all! There were two games that followed eerily similar paths to the livestreamed Championship (Games 15 and 19), where Louis snowballed off an early dogpile of Darius. That turned out to be a very unlikely occurrence though, as Louis was rarely left alone for long enough to muster a successful conquest, and even when he did, he was also thoroughly capable of bungling a surefire win afterwards. Instead, it was the other two northern leaders, Ramesses and Saladin, who found themselves competing for the wins repeatedly. Meanwhile, Darius turned out to be a dark horse candidate, while Louis (and Churchill) could only win through outlier scenarios, and Mehmed couldn't win at all. Broadly speaking, there were two overarching factors that affected how these games played out, cultural emphasis and diplomacy. Both favoured Ramesses and Saladin, and this was the reason for their success; meanwhile, Churchill benefited only from diplomacy, Louis only from culture, and poor Mehmed benefited from neither (so did Darius, but he scammed a few wins with Financial anyway). Viewing the game through this lens creates a narrative that explains why Ramesses and Saladin were so strong, and why Louis' win was unlikely but not impossible.

The championship map is relatively lush compared to our normal games, with very strong 'Big Fat Crosses' available all over the map. This means that leaders who emphasise early border pops tend to snowball heavily on this map compared to leaders who don't, with Creative trait being uber-powerful in this context. A clear comparison can be drawn, for example, between the usual fourth cities of each leader (pictured below). For Louis, Rheims became immediately productive soon after settling (usually ~T40), able to work furs, pigs, and rice tiles. Meanwhile, Mehmed's Bursa lacked food resources entirely until it could pop borders, usually with libraries, which basically didn't happen until T90 or even T100. That was a long time when Louis could work his food resources and Mehmed couldn't! Although these two were the most extreme in either direction, this dynamic played out all over the map, as the three northern leaders (Louis, Saladin, Ramesses) used Creative trait, religion, and monuments to pop their borders rapidly, while the three southern leaders (Darius, Churchill, and Mehmed) typically failed to pop their borders for millenia. The knock-on effects appeared in every game: the northern leaders were dominant in terms of land, economy, and production after the landgrab, with the three southern leaders invariably forced to play catchup. In particular, Louis always came out of the landgrab the strongest, ahead in score usually by at least 100 points, and near-double the laggards. He would be followed by Ramesses and Saladin in some order, and then there would be a big gap to the other leaders.

It was a pretty sad comparison between Bursa and Rheims; this imbalance between the north and south occurred all over the map.

The reader might reasonably wonder then: if Louis was so strong after every landgrab, why did his win turn out to be an outlier, with the other two northerners competing for the overall trophy? The answer is equally straightforward: diplomacy. The diplomacy for this game played out in a way that makes perfect sense in retrospect, but that I don't think I (or the broader AI Survivor community!) expected beforehand. These leaders can easily be split into a 'Good' group (Ramesses, Churchill, and Darius) and an 'Evil' group (Saladin, Louis, and Mehmed). Traditionally in AI Survivor, the Evil leaders perform better than the Good ones, but that wasn't the case here, for two very clear reasons. The first was that while all members of Team Good were able to contribute in various ways, Team Evil was stuck with a member who bordered on useless. Mehmed was a total catastrophe across these games, the weakest leader on nearly every occasion. This was partly for the cultural factors discussed above, as Mehmed repeatedly failed to pop his borders for over a hundred turns, far worse than even his fellow southerners. However, even in the games where Mehmed wasn't crippled from the start, he played inexplicably poorly, particularly when fighting wars. For example, in Game 13 Mehmed lost a war to Churchill while having double his power, and in Game 11 he gave away a core city for peace in a war he was unequivocally winning. Most spectacularly, in Game 15 Mehmed conquered Churchill down to four cities, before losing all 5 that he'd taken, PLUS giving away an Ottoman city for peace. A common theme here was Mehmed failing to bring enough siege units, much like we saw in Game 6 earlier this year. There were seriously Montezuma levels of incompetence on display here. In any case, it was bad news for Louis, who found himself fighting a coalition of Good leaders in the mid-late game without much support. The closest comparison to Mehmed here was Churchill, who also found himself very behind after the landgrab, but Churchill was consistently able to fight his way back to relevance, in stark contrast to Mehmed's ineptitude, and was rewarded for his effort by a series of second place finishes and kills (most of which came from, you guessed it: Mehmed). Those Charismatic/Protective redcoats meant business!

