Game Two 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. Game Two saw Suryavarman shake off a modest early game and snowball into runaway status, aided by Brennus right up until the Khmer plunged the dagger into his back at the conclusion of the match. 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:
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 alternate histories suggest that the actual Game Two was a typical result, although there was more variability in this game than in some of the other matches. Suryavarman was the winner of Game Two and he did indeed emerge as the victor more than anyone else, taking home nine first place finishes out of the twenty replays that I watched. Similarly, Louis came in second place in the Livestreamed game and he also had more runner up finishes than any of the other AI leaders. The second place finishes were evenly split across a whole bunch of different competitors, however, and there was no clear favorite in that category. Once place where Game Two wasn't typical at all was seeing Catherine as First to Die. This didn't happen often and she survived to the finish in 75% of the alternate histories. Sitting Bull and a slightly surprising Wang Kon were the most likely AIs to be eliminated first although once again there was no true runaway favorite in that category either. And there was even one game where *NO ONE* died despite lasting for 379 turns - what the heck was going on there?!
If I had to summarize this map in one sentence, it would be that the low peace weight leaders generally dominated the action while the high peace weight leaders tended to become marginalized or eliminated. All of the low peace weight leaders had low elimination percentages and all of them outscored the high peace weight leaders in both finish points and kill points. There were essentially three tiers of AI leaders: Suryavarman/Louis who won 75% of the games between the two of them, Brennus/Catherine who typically survived to the end of each match but rarely won, and then the high peace weight trio of Sitting Bull/Wang Kon/Roosevelt who were usually eliminated in each game. The fact that there were 4 low peace weight leaders as compared to 3 high peace weight leaders was a big factor in terms of how this map shook out. However, the personalities of each AI leader also played a significant role. The high peace weight leaders were less aggressive as a group (as one would expect) and often let chances to expand and become stronger slip through their fingers. Roosevelt in particular was sometimes in a strong position coming out of the landgrab phase and would sit back doing nothing while his allies were picked off one by one. Suryavarman and Brennus and Catherine were more likely to keep snowballing themselves ahead once they took an initial lead, thereby preventing anyone else from catching up.
There was a lot of warring on this map but those conflicts frequently ended up in stalemates. I noticed that there were a lot of First to Die eliminations that took place at late dates, often occurring after Turn 200 and not occurring at all in the wacky Game #16. I think that this was due in part to having two Protective leaders (Sitting Bull and Wang Kon) among the high peace weight leaders which allowed them to drag out conflicts for extended periods of time. We experienced this in the actual Game Two on the western side of the map where Louis and Wang Kon and Roosevelt were stuck in an endless violent stalemate with little to no territory changing hands. Generally speaking there needed to be a 2 vs 1 situation for conquests to start taking place and sometimes that could take a long time to develop. For all of the fighting this was a series of alternate histories with a low number of kills, only 63 of them in total despite having an extra AI competitor (six instead of seven). I think that this was due to the fact that there was no clear punching bag to get knocked out quickly and then snowball one AI leader into runaway status as we've seen in some of the other alternate histories.
The victories that took place were characterized by a lack of Domination finishes, only two of them across these twenty games. The actual Game Two was atypical in that respect. Part of the issue here was Louis who chased after a Cultural ending in virtually every game. Even his Domination win in Game #20 was effectively a backdoor Cultural win since he stopped researching and began running the culture slider despite having 50% of the world's land area under his control. There was also a shocking five different Diplomatic victories and this happened too many times to be a coincidence. I'm guessing that this was due to the lack of a clearly dominant AI leader in most games combined together with all of the low peace weight leaders surviving to cast votes in the United Nations. They liked one another enough to vote for a winner and it kept happening in unexpected fashion. Even though Suryavarman won nine different times, he rarely ran away with things in the way that we tend to see winners do. A lot of these were close, contested victories where he pulled out a win against tough competition. Unlike Louis, who basically could only win by culture, Suryavarman won all four available victory types and that's a sign that he was pretty clearly the best AI leader for this scenario. The other five leaders could only win when all of the cards lined up just right for them and it was an uncommon result to see them come out on top. Catherine and Brennus had some strong performances but it was hard for either of these leaders to translate them into an actual victory.
