Civ4 AI Survivor Season 8: Playoff Game Three Alternate Histories


Introduction

Playoff Game Three Alternate Histories Spreadsheet

One of the recurring features of past seasons of AI Survivor have been our "alternate histories", running additional iterations on the same maps to see if the same events would play out again. Playoff Game Three turned into a dominant Churchill victory after he successfully united northern and southern England, followed by running over the rest of the field. 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. First I'll post the resulting data and then discuss some of the findings in more detail. Keep in mind that everything we discuss in these alternate histories is map-specific: it pertains to these leaders with these starting positions in this game. As Wyatan mentioned, an AI leader could be a powerful figure on this particular map while still being a weak leader in more general terms. Now on to the results:

Season Eight Playoff Game Three

Game One | Game Two | Game Three | Game Four | Game Five

Game Six | Game Seven | Game Eight | Game Nine | Game Ten

Game Eleven | Game Twelve | Game Thirteen | Game Fourteen | Game Fifteen

Game Sixteen | Game Seventeen | Game Eighteen | Game Nineteen | Game Twenty



(Note : "A" column tracks the number of war declarations initiated by the AI, "D" the number of times the AI is declared upon, "F" the points for finish ranking, and "K" the number of kills.)

The pregame thinking for Playoff Game Three was that this was a mostly peaceful field of leaders, and that the high peaceweight leaders on the eastern side of the map would work together to eliminate Mehmed/Qin before racing once another to space. Thus Victoria was the favorite to win the picking contest and Spaceship was overwhelmingly the most likely victory condition chosen. However, the game that we watched on Livestream didn't follow this pattern at all, with Pericles invading Sitting Bull while Churchill spent the whole early game battering Victoria. That allowed Mehmed to become one of the top AI leaders before eventually running afoul of Churchill in the lategame, followed by his collapse backdooring Qin into an undeserving Championship spot. I was interested enough in this game that I used the one week break before the Championship to run the initial five alternate histories on Livestream, pulling back the curtain to show the viewers what the process looks like to do these repeat map playthroughs. Then six months went by before I could get around to finishing the remaining 15 alternate histories in what had to have been the longest break between starting and finishing one of these projects, heh.

So what did replaying the map all of those additional times reveal? The biggest surprise for me was the relevation that this setup essentially had two co-favorites, Churchill and Sitting Bull. They wound up with eerily similar results in the data: each had seven victories, each had four Runner Up finishes, and Sitting Bull wound up with 20 kills as compared with 19 kills for Churchill. It's hard to get much closer than that and 20 map replays clearly weren't enough to determine which had the edge. Thinking of them as co-favorites is the best we can do with the data on hand. What was interesting was how the two of them ended up in the same place despite taking two distinct paths to get there. Churchill had the safest spot on the map and benefited enormously from favorable diplomacy. He was rarely attacked by anyone and never dogpiled, enjoying the gigantic benefit of avoiding the 1 vs 2 conflicts that always seem to torpedo the AI in Civ4. However, he had a more cramped position and a lot of desert in the area around his capital, leading to slower starts on average. Basically Churchill couldn't win based on solely his own portion of the map; he needed to conquer someone else and snowball ahead, but he could pick and choose his spots for aggression better than anyone else.

By way of contrast, Sitting Bull had what was objectively the best starting position on the map. The Native American leader had a beautiful floodplains river valley with a gold resource present along with enough space to expand in every direction simultaneously. He was frequently the top leader during the landgrab portion of the gameplay and ran out to a bunch of early score leads. Unfortunately for Sitting Bull, his central starting position also opened him up to aggression from all sides and he was emphatically not safe from supposedly peaceful leaders like Pericles and Churchill. Sitting Bull was attacked more often than anyone else and it wasn't particularly close, getting dogpiled repeatedly from east, north, and west simultaneously. There wasn't much he could do in situations like Game #12 where he was invaded four times by three different AI neighbors before Turn 150. Sitting Bull also didn't help himself by repeatedly founding his own religion and then failing to spread it, with the gold resource being a bit of a poison pill in that regard. He often had a poor diplomatic situation due to being religiously isolated, drawing hatred from Mehmed/Qin due to peaceweight concerns as well as religious friction with Pericles/Victoria/Churchill.