Even when Mehmed was having an unusually good game, he was more than capable of screwing it up somehow - in this case (Game 7), Mehmed snowballed via eating Louis, which left him diplomatically isolated for the rest of the game. Also note Louis' initial scoreboard dominance, and Mehmed's initial weakness.

Secondly, while Team Good was a generally strong alliance who rarely attacked each other (particularly Ramesses and Churchill, who formed a rock-solid bond over these games), Team Evil had a semi-consistent turncoat in Saladin. The Arabian leader usually took Louis' and Mehmed's sides during the various wars, particularly against frequent heretic Ramesses, but he was also very capable of using religion to swap sides. These religious backstabs were often quite well-timed, such as in Game 18 when Saladin annexed Louis, or Game 8 when he ate Mehmed; each was key to Arabian victory. This ruthless backstabbery worked out very well for Saladin, and was responsible for him taking a strong second in the overall scoring, but it was a disaster for Louis and Mehmed, who had one less ally to consistently rely on, and found themselves in two-front wars repeatedly as a result. In general, Team Evil fell to infighting a lot, while Team Good did not, as Louis and Mehmed also fought surprisingly often. For example, they clashed in Games 7 and 17, which always ended disastrously. Considering they were mutually Pleased at 0 relations, this was surprising, but can be explained by a combination of border tension and religious differences. When combined with Mehmed's general incompetence, these factors combined to prevent Louis from repeating his dominant snowball more than a handful of times, and allowed Ramesses and Saladin to fill that power vacuum. Meanwhile, although Darius managed an economic snowball a few times, he was typically savaged by Louis or Mehmed in the early game; even once Ramesses or Churchill cleaned up Team Evil, Darius would be left crippled, incapable of truly contending. To be honest, it's a bit of a shame we didn't see either of Saladin or Ramesses win - I have a feeling that the AI Survivor community would have been very surprised!

Now for a look at the individual leaders:

Ramesses of Egypt
Wars Declared: 29
Wars Declared Upon: 33
Survival Percentage: 60%
Finishes: 7 Firsts, 4 Seconds (43 points)
Kills: 11
Overall Score: 54 points

Ramesses narrowly graded out as the top leader on this map, and I think that was a deserved result. He won a third of the time, was equal for second place finishes until Game 20, and only died early to coordinated dogpiles from Team Evil. Ramesses was very effective at building and expanding on this map, usually having among the most cities at the end of the landgrab while simultaneously building multiple key wonders (Ramesses loved both Stonehenge and the Pyramids). He also founded the non-Saladin religion in most games, and built a series of very early shrines using the Egyptian obelisk (the earliest I recorded was T42 - that's pretty disgusting!). Combined, this all meant that Ramesses could become an economic monster if left alone, and it also created a positive diplomatic environment for him, as his religion was the most likely to spread across the map in the early game thanks to that shrine. Ramesses was, if anything, more capable of snowballing peacefully than Darius, despite lacking Financial trait, as his burgeoning collection of wonders was just as powerful as any trait; he was also less likely to get attacked thanks to his lower peaceweight. It's easy to see what went wrong on Livestream: Churchill attacked Ramesses quite early, which was never replicated in these games (the closest was Game 17, where a snowballing Churchill bulldozed Ramesses in the lategame). It was also responsible for Team Evil winning, as Team Good uncharacteristically sabotaged their own game. The more common issue for Ramesses here was that although he was decent at fighting at tech parity, he could fall behind in military technology despite being ahead in overall beakers. Saladin especially would often rush Rifling, and opportunistically use this window of disparity to cripple the Egyptians. This was reflective of a general trend, as Saladin and Ramesses fought constantly in these games, due to founding both the early religions. The winner of these wars would typically go on to win the game (they split around 2/3 of the total victories), while the loser would become irrelevant (though at least one of these leaders advanced out of 17/20 games, they advanced together only thrice - in all three, Ramesses did not found the other early religion). In the games where either Ramesses or Saladin was weak, it was usually because one had crippled or killed the other (or they'd fought each other out of contention, as in Game 7), and if these two had just been a bit friendlier, I suspect they would have basically split all of the victories between them.