Now for a look at the individual leaders:
Suryavarman of the Khmer
Wars Declared: 56
Wars Declared Upon: 34
Survival Percentage: 65%
Finishes: 9 Firsts, 2 Seconds (49 points)
Overall Score: 57 points
Suryavarman ended up winning the most alternate history matches by a clear margin and therefore backed up his victory in the actual Game Two. As mentioned above, he was not especially dominant in these wins as demonstrated by the ending screenshots from Game #5 or Game #10 or Game #13. But the main thing that matters in this competition is crossing the finish line first and Suryavarman did that repeatedly across these alternate histories. He had an amazing hot streak from Game #8 to Game #14 where he rattled off victories in six of the seven games that I watched. That reward didn't come without risk, however, as Suryavarman was also eliminated more often than any of the other low peace weight leaders and was even First to Die on three different occasions. This was a case of "Austria in Avalon Hill's Diplomacy", a nation that has both the highest chance to win and also the highest chance to get knocked out quickly. For all of Suryavarman's success, it wasn't unusual to see the Khmer in a weak or bullied state because they were highly susceptible to getting stuck in a 1 vs 2 situation. When the diplomacy broke in Suryavarman's favor he could snowball ahead of the rest of the field with ease, but if it broke in the opposite direction he was often in serious danger.
Part of the reason why Suryavarman won the most games was due to having many different paths to success. He was frequently able to take advantage of Sitting Bull's horrible diplomatic situation and participate in a dogpile of the Native Americans. This was what happened in the actual Game Two and it was repeated in many of the alternate histories. Another common path involved expanding through Wang Kon who was also a frequent target of the low peace weight leaders. This allowed Suryavarman to pair together with Louis or sometimes with Brennus as the Celts fought a large number of cross-map wars in the alternate histories. Roosevelt was a less common victim of Khmer aggression, at least out of the gate, although America provided yet another path for potential expansion. None of the other low peace weight leaders had as many chances to snowball ahead by devouring their rivals. Perhaps the most surprising thing about Suryavarman was his lack of kills and I honestly don't know why he only ended up with 8 of them. A lot of this was due to bad RNG luck in terms of getting the final city kill against some of his rivals but I think it also speaks to how Suryavarman's victories generally weren't taking place from a position where he was the runaway. His finish points suggest that Suryavarman was by far the top leader on this map but the lack of kills resulted in the Khmer leader only finishing modestly ahead of Louis. In fact, Louis would have taken the top spot if Game #10 had gone differently and that was a match where Suryavarman's spaceship arrived with Louis' third almost-legendary city sitting at 48k culture. Sometimes the margin of victory can be a narrow thing indeed.
Louis XIV of France
Wars Declared: 53
Wars Declared Upon: 40
Survival Percentage: 85%
Finishes: 6 Firsts, 5 Seconds (40 points)
Overall Score: 52 points
Louis was blessed with an exceedingly strong starting position on this map and a generally favorable diplomatic environment. He did take advantage of these opportunities by collecting six victories and a bunch of second place finishes, therefore having better odds to move on to the playoff round than anyone else. The French leader also suffered the fewest eliminations with only three of them across the twenty alternate histories. However, it still feels to me as though Louis left a lot on the table in this game given his starting position and stronger AI leaders would have been able to do more with this setup. The biggest problem for Louis was that he spent too much time building wonders. This caused him to get out-expanded on the western side of the map and therefore he often found himself in a weaker position than he should have been. There was no reason why he couldn't have crushed his usually weak AI competitors in Wang Kon and Roosevelt considering how Louis had an amazing capital and oodles of happiness resources in his core. Instead, Louis tended to get stuck in lengthy unproductive wars that went nowhere with one or both of his neighbors. This was what we watched in the actual Game Two and I saw it play out repeatedly over and over again in the alternate histories. A better AI leader would have been winning these wars instead of stalling out for long turns on end.
