Civ4 AI Survivor Season 5: Playoff Game One Alternate Histories


Introduction

Playoff Game 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. I was one of the many viewers who felt that the Playoff Game One that we saw had been a weird sequence which was unlikely to occur again. I mean, Washington as the overall winner? Louis backstabbing Pacal right at the moment when the Mayan leader was about to start snowballing into runaway status? Those had to be flukes, right? 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:

Season Five Playoff Game One

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.)

Surprise! It turned out that I was completely wrong and the game that we watched ended up being fairly typical for this map. Washington genuinely was the leader most likely to win, and although Pacal did score a series of victories himself, he clearly had a rougher time getting across the finish line than his American counterpart. No one else even came close in terms of wins: Washington was the favorite with Pacal a secondary threat to emerge on top. The other leaders only won in unusual circumstances that rarely popped up - those were the true oddball games. So what happened here? How did the favorite in the First to Die contest who was picked by less than 5 percent of the contest entrants to win the game end up being the strongest AI on the map? This wasn't a case of a weak leader lucking their way into a victory as we'd seen in some other cases. What happened here to throw off the community?

The key to understanding this map was the geography surrounding Alexander's starting position. I do a basic check of all the AI capitals to ensure that they're reasonably balanced before the games start, and part of that process involves making sure that all AI leaders have reasonable access to a metal resource of some kind. It might be iron or copper and it doesn't have to be right on top of their initial settlers but I won't let anyone get locked into a metal-less position because that means certain doom in Civ4. Well, to make a long story short, that process ended up being a failure on this particular map. Alexander had iron exactly four tiles away from his capital and I thought that this would be sufficient to secure him access to the critical resource. What I didn't expect was Pacal sending his starting settler to that exact spot in every game. Not once or twice, not most of the time, literally *EVERY* game. Alexander never managed to get control of that iron resource even one time and this left him without any source of metals, i.e. a hopelessly crippled civilization.


The iron resource in question

Total failure on my part, argh. Everything else that took place in this game flowed downstream from this initial event. Alexander was the first casualty as he was left with no shot at all. His Greek civilization was wiped out in 18/20 games, fully 90% of the matches that I watched, and Alexander was the First to Die in 11/20 games. The Greeks never emerged as a dominant power and were rarely even mediocre in strength. The early action in each game was nearly always dominated by someone running over the hapless Greek civilization and then snowballing off their easy conquest into one of the top spots. The leaders best positioned to take advantage of Alexander's stumbles were Washington and Pacal, and thus it should come as no surprise that they ended up winning the most games. (I'm a little bit surprised that Charlemagne didn't have more success in this regard but for whatever reason he didn't conquer Alexander as much as you might think. Perhaps it was his lack of plotting war at "Pleased" relations; Charlemagne didn't keep the conquest train going that often even when he did have early success.) There were a fair number of community members who sniffed out Alexander's potential problems and picked him as First to Die. This was absolutely the correct choice and they were rewarded with points in the game that we ended up watching.

So Alexander was a complete weakling who couldn't serve as more than a speedbump for other leaders due to his lack of copper and iron. A warmonger who can only build chariots and horse archers doesn't make for an very effective AI. But it wasn't just Alex: it turned out that Louis was also a bit of a featherweight as well. We saw an unusually strong France in the game that we watched thanks to the backstab of Pacal. In most games Saladin was stronger and Louis was weaker, as the French leader was only able to manage a single victory in Game #7. I was correct that Louis didn't fight Pacal very often, roughly once every 3 or 4 games, but it also wasn't some kind of impossible fluke either, just unlikely. Louis was never First to Die and he would usually spend his time clashing ineffectively with Saladin or occasionally Washington before getting eliminated sometime in the midgame or lategame. He actually mananged fewer kills than Alexander which is just sad.

With a crippled Alexander and an ineffective Louis to either side, it was Pacal who suffered the most. Time and again Pacal would emerge in the midgame as the strongest leader on the map only to be the victim of a dogpile from the high peace weight leaders. He was an effective enough leader to win six different games anyway (if you look closely you'll see that Pacal's wins corresponded with few wars being declared against him) but he could have and would have won several more without having the bad luck to be the only low peace weight leader left. This map turned out to be heavily stacked in Washington's favor, between his amazing starting position and a pair of moderately aggressive "won't declare war at Pleased" leaders flanking him on either side. Washington could snowball off Alexander or through a dogpile against Pacal, and once he had the most land, it was easy to lock in a bunch of wins with neighbors who would never declare war on him. Some of these matches had very boring endgames where everyone loved each other. I'm not convinced that Washington was unusually effective as an AI leader in these games, rather it seemed to be a perfect setup for him to dominate.

