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 Three featured a drama-less roll over the map by Darius, with a strong Saladin finishing second. 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:
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: After the first four games, three of which were near-identical to the livestream, I thought Playoff Three was basically 'solved,' but then there were a few twists afterwards. Even so, the game we saw on livestream turned out to be a very typical outcome, with Darius being the most likely winner, Saladin the most likely second-place, and Pacal having a decent chance for first to die. As these maps so often have, there was clearly a default outcome that occurred if no one could snowball out to an unlikely victory. In this case, the default outcome saw three extended 1v1 wars between Darius/Qin, Pacal/Gandhi, and Saladin/Alexander. In this scenario, Darius was typically the strongest leader, able to muscle out the rest in tech and kill 1-2 opponents in the mid-late game. Oftentimes Saladin, Qin, or Alexander would be able to kill an opponent too and also be snowballing, but hilariously slowly compared to the Persian leader, who won a full half of the games in this fashion. The other leader who turned out to be competitive for the win was Pacal, which probably won't surprise many people. Pacal's classic trait combo of Financial/Expansive Low-Peaceweight proved enough to win several high-risk, high-reward games. Pacal's performances swung heavily on whether he was attacked throughout the game, with his successful games having mostly offensive wars (see Games 5, 15, and 16) and his worst games having defensive ones (Games 2, 3, and 4). If Pacal was attacked early on, he would typically collapse, but if he was allowed to snowball and choose his targets, he was genuinely competitive with Darius for the win.
So why was Darius so strong? In short, land and diplomacy. Darius' capital had wheat and ivory, which Darius could improve straight away using his starting techs. At his second city, he also had gold for a second early luxury, while his capital also held copper. From there, Darius had the biggest cities and the early power lead that would have had more competent AIs salivating. He could expand northwards at his leisure (Gandhi was shockingly bad, at both winning and expanding) and typically ended up with the most cities, the best cities, and a massive tech lead. The diplomacy also lent itself very well to Darius' position here. The low peaceweight leaders were not harmonious at all in these games, rarely able to work together. The most common infighting was between Saladin and Alexander, which happened almost every game, but Alex generally ruined things for everyone as he repeatedly attacked Qin and Pacal as well. Even when the low peaceweight leaders could get a coalition going, it was usually against Gandhi, whose slightly higher peaceweight (and much lower power) made him a much more tempting target than Darius - this was responsible for Gandhi's high first-to-die rate. Darius could use Gandhi to buy time, such that even when the other leaders did finally get around to cooperating against him, it was too late, and he would eviscerate their macemen with infantry (see Games 3, 9, and especially 11 where Darius took on an extended 3v1… and won it!). Even that was in the rare scenario that the other AIs could cooperate though, and it was more common for Darius to pick them off one by one while they engaged in petty squabbles. Really, I want to make it clear the extent to which Darius was essentially gifted this game; I honestly think it is a poor indictment on his AI ability that he 'only' managed to win half the time.
The games that Darius lost were a lot more interesting, and tended to go in a few directions. The first and most obvious outcome was Pacal winning instead. This was very rarely due to Pacal actually outracing Darius in a fair fight (the Mayan starting position simply did not compare to the Persian one), and more commonly because something had gone wrong for Darius. Occasionally, Qin Shi Huang could beat Darius in the 1v1 fight (this alone shows Darius' incompetence - he should have won that war every time). When this happened, someone else had to fill the Financial void, and Pacal usually stepped up, as in Games 5 and 18. Alternatively, the dogpile of Darius could be uncommonly successful, usually when it happened earlier than normal. Finally, Pacal could occasionally get a quick solo-kill on Gandhi or Saladin, which allowed him to remain competitive with the Persian monster (see Game 8). The rare games where neither of these leaders won were the most interesting though. In Game 1, Alexander settled ON copper, ON a river, which connected copper to all three of his cities on T32. He then attacked Saladin with a Phalanx-stack on T44, which ended about as gruesomely as you'd expect (Saladin's death on T83 would have broken the AI Survivor record, and he was honestly very lucky to survive that long). Meanwhile, Qin's sole win (Game 12) saw him roll up Gandhi and Darius en route to a cruisy domination win; if you'd only seen that game, you'd probably think he was a monster on this map! I'm not sure what was so weird about that one. Finally, Saladin's two wins each saw him take land early on from Alexander (a culture flip in Game 7, and an early war in Game 17) and snowball that into a win - Darius also had to die for this possibility, though. In general, all three of these leaders fell in the 'okay' category. They weren't total pushovers like Gandhi, but they were all also clearly unlikely to win, requiring things to go pretty much perfectly.