The six victories for Louis are also somewhat deceptive because he won five of those six games by culture. Now a win is a win, don't get me wrong, but typically the winning AI leader is also tops on the scoreboard and that was not the case for Louis in a number of these games. He was behind Sitting Bull when he won in Game #3 and at best tied for the lead in Game #6. Louis did have some strong performances like in Game #17 and Game #20 but they tended to be the exception more than the rule. His best result took place in the final Game #20 where Louis solo conquered Wang Kon at a very early date and then proceeded to consume Roosevelt and Suryavarman. He had 50% land area shortly after Turn 200 and could have walked with ease to any victory condition imaginable. So what did the French leader choose to do? Sit back passively and start running the culture slider. He did win of course, eventually, but this little vignette demonstrated how Louis lacks the killer instict of the top AI leaders. It was good enough to be one of the top leaders in this scenario but the map isn't always as kind as it was for this game and Louis could have, should have, been more dominant than he was here.
Brennus of the Celts
Wars Declared: 49
Wars Declared Upon: 19
Survival Percentage: 75%
Finishes: 1 First, 4 Seconds (13 points)
Overall Score: 33 points
By way of contrast, Brennus probably maximized his opportunities from a much weaker starting position. The Celts had a corner start and didn't have many good opportunities for expansion on this map. The two AI leaders that bordered Brennus were Suryavarman and Catherine, two other low peace weight leaders that formed natural allies with the Celts. None of the high peace weight leaders directly bordered Brennus and he rarely had the same kind of easy chances to snowball enjoyed by many of the other leaders. Brennus typically took the only option available to him by attempting to fight his way out of the corner and he had more success at this plan than I expected to see. Brennus was rarely eliminated and he became quite strong in a number of these matches such as Game #15 and Game #19. His best opportunities came from conquering either Sitting Bull or Catherine, the latter of which we experienced in the actual Game Two. But Brennus also engaged in a lot of cross-map wars and attacked Wang Kon in particular repeatedly. Sometimes these wars stalled out and were completely ineffective (the usual fate of long distance fighting) but there were also times where the Celts managed to pick up a legitimate second core of cities off in the west. As I mentioned before, Brennus was more successful in these conflicts than he had any real right to be.
I also need to call attention to the fact that Brennus wound up with the most kills in this game by a wide margin. He had double the kills of anyone else until Louis managed to bag a trio of AI rivals in the last match. It's not fully clear to me why this was the case, as Brennus wasn't declaring war more often than the other low peace weight AI leaders and it's not as though he was running away with a lot of these games. He just seemed to have a weird knack for getting the last hit on AI capital cities and swiping the kill credit for himself. I think that this is probably a result of small sample size bias and if we were to run 100 of these games instead of 20 we would see this strange quirk in the numbers disappear. Brennus had 11 kills in matches where he didn't finish in first or second place and that's really unusual. By way of comparison, every other AI leader combined had 8 such kills. For whatever reason, he was in the right place at the right time over and over again. (Ironically we had the opposite result in the actual Game Two, where Suryavarman seemed to land every 50/50 chance to capture a city when the two of them were at war against mutual opponents.) Anyway, Brennus was clearly a lower tier than the Suryavarman/Louis combo in this game but nonetheless still managed to be effective at warring and eliminating his competitors. It was a good performance from a subpar starting location.
Catherine of Russia
Wars Declared: 57
Wars Declared Upon: 29
Survival Percentage: 75%
Finishes: 2 Firsts, 4 Seconds (18 points)
Overall Score: 27 points
Catherine's situation in this game was pretty similar to Brennus' situation and it's not surprising that they wound up with a similar set of outcomes. Catherine had a better chance to expand outwards thanks to her Imperialistic trait and the presence of diplomatically unpopular Sitting Bull next door. However, Catherine's land was weaker than the Celts and she had what was very likely the poorest local area in terms of land quality on the map. There was a lot of tundra and ice in the region and Catherine tended to struggle in terms of breaking out of this area. She didn't have a whole lot to work with and this was responsible for several early eliminations including three First to Die exits. These were the games where Sitting Bull and Brennus teamed up together to attack Russia and left Cathy with essentially no options. We watched this happen in the actual Game Two and the same scenario repeated itself in Game #5 and Game #10. Catherine's weak start necessitated the diplomacy breaking in her favor and that didn't always materialize.