Now for a look at the individual leaders:

Leader Summaries


Washington of America
Wars Declared: 39
Wars Declared Upon: 57
Survival Percentage: 60%
Finishes: 8 Firsts, 2 Seconds (44 points)
Kills: 18
Overall Score: 62 points

Washington was the best performer on this map with 9 total victories (including the real playoff game) in 21 matches. I explained some of the major reasons why in the overview paragraphs above: a weak no-metals neighbor to devour for additional land, generally friendly neighbors located on both flanks, and his biggest rival AI leader being someone who tended to get dragged down by a poor diplomatic situation. It also helped that Washington had a dream capital that lined up well with America's starting techs, a wet corn and a gold resource along with six floodplains tiles. He even had a stone resource that often turned into Stonehenge or the Pyramids. Washington's two main routes to success came from crushing Alexander or being the leading force in a coalition that took down Pacal. He was well positioned to take advantage of either situation and it was common for both to happen in a single game. Washington struggled when his high peace weight made him a target for a whole bunch of the other leaders simultaneously. We knew that this was a possibility and it manifested in a minority of the games, resulting in Washington being the First to Die on three different occasions. Saladin mostly went to war with Louis early on, but if the Arabs attacked to the west it was almost always bad news for Washington. Similarly, Washington rarely clashed with Charlemagne and it was typically a disaster when that scenario popped up. A disproportionate amount of Washington's eliminations came from getting attacked simultaneously from the north as well as the east or west.

Washington's performance was highly correlated with how often he was attacked by the other leaders. In many of his wins, he faced only one or two (or even zero) attacks from his competitors. By way of contrast, Washington was the subject of gang invasions in many of his eliminations, especially Game #6 where he was attacked seven different times. To his credit, Washington also faced seven invasions in Game #17 while still going on to win but that was a clear outlier result and normally lots of defensive fighting meant a poor performance. It became clear that Washington is a moderate to low aggression AI leader as his dominant position didn't translate into more offensive wars and more kills. He only declared war 39 times which isn't a lot given the leading position that America occupied in many of these games. Washington was also more willing to sign peace treaties than many of his competitors, and that's part of the reason why a significantly less dominant Charlemagne managed to tie him with 18 kills. Overall this seems to have been a really, really good setup for the guy. He found himself in an ideal position and did a reasonably good job of capitalizing on it.


Pacal of the Mayans
Wars Declared: 23
Wars Declared Upon: 54
Survival Percentage: 50%
Finishes: 6 Firsts, 2 Seconds (34 points)
Kills: 13
Overall Score: 47 points

Pacal was the second-most likely leader to win on this map by a clear margin. He can best be compared to other "feast or famine" leaders like Mansa Musa or Willem or Gandhi as Pacal generally either scored a top two finish or was eliminated from the game. In fact, Pacal had a long streak of taking one of the top two spots in every single game where he survived right up until the last three matches. Pacal's gameplan was heavily economic-focused and based around out-teching his competitors using his Financial trait. He wasn't the best at expanding and he wasn't particularly aggressive at all, with 23 offensive wars stacked against 54 defensive wars. As you would expected, only one of Pacal's wins came by Domination with the other five blasting off into space. There were many situations where Pacal could have looked to muscle out his competition once he had pulled ahead and the Mayan leader simply preferred to remain at peace and continue building. This is the biggest flaw in his AI personality and the reason why Pacal is a tier below someone like Huayna Capac.

There were two general routes to victory for Pacal in these matches. The first was to capitalize on a dogpile of Washington in the games where Louis or Saladin or Charlemagne attacked the American leader. Washington was eliminated in five of Pacal's six victories and that wasn't a coincidence, as a weak American civ meant that the Mayans were free to dominate the middle of the map. The other path to success was to be the civ that wiped out Alexander's weak holdings to the west and grow off that additional territory. This was the riskier path though because Pacal could easily face an attack from Washington or Charlemagne or even occasionally Saladin while he was tied up with Alexander. That was the exact scenario that took place in the playoff game that we watched and it happened repeatedly in these alternate history scenarios. Pacal would need 25-30 turns to conquer Alex and then another 25 turns to consolidate afterwards to be safe, and he simply didn't get that much time in many games. On the other hand, once Pacal did get out in front of the pack it was nearly impossible for the other leaders to catch him thanks to his Financial trait and econ-heavy build preferences. Pacal could sometimes ignore Rifling and Assembly Line techs for a dangerously long period of time and this did cause problems in some games. He likely cost himself another victory or two by not concentrating more on military technology.