But if Sitting Bull could dodge enemy attacks and limit his wars to one opponent at a time, he usually ended up winning the game. It bears repeating that his starting position was genuinely excellent and Sitting Bull usually out-expanded the rest of the group. He was also very well positioned to expand further through conquest, as both Mehmed and Qin were unpopular leaders as well and any dogpile against the western AIs would inevitably see Sitting Bull reaping most of the spoils. There were several games where Churchill or Victoria went racing off to attack Mehmed, only for Sitting Bull to jump in as well and claim nearly all of the territory. In a setup where the other AI leaders tended to have poor economies and the overall pace of the game was slow, Sitting Bull's territorial edge translated into a lot of victories. The survival rates tell the story: Sitting Bull survived to the finish in 12 games and claimed a top two spot in 11 of those 12. Churchill had a much higher 85% survival rate but also six different games where he survived but didn't advance. Sitting Bull clearly had the higher difficulty setup while Churchill was playing on easy mode.

Most of the other leaders found themselves in a situation where it was possible to win if everything broke in their favor while being unable to do so with any consistency. Pericles had a whole bunch of Runner Up finishes while almost never being in a position to win outright. He was simply too cramped on territory and had no obvious way to break out, nor was he economically strong enough to go for the Cultural path. His only victory came in Game #16 where he had a very unlikely solo conquest of Victoria and snowballed from there. Mehmed and Victoria both managed more victories while also having terrible survival rates as they usually didn't make it to the finish line. Mehmed needed to win via the military route and he pulled this off on a trio of occasions, all three times when the eastern AIs were stuck in long unproductive wars which gave Mehmed the time and space to win repeated 1 vs 1 conflicts. Victoria needed to do something similar and avoid early aggression from Pericles and Churchill which almost never happened. There were exactly two occasions where she managed a full conquest of Pericles and both times she won outright. Usually though, she lacked enough territory to be truly competitive and was run over by someone else during the midgame.

The one truly helpless leader in this game was Qin who was never even remotely competitive for a victory. His Runner Up finish in the actual Playoff Game Three was more or less a joke, getting backdoored into advancing by Churchill's late conquest of Mehmed, and the alternate histories fully backed that up as Qin managed only a single second place finish in 20 map replays - which was also a Churchill backdoor! Qin was lucky to be sharing Churchill's religion in that game or he would have been toast again. The Chinese leader expanded poorly on this map and struggled to expand his borders in game after game, leaving him with low population totals and poor numbers on the Demographics screen. Qin was also diplomatically unpopular and couldn't count on Mehmed to be his ally, with the two of them clashing frequently as well as with the eastern AI leaders. This was a rough setup for Qin and his AI personality was nowhere near up to the task of making something out of this start.

Now for a look at the individual leaders:


Sitting Bull of the Native Americans
Wars Declared: 29
Wars Declared Upon: 48
Survival Percentage: 60%
Finishes: 7 Firsts, 4 Seconds (43 points)
Kills: 20
Overall Score: 63 points

This was legitimately one of the most impressive performances I can remember seeing from Sitting Standing Bull. Having run a whole bunch of these alternate histories, I can attest that it's extraordinarily rare for the most-invaded AI leader to grade out with the highest score, even if it was a statistical tie with Churchill. Sitting Bull was attacked 48 times compared with a mere 11 invasions faced by Churchill, a true case of one leader playing the game on Deity setting while the other was on Chieftain. If Sitting Bull has a unique niche as a leader, it's probably being really good at mounting a hopeless defense when outnumbered and facing multiple attackers. He was able to use his Protective trait and Totem Pole-enhanced archers to stall out a number of dire situations long enough for another leader to come to his assistance, something that doesn't work in most games but did have a chance here in a high peaceweight field. With that said though, the weaknesses of Sitting Bull were very much also on display here. His non-Domination victories were painfully slow (including a Turn 413 Cultural win!) and he did a poor job of spreading his religion in the games where he founded one. It's easy to see how he gets left behind when facing leaders with better economic traits, and I suspect that someone like Justinian or Huayna Capac would have won roughly 15 games from this start. Nevertheless, despite the dice being stacked against them from a diplomatic perspective, the Native Americans were a force to be reckoned with on this map.


