Playoff One 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 One was billed as a duel between Mansa Musa and Pacal, however it turned out to be a tight race between Mansa and Elizabeth instead. 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 JackRB - many thanks for spending so much time on this task! 
JackRB 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:
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.)
JackRB: My first time running AI Survivor Alternate Histories, and I was particularly interested in running Playoff 1 as my first game to determine two simple questions. Firstly, whether Mansa or Elizabeth could decisively prove to be the best leader on the map, and secondly, if Pacal could redeem himself with some wins of his own given his atrocious performance in the actual game.
As it turned out, the actual game we watched was a pretty standard game in the grand scheme of things, with Mansa 1st / Elizabeth 2nd / Gilgamesh First to Die all being the most likely in their respective categories over the series of repeats. What was unique was the total self-destruction of Pacal, which freed up Elizabeth to begin the dogpile on Gilgamesh and take a significant chunk of the spoils herself, and more importantly away from Mansa. This, combined with a few lucky barb captures, demonstrated a greater degree of "over-performance" from Elizabeth relative to Mansa Musa, made evident over the course of watching 20 extra replays.
The typical script the game followed, when Pacal didn"t commit total self-sabotage, were two different 2v1"s centred against the low peaceweight leaders. Pacal would almost always be attacked by Hammurabi and Elizabeth early, either due to Pacal attacking Hammurabi and Elizabeth jumping in to help (or vice versa), or simply victim to both his neighbours individually deciding to attack him without provocation. 2v1 wars are always dangerous for the AI, and while Pacal was frequently the strongest of the trio, he struggled to make headway with 2 borders to defend and 2 civs worth of production queues to grind through. After much toil and struggle, the Mayan lands would be partitioned and roughly split between the Babylonians and the English, with the breaking point usually coming when one side re-entered the fray after some period of recuperation of which Pacal was never afforded.
Running in parallel on the other side of the map, Gilgamesh would almost always declare on Lincoln due to the amount of border tension and was usually successful at getting an early conquest or two, rendering Lincoln irrelevant in the grand scheme of things. Meanwhile Mansa, frequently the last to involve himself in a war, simply waited his time and blindsided Gilgamesh at an opportune moment with a numerically and technologically superior army. How well Gilgamesh played up until this point merely dictated how long it would take for him to be crushed by the Malinese, not whether he could survive the onslaught. Sometimes Lincoln survived, sometimes not, but it never mattered as Mansa would invariably grab all the spoils and have two civs worth of territory around the Turn 200 mark.
These two threads of the game would run concurrently, with Gilgamesh usually being the first of the low peaceweights knocked out, unlocking Mansa (and even Lincoln) to pile on the misery for the Mayans and wipe any trace of low peaceweights clean. After the smoke of war cleared, Mansa would emerge as the largest civ and race ahead to the victory condition of his choice with Elizabeth a close (if not as close in the actual game) second. The Malinese leader would frequently be playing from behind technologically due to his weaker land and Elizabeth frequently building every wonder under the sun, but once he integrated the Sumerian cities into his economy he blasted ahead and sailed to some kind of victory, all winning by Culture, Spaceship, Domination and Diplomatic victories multiple times. Diplomatic would actually be the most common victory, as with the shared wars and similar peaceweight, Elizabeth having Free Religion as her favourite civic and all the other miscellaneous relations bonuses - there"d be a good 50+ turn gap between the UN being built and a non-Diplo victory being achieved, giving plenty of time for the AI to randomly cut the game short via the victory vote (In Game 11, a Diplo victory vote succeeded the same turn the Spaceship arrived but Diplo victory triggered first, the more you know!). Frequently this just cut down on waiting and gave it to the leader who was already winning, although Mansa did manage to scam two more wins from losing positions, both Elizabeth cultural pushes (Games 13 and 14).
The mantra of "peaceweight is destiny" was in full effect here, and it"s notable that religion was not a reliable counterbalance to this. Frequently religious alliances across the peaceweight divide would get ignored and devolve in war anyway, but this is because the religious map as a whole was very fractured and ever-changing game to game. Pacal reliably founded the first religion but beyond that was anyone"s guess, with everyone else founding their own competing religion multiple times so any meaningful religious bloc would crumble to pieces at the early-to-mid-game where the Code of Laws and Theology religions were picked up - with every AI wanting to run their own self-founded religion instead. This left peaceweight as the lowest common denominator for AI"s picking wars, and would need a few low-odds dierolls to be convinced otherwise.
There were a few games that did deviate significantly from this typical script, something that seems inevitable when you roll dice enough times. For example, Game 7 saw Hammurabi join forces with Gilgamesh to set up an early 2v1 on Mansa leading to Hammurabi"s sole victory, Game 16 saw Pacal declare Elizabeth on Turn 57 (!!) and capture London on Turn 80, while Game 17 saw an unusually weak Mansa allow Gilgamesh to run over the whole map in an astonishing four kill Domination game. However, these were few and far between and on a macro level the games were pretty consistent, which is why there"s a clear divide (just like the actual game) between the winners and the others.