On the other hand, Catherine's low peace weight was a real advantage in this world and she was often able to survive in third or fourth place while avoiding elimination. The Game Two that we watched was not typical in this regard as Catherine was only eliminated 25% of the time. There were also several games where Catherine was able to break out of her corner and start rampaging across the map, and although that was a rarity on this map, it was a terrible sign for the other leaders if Cathy could ever get going. In Game #4 she broke through a long stalemate to conquer the entire eastern part of the continent for an easy victory while in Game #7 she was narrowly beaten to space by an unusually strong Wang Kon. Cathy started more wars than anyone else (if by a narrow margin) and she was unquestionably one of the most aggressive leaders on the map. This was a tough starting position and she managed to wring a lot of value out of it, better than I think most of the other AI leaders would have done. I was expecting her to perform worse in these alternate histories than what the data ended up showing. It's been a rough few seasons for Catherine in AI Survivor but I think she's still one of the better leaders based on what I saw here.
Sitting Bull of the Native Americans
Wars Declared: 27
Wars Declared Upon: 70
Survival Percentage: 25%
Finishes: 1 First, 3 Seconds (11 points)
Overall Score: 17 points
We tend to view Sitting Bull as a joke leader because he's been so awful in past seasons of AI Survivor, the only leader never to score a point or make it to the Wildcard round. However, Sitting Bull was clearly the strongest of the three high peace weight leaders on this map and he was legitimately strong in some of these matches. And yes, Sitting Bull not only was the runner up on three different occassions, he really did win one of the alternate histories! It's true, look at the screenshot for Game #11! Sitting Bull was fortunate to have a good starting position with food resources that perfectly matched his starting techs and a copper resource for early metal-based units. Most AI leaders who rolled his starting position would have been the favorite to win this game. Unfortunately that start had to be played out by Sitting Bull and this guy really does stink up the joint in terms of his AI personality. To his credit, Sitting Bull was genuinely strong in several different alternate histories, including Game #11 which he won along with Game #3 and Game #6. Sitting Bull was the runner up in three of the first six games that I watched and I was wondering what in the world was going on. Then I played the rest of the matches and Sitting Bull was eliminated in 14 out of 16 games so... yeah, about what you would expect.
Sitting Bull had the best starting position and the best prospects for expansion amongst the three high peace weight AI leaders. The gigantic problem for Sitting Bull was his unpopularity with the low peace weight leaders, all of whom pretty much hated his guts from start to finish. The Native Americans bordered Suryavarman and Catherine and were very close to Brennus, with all three of these leaders frequently going to war with Sitting Bull. It was rare to find a game where Sitting Bull wasn't being double or triple teamed by his neighbors. In fact, Sitting Bull was invaded 70 times (!) across these twenty alternate histories, the most that I've recorded to date in running these exercises, including an astonishing nine attacks in the bizarre Game #16. It's hard to believe that there was a game where Sitting Bull initiated 0 offensive wars while being attacked 9 times but it really did happen. He even survived to the finish in that match! Anyway, if Sitting Bull could somehow avoid being dogpiled he did pretty well with a victory and three runner up performances across the five games where he survived. But of course he did get dogpiled in almost every game that I ran and that pretty much sealed his fate. Sitting Bull is a poor AI leader and even with favorable terrain at his capital he wasn't able to accomplish a whole lot.
Wang Kon of Korea
Wars Declared: 33
Wars Declared Upon: 53
Survival Percentage: 30%
Finishes: 1 First, 1 Second (7 points)
Overall Score: 10 points
Wang Kon was an ineffective leader on this map and didn't have many opportunities for his typical trolling antics. He had the northwest starting position and struggled mightily to expand out of that corner. Wang Kon's high peace weight made him a natural target for Louis and their nations clashed over and over again in warfare across these alternate histories. Most of the time these wars tended to stall out without making progress in either direction, at least until another AI leader intervened, and it was far more likely to see another low peace weight leader team up against Wang Kon than the other way around. The Koreans were hurt in this respect by the stagnant gameplay of Roosevelt; the Americans were the natural allies for Wang Kon but Roosevelt initiated the fewest offensive wars of anyone in this game and often did nothing as Wang was wiped from the map. As a result, Wang Kon ended up suffering the First to Die fate more than anyone else at six times in all across these games. He needed help to survive and he received very little of it.