Churchill of England
Wars Declared: 48
Wars Declared Upon: 11
Survival Percentage: 85%
Finishes: 7 Firsts, 4 Seconds (43 points)
Kills: 19
Overall Score: 62 points

Churchill's success in the alternate histories was easy to explain: he simply didn't get attacked by anyone, facing about 0.5 invasions per game. That included a staggering twelve games where he was never invaded even one time! His corner starting position gave him near-total protection from warlike Mehmed and Qin, neighboring only the peaceful Victoria to the north and the "I have more than enough enemies already" Sitting Bull to the west. Instead it was Churchill who nearly always decided when and where he wanted to launch his aggression, a huge advantage for the AI which struggles to deal with unexpected attacks in Civ4. The most frequent target for Churchill's military turned out to be Victoria and the two of them clashed in almost every game. This was one of the biggest surprises of the map to me as I thought they'd be able to work together, but nope, apparently border tension overcame relatively compatible peaceweights. Absorbing Victoria's cottaged terrain and wonder-heavy cities was typically the biggest springboard to a victory for Churchill and the game we watched on Livestream was absolutely the norm in that respect. In games where Churchill wasn't able to snowball ahead, he still almost always survived to the end by virtue of his safe position. Churchill rarely won by Domination and his victories weren't particularly flashy or achieved quickly, but he had the diplomatic golden ticket in this game and used it effectively.


Pericles of Greece
Wars Declared: 25
Wars Declared Upon: 31
Survival Percentage: 70%
Finishes: 1 First, 7 Seconds (19 points)
Kills: 9
Overall Score: 28 points

Pericles was the least interesting leader in this game as far as I'm concerned. He was just kind of there, lurking around in the background most of the time, hanging out as a second or third tier leader without doing all that much. He had an average amount of offensive and defensive wars, scored an average amount of kills, and so on. Most of the scoring achieved by Pericles came from his seven Runner Up finishes, which in practice was a lot less impressive than it might sound on paper. Sometimes a leader who scores a bunch of second place finishes was in tight competition for the overall victory, only to have bad luck break against them at inopportune moments. That was not the case with Pericles who rarely seemed to break out of his corner of the map; even in games where he was relatively strong, like in Game #5 and Game #17, he was still tagging along in the wake of a runaway Domination winner. The one Pericles victory in Game #16 was the only time that he was truly competitive for first place and he was never a serious threat in the other 19 alternate histories. Pericles suffered in this game from not doing any one thing better than the rest of the field: Sitting Bull expanded better, Churchill had better diplomacy, Mehmed fought better, Victoria had the better economy when she was still alive, and so on. It was easy to forget that this guy was even present much of the time, and having only a single victory despite a 70% survival rate and a favorable diplomatic field has to grade out as a pretty weak performance.


Mehmed of the Ottomans
Wars Declared: 34
Wars Declared Upon: 35
Survival Percentage: 35%
Finishes: 3 Firsts, 2 Seconds (19 points)
Kills: 7
Overall Score: 26 points

Mehmed and Victoria were both high risk / high reward leaders, if for completely opposite reasons. The Ottoman leader had an obvious weakness in this setup due to being a low peaceweight leader in a field of mostly high peaceweight opposition. He also had a lot of jungle near his starting position which held good longterm potential but didn't do much for the early game. The only realistic path to victory for Mehmed was to go conquering and snowball over the rest of the map since there was no way that he could out-build Victoria or Churchill if the game stayed peaceful. To his credit, Mehmed pulled this off three different times and performed much better than the community had been expecting where he was one of the favorites to be First to Die. Mehmed's best results actually came where he crushed the even weaker Qin and then used Chinese territory as a springboard to push further east. He could only pull this off when the Ottomans dodged aggression from the other AI leaders, however, and that was rarely the case. There were a bunch of games where Mehmed was simply dogpiled into oblivion, including Game #2 where he was invaded 5 times in the first 170 turns and then Game #10 where the Ottomans suffered an insane 7 attacks (!) So Mehmed absolutely needed the diplomatic roulette wheel to spin in his favor, and while that didn't happen often, the potential was there to do well if he could survive the early turns. We saw a strong Mehmed performance in the real Playoff Game Three and then an outright victory in the third alternate history done on Livestream, leaving video evidence of how he could pull this off if anyone wants to go watch the footage.