Saladin of Arabia
Wars Declared: 35
Wars Declared Upon: 28
Survival Percentage: 55%
Finishes: 5 Firsts, 4 Seconds (33 points)
Kills: 15
Overall Score: 48 points

Saladin was the second-strongest leader here, with a big gap between him and third-place Churchill. This also felt fair, as Saladin was consistently in the mix without becoming truly dominant in many games. Despite starting with The Wheel and Mysticism, Saladin was one of the best expanders here due to popping his BFC borders consistently - this supports a growing consensus in the AI Survivor community that Mysticism is actually one of the best starting techs for (non-Creative) AIs, to prevent cultural self-sabotage. Combine this with founding his own religion, and Saladin shared many of Ramesses' strengths, excluding the wonder-based economic dominance. Saladin made up for this by fighting more successfully, especially around the period of rifles and onwards. It was very rare for Saladin to conquer a neighbour before then, but he often used rifles to get the snowball cooking (see Games 1, 4, and 18). Overall, Saladin's gameplan was to do everything competently - expand, research, fight, negotiate - without excelling in any one category. It wasn't enough to quite match the Egyptians, but it was a successful strategy in its own right, even if it was quite boring. Saladin's other area of strength was diplomacy, where he was the only leader capable of consistently playing both sides of the Good/Evil dynamic. Saladin more frequently joined Team Evil, particularly because of his feud with Ramesses, but he was also capable of successfully switching sides, as in Games 2, 3, 8, 13, and 18. On the other hand, Saladin was the jack of all trades, but the master of none. He could fall behind other leaders if they managed to execute their narrower game plans properly. This could be Ramesses or Darius out-teching the field, and Louis or Churchill snowballing militarily (Mehmed doesn't warrant discussion here). This was the main problem for the Arabian leader: he would win if no one else snowballed properly, but had no effective counter when they did. He could also falter if his religion failed to spread to other leaders, but that's just a part of life for the religious leaders. Overall, it made for a boring but effective leader, who deservedly came second in the Livestream game.

Churchill of England
Wars Declared: 30
Wars Declared Upon: 35
Survival Percentage: 65%
Finishes: 2 Firsts, 5 Seconds (20 points)
Kills: 13
Overall Score: 33 points

If I had to sum up Churchill's performance here in a single word, it would be 'consistent.' Churchill's performance changed little between games, as he always played the same way, with solid-if-unspectacular results. He was the only leader who never died first, and was usually allowed to settle his 8-10 natural cities in relative obscurity while the rest of the world exploded into violence. The only leader Churchill fought consistently with was Mehmed, and Churchill almost always got the better of their little feud. This was initially hard to explain, as Mehmed usually exited the landgrab in a similar position to Churchill, and could usually snipe 1-2 cities in his initial attack - shouldn't the Ottomans have been stronger at that point, and able to finish their conquest? Certainly, one wouldn't have expected Churchill to consistently turn these wars around, retake his lost cities, and push through into Ottoman territory. But that's exactly what happened, time and time again, as almost all of Churchill's kills came against the hapless Ottomans, and all of his placements came from those kills. Now it would be easy to say something along the lines of “well of course Churchill killed Mehmed - Mehmed sucked!” And that is true, to a certain extent. But there were also many, many games where Mehmed had Churchill on the ropes, and by rights the English should have died, only for Protective/Charismatic redcoats or longbows to break onto the scene and change the course of history. This wasn't just dumb luck, either - despite being considerably behind Darius, Saladin, and Ramesses in tech, Churchill beelined Rifling very often, and seemed highly cognizant of the need to use his redcoats before they obsoleted. Almost all of Churchill's kills were scored with redcoats specifically, and he rarely started wars before unlocking them. Now, this was all luck to an extent, particularly the war-timing, but Churchill's aggression rating and tech flavours worked exceedingly well for him here. The problem, on the other hand, was that Churchill absolutely needed those things to work, because by T100 he was always left with a massive hole to dig himself out of. Churchill rivalled only Mehmed in his ability to ignore border pops, and this meant that each of them was consistently out of the running by T150 in most cases. The difference was that Churchill was able to claw his way back to relevancy in most games, and in some extreme cases, snowball out to a win (in Games 9 and 17, he killed multiple opponents en route to dominant victories). That was certainly more than Mehmed could do, and Churchill was a very dangerous combatant in these games, even if he couldn't compete economically with the top leaders.