As usual, a final tidbit: in this case, I want to focus on religion. Who founded the religions here was massively variable, with everyone except Alexander having a decent chance (even he did on one occasion!). This led to wildly different worlds each game - I saw everything, from Qin's Taoism spreading to literally everyone (RIP Saladin ), to Darius bum-rushing a stone-age Gandhi (Game 11), to 5 different leaders following their own faiths. It usually only had a moderate impact on the game, less than I'd have expected (for example, Alexander didn't care if he'd converted to Islam - he just attacked Saladin anyway), but it was interesting to note. The biggest impact was simply that the leaders who did found religions outperformed those who did not, as the cultural benefits were on full display. I mainly bring this up to say: we cannot simply assume that the 'religious' or Mysticism leaders will grab the religions like we often do, as it happened far less than expected here. That might be worth considering for future games.
With that said, here is a look at the individual leaders' performances:
Darius of Persia
Wars Declared: 38
Wars Declared Upon: 40
Survival Percentage: 70%
Finishes: 10 Firsts, 1 Second (52 points)
Overall Score: 66 points
Darius was clearly the strongest leader on this map, but it wasn't for lack of trying! I've discussed at length already how he was basically gifted the win, so here I thought I would discuss the different ways he managed to bungle it. Occasionally, another leader would snowball out of control, most notably Alex in Game 1 - this was not Darius' fault. It was rare though, and he usually displayed shocking incompetence here. Darius expanded really badly (he got the most cities through his silly start, but early on he struggled despite his favourable resources), mostly because he usually neglected culture. The games where Darius aimed for a religion were some of his best (e.g. Game 2), but they were rare, and it felt like watching Genghis Khan sometimes. Darius was also far too peaceful for his own good. Occasionally, he put his foot down and pushed to victory (Game 19), but so often, he would allow another leader to become competitive through inactivity as they conquered 1-2 neighbours. Now, Darius' start was so ridiculous that he usually won anyway, but in normal circumstances this would have been crippling for the Persian leader. Finally, even when he did fight wars, Darius was very bad at it from tech parity, losing wars repeatedly to Qin in particular (not exactly a God of War himself). Now, with all that said, Darius did win at least half the time, and it was through his strengths. He teched very well (well, the gold and floodplains probably helped, but I digress ), and was semi-consistently able to kill a neighbour using that tech lead. If the game stagnated, Darius usually pulled out a win, which is how his programming is designed. It's just that this map also covered for Darius' massive weaknesses, and frankly after watching these games, I'd be very surprised if he got to the championship again anytime soon.
Pacal of the Mayans
Wars Declared: 31
Wars Declared Upon: 24
Survival Percentage: 55%
Finishes: 6 Firsts, 1 Second (32 points)
Overall Score: 40 points
Pacal played second fiddle to Darius, and was a lot more impressive in doing so. I didn't really cover this above, but Pacal's starting position was woeful, surrounded by jungle and enemies, and forced to expand northwards (away from all neighbours) to settle anything worth having. If Pacal was attacked early on, it was usually a disaster, and this was responsible for several early exits (Games 2, 3, 4, and 13). This was much in line with what we saw on livestream (even if Gandhi had perhaps been slightly lucky to do the killing ), and after Game 4 I thought Pacal was practically dead weight here. I was wrong though - Pacal came back to win a third of the games, and was clearly the only leader who could remotely compete with Darius consistently. This was entirely because of his Financial trait, as Pacal could keep up with the Persians in tech without the massive land advantage that the other leaders needed - he just needed a small one instead. Pacal's game were the most variable here, as his central position did create a lot of disaster scenarios. He also had a very variable landgrab, particularly the northern section of land between Pacal and Saladin - one of these leaders advanced in nearly every game, but they literally never advanced together, with securing that northern peninsula being hugely important for either leader's success. Pacal started more wars than he faced, and did far more than I expected to break out of his central deathtrap. His most common and most successful target was Gandhi, who typically collapsed like a house of cards; Pacal's less successful attacks came against Saladin and Qin, who could fight back much better. In the lategame, Pacal was never able to fight Darius in a 1v1 war, but thanks to his peaceweight, could semi-consistently bring in Alex, Saladin, or Qin as an ally. As usual for Pacal, it felt as though his real traits were Financial/Peaceweight 2, and as usual for Pacal, that was enough to scam quite a few wins through diplomacy. He faced the least defensive wars in the game, despite his map position, which is a clear sign of how much his peaceweight helps him. But in sum, this was a pretty impressive result for a leader in a grim central position, and Pacal definitely played the best games here. It just turns out that most of the time, that wasn't enough against the Persian juggernaut.