On the one hand, Pacal was somewhat unlucky in these games and he could have scored several more first place finishes with slightly different circumstances. He lost an exceedingly close space race to Washington in Game #13, his spaceship about five turns slower than the American one, and he was the single most powerful leader in Game #18 before having the bad luck to declare war right after Saladin signed a defensive pact with Washington. I kept watching games where Pacal was just about to turn the corner into an unbeatable position, right when he would suffer another ill-timed invasion that unraveled everything. On the other hand, the very fact that these "bad luck" things kept happening was a reflection of the poor strategic position that Pacal occupied. His low peace weight compatriots were both weak entities on this map and frequently left Pacal as the only survivor facing down multiple high peace weight opponents. Both Louis and Alexander can plot war at "Pleased" which meant that they were never truly reliable partners and always subject to backstabbing him. While Louis didn't attack Pacal very often, when it did happen it tended to be devastating as we saw in our Livestream game. Pacal simply needed a lot more things to go his way than Washington did - the degree of difficulty was noticeably higher. The fact that Pacal still managed to win six games regardless is a sign of his overall strength as a leader for AI Survivor. I think he's clearly the best leader out of this group even if he wasn't the best option to pick on this particular map.


Charlemagne of Holy Rome
Wars Declared: 54
Wars Declared Upon: 14
Survival Percentage: 55%
Finishes: 2 Firsts, 6 Seconds (22 points)
Kills: 18
Overall Score: 40 points

There was a debate in the community as far as whether Charlemagne or Saladin would be the best choice for the Runner Up category, and it turns out that the two of them were pretty much a dead heat to finish second. (Even the playoff game that we watched had them jockeying for a close second place finish!) Charlemagne was typically in the mix in each game while rarely being in position for the outright win. He was also First to Die three different times which was a real surprise to me given his sheltered starting position and Protective trait. Charlemagne's path to victory pretty much always unfolded the same way: an early conquest of Alexander that doubled his territory and served as a springboard for further military ventures. There were no games in which both Alexander and Charlemagne were strong; all eight of Holy Rome's top two finishes saw Greece eliminated, usually early on. Charlemange could also benefit sometimes from a wider dogpile on Pacal although he would typically get only a small part of the spoils since his territory didn't directly border the Mayans. The weirdest games were the small few where Charlemagne went after Washington at the outset, typically driven by the two of them practicing rival faiths. That was a rarity but it did pop up in two or three games. This usually spelled disaster for both leaders, such as in Game #15 where they were both eliminated and Alexander survived to game's end. Somehow both of Charlemagne's victories were Diplomatic wins, which both speaks to his popularity amongst the high peace weight leaders and still has to be some kind of crazy fluke outcome.

Charlemagne was the most aggressive leader according to the war declaration stats, initiating 54 wars while only facing 14 attacks. This was no doubt driven by his corner starting position and the fact that Washington and Saladin (his most common allies) not being willing to declare at "Pleased" relations. Charlemange launched a disproportionate number of his invasions against Alexander and Pacal, again and again taking advantage of the metal-less Greeks to munch on their territory. This was also the reason why Charlemagne tied for the most kills with 18 despite winning so many fewer games than Washington; he was the one scoring the kills of Alexander in game after game. I'm a little bit surprised that Charlemagne didn't do better in these matches given how the map was set up favorably for him. I think this ties back into his terrible Holy Roman starting techs, with Washington just getting out to much faster starts and therefore being in position to accelerate his opening better than Charlemagne. The Burger King ended up being his usual self, not especially dynamic but kind of hanging around until he managed to find himself in second place yet again. I think we have a decent read on his AI personality by now: a solid leader that falls somewhere above the middle of the pack without being top tier.