Darius of Persia
Wars Declared: 29
Wars Declared Upon: 44
Survival Percentage: 50%
Finishes: 4 Firsts, 1 Second (22 points)
Kills: 7
Overall Score: 29 points

More than any leader I've seen in an Alternate History thus far, Darius here demonstrated the importance (and randomness) of diplomacy in overall performance. It's hard to overstate how strongly Darius' good games correlated with facing few defensive wars, and vice versa. To quantify it, in the five games where Darius placed, his offensive to defensive wars ratio was 16:6, and in the other 15 games, it was 13:38 - that's a pretty enormous difference! So yes, the typical Financial scam was in full effect here, as Darius could absolutely run away with the game if he was left alone for the first 200 turns, but tended to fold totally if he was attacked before then. Louis was the biggest problem for Darius here, as they clashed in nearly every game, while Mehmed was more likely to fight Churchill for whatever reason. It was when Mehmed and Louis did work together that Darius performed worst; this occurred in Games 1, 4, 8, 15, and 19, all of which ended in Darius dying first. This was also what happened on Livestream, as we witnessed pretty much the worst-case scenario for Darius there. It was these games that tended to end well for Team Evil, as they were able to knock out a member of Team Good early enough to overcome Mehmed's weakness. Ultimately, I'm not sure what else to say for Darius here. He performed well when the diplomacy broke out in his favour, and more frequently collapsed or became irrelevant when it didn't. There honestly wasn't much more to it than that.

Louis of France
Wars Declared: 38
Wars Declared Upon: 42
Survival Percentage: 35%
Finishes: 2 Firsts, 4 Seconds (18 points)
Kills: 10
Overall Score: 28 points

Well, this was embarrassing... after a dominant performance in his other Alternate History appearances this year, Louis completely dropped the ball here, only able to replicate his victory twice, and otherwise among the least likely leaders to even survive the game. Indeed, until Game 15, it was actually looking like Louis might join Mehmed in the no-win club. In the two games where Louis did pull through to win (Games 15 and 19), he received most of the spoils in an early dogpile of Darius, just like the Livestream game. From there, he and Mehmed would roll over the rest of the map, and in each of these games Louis similarly appeared as the dominant leader. The rest of the time though, Louis was an incoherent mess, trying to do everything at once, and generally making a mess of it all. This was particularly surprising given that Louis invariably exited the landgrab phase as the unambiguous top dog, clearly leading and apparently set up to crush one of his rivals. He had phenomenal starting techs and the best Championship-map trait (Creative), seeming poised for greatness. However, it was in the 'crush a rival' phase of Louis' game that things typically went wrong. Because he expanded so much faster than everyone else, Louis usually started the game's first war, and it could be against Darius, Saladin, Mehmed, or Ramesses, depending on how the religions shook out. However, these wars were usually started before catapults, leaving Louis to ram his units into his enemies' walls while the other leaders, especially Ramesses and Saladin (who were more passive, especially early on), caught up to him. Indeed, by T150, Louis' lead had usually evaporated, as his wasted production spiralled out of control. Then things would go from bad to worse, as one of the rivals that he hadn't initially attacked (or who wanted revenge for an earlier war) would join in and put Louis into a 2v1. Louis would usually hold on for a while, evidenced by even his first-to-die deaths all occurring after T200, because he was so strong to begin with. But Louis fought a lot, with the second-most offensive and defensive wars, and inevitably these wars would catch up to the French economy. Louis would fall behind in tech, or production, and be slowly taken apart - it was a pretty amazing fall from grace, considering the position of strength that he always started in.

Louis also had issues with deciding when to turn on the culture slider, as Slashin' had already noted in Game Eight earlier this year. He would often turn on the slider ridiculously early, before even researching Rifling, and this got Louis killed repeatedly, or at least cost him the win, in games that he should have been competing for (Games 1, 5, 11). In several other games, he also just had a terrible tech path, putting on his Willem hat and ignoring Rifling to farcical levels (Games 10 and 18). Louis then wasn't helped by the extreme incompetence of his supposed natural ally Mehmed, who both attacked the French repeatedly, and basically donated his land to Churchill or Darius in game after game. Nor by Saladin's willingness to throw him under the bus in the lategame. Basically, the setup for this game was significantly biased against Louis, despite the position of strength that he always occupied after the landgrab. Louis himself also had multiple issues that cost him several games he should have won, and I suspect a more coherent AI package would have helped the Frenchman greatly. Unlike on the vastly unfair maps that Louis lucked into for his first two games this season, he was only able to overcome these problems a few times - it just so happens that we got one of them on Livestream. A pretty underwhelming performance from the Season 7 Champion, I'd say.