Now for a look at the individual leaders:
Mansa Musa of Mali
Wars Declared: 35
Wars Declared Upon: 11
Survival Percentage: 95%
Finishes: 15 Firsts, 2 Seconds (79 points)
Kills: 19
Overall Score: 98 points
Mansa Musa continues to assert his credentials as one of the best AI"s for this competition, just falling short of breaking the 100 point barrier. We all know his economic game is second-to-none, but in this set of games what impressed me most was his repeated success in warfare, backed up by grabbing an impressive 19 kill haul in just 35 offensive wars, all primarily inflicted on poor Gilgamesh. His starting land wasn"t fantastic, he would usually be middling in technological prowess vis-a-vis Elizabeth, but consistently strong expansion and smart tech paths laid the groundwork for repeated military success - with Mansa"s few failures to place in the top 2 revolving around his inability to snowball early. I noted that Mansa had a tendency to go early Banking which meant he was running around with Knights well ahead of everyone else, and saw said Knights sweep across the Sumerian countryside time and time again. Once Sumeria was conquered, Mansa had plenty of ways to pursue a victory of his choosing. He"s usually cash out with a Diplomatic victory with all the friendship built up over the course of the game, but was also willing to go Space, Culture (not as frequently as you"d think, lacking the technological strength to found enough religions) and even Domination if feeling bloodthirsty late on, usually steamrolling Hammurabi in the process.
Elizabeth of England
Wars Declared: 36
Wars Declared Upon: 22
Survival Percentage: 90%
Finishes: 3 Firsts, 10 Seconds (35 points)
Kills: 11
Overall Score: 46 points
Elizabeth had a great breakthrough season this year, finally getting a friendly environment to show off her technological prowess. She did exhibit the usual weakness of peaceful builder AIs (exacerbated by the Stone/Marble), frequently stuck at 4-6 cities. However they were extremely high quality, the oodles of wonders ensuring she was the tech leader early on, and she was frequently able to expand militarily despite her small size. Elizabeth was in the best position to vulture territory from both Pacal and Gilgamesh, and she was consistently both proactive enough, and sensible enough to get necessary military tech (i.e. Redcoats) to take advantage. On the whole though, her set of alternate history games were akin to her actual game, solid performances that easily cashed a top two finish, just frequently unable to beat Mansa in a victory footrace, coming up a handful of turns short time and time again - but to even be in a position to give Mr Moneybags himself a run for his money is no mean feat.
Hammurabi of Babylon
Wars Declared: 35
Wars Declared Upon: 26
Survival Percentage: 70%
Finishes: 1 First, 3 Seconds (11 points)
Kills: 6
Overall Score: 17 points
Hammurabi was the exact AI that Mansa and Elizabeth needed to be as successful as they were on this map. A high peaceweight leader, not too aggressive but willing to cooperate in taking down the low peaceweights (primarily Pacal), yet having essentially no chance of being victorious himself due to his relatively mediocre trait combination and economic incompetence. He got a few 2nd places here and there due to unusual (a.k.a lucky) success in the early warfare, and his sole win happened when Hammurabi switched sides and collaborated with Gilgamesh to cripple Mansa early. But the vast majority of the time he was a useful idiot and extremely passive once Gilgamesh and Pacal had been wiped from the map.
Gilgamesh of Sumeria
Wars Declared: 41
Wars Declared Upon: 31
Survival Percentage: 15%
Finishes: 1 First, 1 Second (7 points)
Kills: 8
Overall Score: 15 points
Gilgamesh really tried his best, reliably warring Lincoln with good success but Mansa"s subsequent invasion often rendered his efforts pointless, all a well played game would usually do was prolong his inevitable death. The diplomacy was just overwhelmingly tilted against him, and just how peaceful leaders like time to build, warmongers need time to focus on a single war to begin the snowball and Gilgamesh was very rarely afforded this chance. Despite this usually uniform outcome, Gilgamesh was more than willing to demonstrate the very best and worst in warmongers. On one hand he had a spectacular 4-kill victory in Game 17, on the other there were a few games where he committed "Season 6 Qin"-tier economic suicides, the worst being Game 16 where he mismanaged his economy to the point of losing over 20 gold per turn at 0% science, limping to Pottery at a record-breakingly slow Turn 162 (!!) and being eliminated a mere 40 turns later. Entertaining to watch, caused plenty of trouble, but with little to show for it.
Pacal of the Mayans
Wars Declared: 26
Wars Declared Upon: 67 
Survival Percentage: 20%
Finishes: 0 Firsts, 3 Seconds (6 points)
Kills: 1
Overall Score: 7 points
Pacal was never as face-plant stupid as the actual game, but he was as capital-D Doomed as Gilgamesh, and one only needs to see the 67 defensive wars faced to see why. He was never able to build any meaningful alliance via religion, and a great capital alone is simply not enough to consistently overcome 2v1s. He was surprisingly good at clinging on to life however, rarely being First to Die, but that only invited more opportunity to face more opponents, demonstrated by there being 4 total games where Pacal faced 6 or more (!) defensive wars in total. No AI could survive such a ruthless onslaught. Pacal"s ability to pick up points was totally out of his hands, and hinged on Mansa attacking Hammurabi early (plus Gilgamesh"s miracle 4-kill game) to have any chance of backdooring his way to a second place finish, never even remotely threatening a top spot.
Lincoln of America
Wars Declared: 13
Wars Declared Upon: 29
Survival Percentage: 45%
Finishes: 0 Firsts, 1 Second (2 points)
Kills: 4
Overall Score: 6 points
Last and certainly least of the high peaceweights, Lincoln was Gilgamesh"s designated punching bag, often weakened to permanent irrelevancy. Even in games where Gilgamesh fell flat on his face and was little threat, he was usually unable to expand beyond his meager territory, and he was mediocre economically and even struggled culturally. I expected very little from Lincoln based on his actual game performance and I was frankly still disappointed.
What we saw on livestream was a pretty standard game in all the ways it mattered, although we got some excellent entertainment in Pacal failing spectacularly and a nailbiting finish near the end that was never quite replicated in any of the alternate histories - the game was kind enough to inject some drama in what was otherwise quite a buildfest. The main takeaway from this map was Elizabeth"s performance, finally getting back into the limelight of good performances after many years of unlucky performances and tough breaks.