If it wasn't bad enough having to border Suryavarman and Louis, the two strongest leaders on the board, Wang Kon also found himself suffering from cross-map attacks over and over again. Brennus in particular seemed to like going after Wang Kon despite having no cities near the Koreans. I suspect that this was due to their nations founding and then practicing two different religions on top of their peace weight disparity. These seemingly nonsensical invasions from Brennus played a major role in dragging down Wang Kon as he was faced with multiple attackers in game after game. He was almost never one of the strongest AIs and didn't lead in score even in the single game that the Koreans managed to win. Wang Kon's second place finish in Game #14 was also a distant result where he wasn't particularly close to Suryavarman. There just wasn't much for Wang Kon to work with on this map and he was largely irrelevant to the outcome of most matches. He had the two earliest First to Die dates (Turn 115 and Turn 123) and largely served as a speedbump for his more successful rivals.
Roosevelt of America
Wars Declared: 27
Wars Declared Upon: 57
Survival Percentage: 30%
Finishes: 0 Firsts, 1 Second (2 points)
Overall Score: 7 points
Roosevelt was a pacifist on this map and played out a series of inert games that accomplished virtually nothing. Roosevelt was only First to Die in a single game and his survival rate was no worse than the other high peace weight leaders. Considering his exposed central starting position things could have been a lot worse for the Americans. They could have been a lot better too, however, as Roosevelt was the only one of the seven leaders who failed to win a game. He didn't seem to have any clear path to a victory either as Roosevelt never emerged as the dominant power in a single game. His best result was the oddball Game #11 where all of the high peace weight leaders survived to the finish but even in that match Roosevelt was well behind Sitting Bull. He wasn't even close to winning in any of the other five games where he survived to the finish and that included Game #7 where Roosevelt survived with a single city only because Cathy couldn't reach the American capital through Korean borders. I suppose that Roosevelt's best path to victory would have been via space but there always seemed to be other AI leaders who could pull that off better. Amongst the high peace weight leaders, Wang Kon's Financial trait gave him the edge in pure economy while Sitting Bull was the only one of the three who ever seemed to be able to conquer additional territory. Roosevelt didn't seem to do anything better than his rivals.
A typical Roosevelt game would see him survive longer than the other high peace weight leaders thanks to turtling up in the center of the map. There were some games where he was reasonably strong in the Turn 150-200 period thanks to staying at peace while other AI leaders were engaged in fighting. However, once some of the weaker AI leaders were eliminated from the map, sooner or later the successful conquerers would turn their attention to Roosevelt and he wouldn't be strong enough to fight off opponents who had snowballed via additional cities and territory. He was the second or third AI leader to perish in game after game. Roosevelt's territory was reasonably strong in this game and there's no reason why he couldn't have put up a better showing here. This seems to be another AI personality that's a real dud, someone who just tries to stay peaceful and build wonders. Unfortunately for Roosevelt, he doesn't have strong leader traits and he's tied to one of the worst civilizations in the game. He tries to run the Mansa Musa playbook but he can't do it effectively enough to translate it into victories.
Once again, the Game Two that we watched play out on Livestream was pretty typical for this scenario. We ended up hitting the most likely winner and runner up according to these alternate histories, although the second place category was pretty much a toss-up that could have gone in a number of different directions. First to Die was the only category where we had a somewhat unlikely result, with Wang Kon or Sitting Bull being far more likely to depart the stage first. We ended up having a typical finishing date and number of wars in Game Two, with this being a map that produced a series of late finishing dates. The average and median finishing date were both around Turn 350 which is quite a bit later than what we usually see for a game of AI Survivor. This scenario seemed to produce a lot of stalemated conflicts and lacked leaders who were strong from an economic perspective, thus drawing these matches out to slower ending dates. This won't be remembered as one of the more famous or interesting AI Survivor games but it was still fun to watch these alternate histories and see how things could have played out. Thanks as always for reading along!