Saladin of Arabia
Wars Declared: 43
Wars Declared Upon: 36
Survival Percentage: 60%
Finishes: 2 Firsts, 7 Seconds (24 points)
Overall Score: 39 points
Saladin graded out about equal to Pacal, although it was via a totally different method of accumulating points. Where Pacal was able to sneak quite a few wins, but otherwise failed quite a lot, Saladin was a far more consistent leader. He rarely won, and both of his wins required pretty unusual circumstances: in Game 7, he flipped Alex's third city around T50, and in Game 17, he slowly conquered an entirely Taoist world, which somehow completely failed to unite against him. However, Saladin was consistently pretty strong, and good at sticking around. Very often, Pacal or Darius would be finishing their spaceship as an Industrial (or even Renaissance) Saladin finally finished off his initial opponent. Sometimes, Industrial/Renaissance Saladin would just have sat on his hands all game, and still manage to come second! Thanks to his position in the turn order, Saladin was the leader most likely to found a religion, and he typically expanded very well thanks to having some very strong land near his capital. However, Medina was settled towards Alexander and pressured the Greeks culturally, which created a bitter feud across most games. Occasionally, Alexander would get the best of this conflict, such as in Games 14 and especially Game 1. It could also happen when Pacal jumped in from the other side, which was rare but plausible (such as Game 15). More often though, Saladin would win, but far too slowly to be competitive, as he was left technologically in the dust. The other interesting thing about this Greco-Arabian relationship was its one-sided nature. Alexander attacked Saladin in nearly every game, but was usually following the Arabian Islamic faith, which meant that Saladin rarely attacked him in turn. Even when Saladin had been attacked once or twice(!) by the Greeks, he was more likely to march across the map to attack Gandhi or Qin than to seek revenge against Alexander. This was very lucky for Alex, who in most games would have been squashed by his stronger northern neighbour. Overall, Saladin played out a series of solid but unremarkable games, pretty much what we've come to expect from the Arabian leader. He's proven over time that he's decent, but has never seemed able to make the jump from decent to powerful; it was no different here.
Qin Shi Huang of China
Wars Declared: 30
Wars Declared Upon: 46
Survival Percentage: 35%
Finishes: 1 First, 5 Seconds (15 points)
Overall Score: 25 points
Qin was pretty bland here, if I'm being honest. He fluctuated somewhat in his performances, particularly in his early expansion, which was sometimes very strong (competing with Darius for best in the game) and sometimes appalling (leaving him the clear runt). The best I could figure out was that when Qin researched Mysticism early he was a lot stronger, but that didn't account for all the fluctuations in performance that I observed. Qin was basically a weaker version of Saladin, probably due to his central position. He was able to fluke out one win thanks to a strong landgrab, IND Stonehenge, and methodically crushing his way through the rest of the map. He was unlucky not to win another (Game 18), when he conquered everyone else but barely lost to a Mayan spaceship. He was able to score quite a few seconds, tagging behind Pacal, Alex, or Saladin (never Darius, whose performance was inversely tied to Qin's). His tech rate was terrible, at least compared to Darius and Pacal's. The biggest difference was his relationship with Darius, as they fought in most games, and could never both do well. Obviously Darius won most of these fights, but honestly given all of his intrinsic advantages, and the fact that he was often helped by Alexander, Gandhi, or Saladin, it would have been a very bad sign if he hadn't. Qin's tech was quite bad, like Saladin, and in both cases it was probably because they started without any major rivers. I'm not really sure what to say for Qin, to be honest. He performed adequately in the games where he wasn't dogpiled, but his central position made that outcome pretty unlikely. It was pretty much the most average performance I've ever seen from an AI, and I'm glad Qin didn't make it to the Championship to do nothing again.