Victoria of England
Wars Declared: 22
Wars Declared Upon: 30
Survival Percentage: 30%
Finishes: 2 Firsts, 2 Seconds (14 points)
Kills: 5
Overall Score: 19 points

The gameplay for Victoria was strikingly similar to Mehmed in terms of her outcomes while looking wildly divergent in action. Victoria also died in most games and had a good chance of winning if she was still standing when the curtain fell. However, her path to victory had nothing to do with military prowess and was instead entirely focused on economy, as Vicky needed to out-tech everyone else with her superior Financial trait. She pulled this off twice: in Game #11, where the other AIs all fought each other to no resolution for 200 turns, and then again in Game #15, where Churchill ignored Victoria for the whole game and allowed her to pull off the extremely rare solo kill of Pericles. Victoria simply needed additional territory as she was more tightly squeezed than anyone else on the map and kept inexplicably building wonders instead of using her Imperialistic trait to claim land. The marble resource at Victoria's start was absolutely a poison pill in that regard and her performance on the Livestream where she went wonder-crazy was repeated over and over again. Even worse, Churchill and Victoria clashed in nearly every game and Churchill simply won that fight almost every time. For that matter, Vicky also started a whole bunch of ill-advised wars with her neighbors so she wasn't exactly an unfortunate victim in many of these games. I actually think that this scenario wasn't far off from Victoria being the dominant AI, as she was *WAY* better than everyone else in the game from a research perspective. But she needed more space to expand and fewer wars with Pericles/Churchill so a Victoria win remained nothing more than a low-odds sideshow instead of the main event.


Qin Shi Huang of China
Wars Declared: 25
Wars Declared Upon: 28
Survival Percentage: 10%
Finishes: 0 Firsts, 1 Second (2 points)
Kills: 2
Overall Score: 4 points

This guy was overwhelmingly the worst AI leader on the map and it wasn't even close. All of the other leaders could be at least situationally strong, and while figures like Mehmed and Victoria had the odds stacked against them, they could be dominant when the luck factors broke their way. Qin was the exception to this rule: he was never competitive in any way, never for the outright victory and almost never even for the runner up spot. Qin spent pretty much every game sitting at the bottom of the scoreboard with poor expansion, low population, and terrible research. He frequently ignored Mysticism tech and was sitting around without borders popped some 100 turns into the game, crippling his early game and rendering China irrelevant from the outset. Combine that together with an abysmal diplomatic situation and weak economic traits, and it was a recipe for a disaster of a game. I mentioned this above but Qin's only runner up finish came from a distant trailing position where he managed to avoid dying to a runaway Churchill in Game #7 by virtue of lucking into his religion. Outside of that fluke outcome, Qin had a 90% elimination rate and a mere two kills across 20 alternate histories. At least he technically wasn't the winner of the First to Die category! This was a downright pathetic set of results.

Conclusions

The game that we watched on Livestream largely fell into the "Churchill victory" category which happened about a third of the time. Another third of the outcomes were the "Sitting Bull victory" scenarios where the Native Americans avoided getting dogpiled, and the final third were the "low odds win by Mehmed / Victoria / Pericles" group. It was very common for Churchill and Victoria to spend a long time warring, and when Churchill could win that struggle decisively as he did on Livestream, he was in an excellent position to win. As for Sitting Bull, he didn't get attacked by Pericles that often, but when he did, it usually spelled disaster. The two strangest aspects of our Livestream game were Mehmed getting so strong before his big collapse, and then Qin landing the second place finish at the end. Churchill's winning games almost always had Pericles or Sitting Bull tagging along as his runner up, typically when they shared a religion together. Otherwise, the general contours of the match we watched were a good representation of how things looked when Churchill took the game. Even the number of wars and slow finish date that we experienced were pretty normal; these alternate histories took a long time to play out with finishing dates on Turn 413, Turn 380, Turn 409, Turn 390, along with only a handful of games under Turn 300. The most interesting part of the alternate histories was watching the competition for the multi-game scoring lead between Sitting Bull and Churchill. They traded the pole position back and forth multiple times before Sitting Bull broke the tie in the very last Game #20 - by killing Churchill himself to score point 62 and then eliminating Mehmed for point 63! That was genuinely entertaining amid a series of otherwise kind of dull games.

Thanks as always for reading, I hope you enjoyed this look back at Playoff Game Three!