Mehmed of the Ottomans
Wars Declared: 48
Wars Declared Upon: 28
Survival Percentage: 25%
Finishes: 0 Firsts, 2 Seconds (4 points)
Kills: 6
Overall Score: 10 points

Mehmed was an absolute catastrophe in these games. He was never remotely in contention for the win, and I want to make it very clear that I think if I had run 100 games instead of 20, he still might not have won any of them. Mehmed failed to pop borders to a hilarious extent, which was a massive contributor to his weakness. He couldn't work food resources or connect luxuries for extended stretches of turns, and this nearly always crippled him by T100. He was also a part of the unfortunate Team Evil, but unlike Saladin and (especially) Louis, Mehmed wasn't let down by his allies - he was the problem! So often, Mehmed would start a war with Churchill or Darius or Ramesses, and just totally stalemate, incapable of making any progress whatsoever despite having the highest power in the game. Eventually, through poor strategic and tactical decisions, those stalemates would start turning against the Ottomans, and suddenly Mehmed would find himself on five cities and about to die (this was especially common against Churchill, who simply outworked Mehmed time and time again). Hilariously, the only opponent that Mehmed could beat militarily was Louis, who he also fought repeatedly (e.g. Games 7 and 12). To his credit, Mehmed did start the most wars of the group, trying to dig himself out; on the other hand, a lot of those wars were very unwise, and got Mehmed killed. He was simply so bad at fighting them that it would have been better to just huddle up in the corner, and hope to be left alone. It wasn't just that Mehmed was weak - he also played horrifically badly, repeatedly failing to build enough siege units, and slamming his janissaries into the walls of London or Persepolis like it was T75, except it would be ~T200. Mehmed's only genuinely strong game was Game 19 (Game 14 was a classic backdoor), when he unusually opened with Mysticism and built Stonehenge, which allowed him to actually expand properly. In this game, he was a genuine threat to Team Good, even if he never looked like winning due to Louis' inevitable culture clock. This was a genuinely embarrassing performance from the Season Five Champion, and frankly I have absolutely no idea how Mehmed ever won a championship three years ago. What a disaster of a leader, man.


It turns out that we got a result on livestream that was certainly plausible, and basically replicated twice, but that was decidedly not the 'expected' outcome for the Championship. Instead, Ramesses or Saladin should probably have come first, while any of Churchill, Ramesses, Saladin, or Louis would have been a reasonable second (at least we hit there!). We also had a low number of wars compared to the games I ran, likely because Louis achieved victory so quickly (the same thing happened in Games 15 and 19, Louis' other wins). This is becoming a trend with the Championships at this point, with all three of the most recent livestream outcomes for these games being proven as outliers - just a weird fluke of small sample size, I suppose.

It is the third AH that I've run (alongside S3G1 and S7P1) where Ramesses has been the strongest AI; it is becoming clear that the Egyptian leader has been unlucky in the past, and is likely underrated by the AI Survivor community as a result. It all makes sense on paper: Ramesses has excellent starting techs, prioritises culture (and thus pops borders), can fight wars effectively, and snowball peacefully through monopolizing the game's wonders. The total package this creates is powerful, and Ramesses is hopefully less likely to be underestimated in future. The same is true of Saladin - it turns out that being a competent fighter, founding a religion, and popping BFC borders early on, are enough to be competitive! - who will similarly likely be more highly regarded in future seasons. Each of these leaders was unlucky not to perform well on Livestream, and I'm particularly sad that Ramesses didn't get his moment in the sun. The other thing to note is that unlike the Livestream game, a lot of these were actually very interesting, close games to run! Frequently there would be a close spaceship race between Ramesses and Darius, or Louis would be pushing culture as Saladin launched the spaceship, or Churchill would fight all comers to come back from the dead. After a few pretty dull games in a row, it was a joy to watch multiple leaders actively contend for the win in game after game, and although it's sad that our Livestream game was among the less exciting, I'm glad I didn't have to sit through 20 more iterations of that. I'm looking forward to the next season, and I wonder if our perceptions of many of these leaders will change as a result of these Alternate History games - I know plenty of mine have.

Cheers ~ Amicalola