Alexander of Greece
Wars Declared: 66
Wars Declared Upon: 25
Survival Percentage: 40%
Finishes: 1 First, 2 Seconds (9 points)
Overall Score: 24 points
Unlike Qin, Alexander was at least interesting. He was very competitive in the first 6 games, scoring as many points as Darius did, and placing in half of the games. This included the never-repeated fluke in Game 1, but also two strong seconds to Pacal in Games 5 and 6. I thought Alexander could actually be quite strong on this map despite the cultural pressure he was under from Arabia, but then he failed to place at all in the final 14 games, and fell off completely. Alexander was a rabid warmonger here, declaring the most offensive wars by far. In particular, he attacked Saladin in basically every game, and was then very lucky to receive a free pass as Saladin rarely sought revenge. Unlike Qin's bad games though, it was very clear what hurt Alexander, which was a refusal to construct cultural buildings until T75 or later in nearly every game. This allowed Saladin to steal the first-ring tiles from Alexander's cities, and though this drew Alex's ire, he was usually so weak by this point that it didn't matter. In the rare cases where Alex did prioritise culture he was actually pretty strong (most notably Game 9 where he founded the Meditation religion), but it was so rare that it has to be considered an outlier. Other than Saladin, Alexander was also pretty likely to attack Qin, Pacal, or Gandhi, as only Darius was safe from an early Greek deathmarch. Alex's performance was pretty binary, in the sense that he either got allies and was the junior partner, scoring kills without advancing (Games 13 and 16), or he had to fight 1v1 and died miserably (Games 17 and 18). He was very unlikely to advance to the Championship, and his mediocre performance on livestream was pretty indicative of his chances here. It was another warmonger who couldn't expand or tech well enough to make consistent conquests, and we've learnt many times by now that that's a recipe for failure.
Gandhi of India
Wars Declared: 11
Wars Declared Upon: 53
Survival Percentage: 30%
Finishes: 0 Firsts, 4 Seconds (8 points)
Overall Score: 8 points
Oof. Poor Gandhi. The Indian leader had a rough time of it here, being attacked by virtually everyone, and with a truly terrible start that made even Pacal's look amazing. Really, in retrospect, it's pretty amazing that Gandhi managed to attack and conquer Pacal on livestream, because he didn't score even a single kill here! That was pretty indicative of his strength as a leader, as Gandhi never once managed to even win a war, let alone conquer someone completely. At best, he would stalemate his attackers, only for someone else to jump in later on and finish the job. Honestly, Gandhi doesn't feature too much in my notes for these games, because he died so unremarkably in so many of them. He couldn't expand well because his start was terrible, and he couldn't compete diplomatically because everyone hated his guts. In the games where he did survive to the end, it was as an also-ran, with Gandhi literally never competing for a win even in his four second places. He pretty much only advanced in those games because everyone was dead. Perhaps Gandhi's primary contribution to this game was in feeding so many victories to Darius and Pacal, who each benefited from Gandhi's severe weakness (Darius by settling so much land, and Pacal by getting several easy conquests). Overall, while Gandhi was apparently slightly more likely to advance to the Championship in these games than Alex, he was obviously the weakest leader, and it was yet another year of failed Playoffs for him. This is really starting to become a curse at this point, poor guy.
This AH revealed the result we experienced on livestream to be a very standard outcome, receiving the most likely winner, second place, and a very reasonable first to die. Even the number of wars was about right, as was the finishing date, which was very early on average compared to other AHs I've run. It was nice to have confirmation that this game went about as expected, although it makes it much harder to write anything interesting about it. I suppose the only thing we missed was the scenario that Pacal won instead of Darius, but that was less likely, and frankly Pacal was very lucky to be featured in this game anyway - good riddance.
For Fantasy purposes, this was the game that locked in my victory, and that turns out to have been somewhere between lucky and expected. Darius was the most likely leader to win, yes, but he wasn't exactly guaranteed, and his place here required me to win the earlier coinflip of Game 6 anyway. We could have been looking at a very different Fantasy championship if Pacal or Qin had advanced (or especially won!). Henrik was semi-unlucky for Pacal not to score bigger (though as mentioned, Pacal should not have been here at all), while Cuthraxys definitely missed out on Qin, who not only should have won his opener, but had a decent chance to advance here as well! It's hard to talk about missed points in a playoff game, since this all relied on opening round chance as well, but those are probably the main points. I hope you enjoyed this Alternate History, and the confirmation that no, not all of the results we experienced this year were weird flukes.
Cheers ~ Amicalola