Saladin of Arabia
Wars Declared: 50
Wars Declared Upon: 24
Survival Percentage: 75%
Finishes: 3 Firsts, 6 Seconds (27 points)
Kills: 8
Overall Score: 35 points

Out of the six leaders in this game, Saladin's typical performance was the most divergent from what we observed in the actual playoff game. Saladin ended up clashing with Louis in almost every single game, and it was rare that the two of them weren't squabbling over the eastern part of the map for the first 200 turns. Unlike what we watched in the official playoff game, Saladin was almost always the stronger of the two leaders and the proof was in his far superior finishing positions as compared to his French rival. Saladin also founded his own religion in most of these alternate histories as opposed to never establishing his own faith in Playoff Game One. The Arabian ruler would love the leaders who shared his religion and hate those who didn't, with the natural spread of Islam often dictating how well Saladin performed afterwards. When Islam was widely spread Saladin would have lots of allies and usually took home a top two finish. When it wasn't, he became diplomatically isolated and tended to be a non-entity. Saladin's corner starting position seemed to help him in terms of survival as he was only First to Die a single time and he was the least likely to be eliminated out of the group. That's not to say that Saladin was always a strong contender for the win, however, as he had a remarkable six different survivals that were outside the top two. We're noticed this same occurrence in past seasons of AI Survivor as well; for whatever reason, Saladin seems to end up in third or fourth place (without dying) a lot.

Saladin's path to victory was straightforward: conquer Louis and absorb French territory. You can spot that this took place in all three of Saladin's victories as he always seemed to be drawn into conflict with Louis for one reason or another. Saladin was also pulled into combat with Pacal more often than I expected to see happen, with religion nearly always being the cause of that dispute when it happened. Saladin rarely fought with Washington but when it did happen it usually torpedoed American chances for a win. Saladin graded out as one of the more aggressive leaders with 50 offensive wars against 24 defensive wars, although that seemed to be the same process at work as we saw with Charlemagne. Corner starting position plus allies who don't declare war at "Pleased" relations tended to mean fewer invasions faced. Unlike Charlemagne, Saladin was much more willing to sign peace treaties ending his wars and therefore only managed 8 kills despite fighting roughly the same number of wars. Some of the AI leaders definitely have more of a killer instinct than others and Saladin wasn't interested in fighting to the finish that often. He grades out as highly similar to Charlemagne overall on this map, slightly more likely to win and slightly less likely to score kills to balance it out. Neither one of them were favorites to win and it was a coinflip as far as who would take second place.


Louis XIV of France
Wars Declared: 40
Wars Declared Upon: 44
Survival Percentage: 40%
Finishes: 1 First, 3 Seconds (11 points)
Kills: 1
Overall Score: 12 points

Louis was the other half of the eastern duo and the one that usually lost out on the power struggle with Saladin. Some of the community members had pegged him as having poor land around his starting position and expected Louis to struggle as a result. They were absolutely correct as Louis rarely accomplished much of anything in these games, managing only a single victory and three second place finishes (all of them following in Pacal's shadow). The path to success for Louis involved winning the duel with Saladin and snowballing off his territory, something that almost never happened in these games. The only other viable alternative was to expand at Pacal's expense and that worked out exactly one time when Louis scored his sole victory in Game #7. When Louis made the uncommon decision to attack Pacal, it typically ended with both of them suffering elimination at the hands of one of the southern leaders. This was a map where there weren't many successful options for Louis and he spent a lot of time floundering about without accomplishing anything.

Louis was the only leader in this game to end up with a balanced ledger of war declarations, roughly 40 offensive wars and 40 defensive wars. I would guess that well over half of those wars involved Saladin in one form or another, either as the attacker or the defender. There was almost always a barbarian city of some kind located in the jungle between their two homelands and the random luck of who took that spot often dictated the outcome of their warring. Louis was comically bad at finishing off opponents as he managed a single kill across all of those wars. There was some bad luck here as he had an elimination credit swiped away at the last moment on a few occasions, but mostly it was inept gameplay on the part of the French leader. Louis had enormous difficulty in snowballing ahead off someone else's territory and rarely pulled it out. He avoided being First to Die even one time and yet also found himself eliminated in almost two thirds of the matches. As the numbers suggest, this meant that Louis was repeatedly eliminated in the midgame with lots of exits between Turn 200 and Turn 300. Multiple times he would backstab someone for a temporary advantage while fatally undercutting his larger strategic position. Louis has an incoherent AI personality (low peace weight warmonger who also likes to build wonders) and it was on full display here. This is an AI leader that doesn't have a clear strategy and it led to ineffective results on this map.


Alexander of Greece
Wars Declared: 24
Wars Declared Upon: 37
Survival Percentage: 10%
Finishes: 0 Firsts, 1 Second (2 points)
Kills: 4
Overall Score: 6 points

Alexander had absolutely no chance on this map and if he were a real person I would apologize to him for muffing up the metals situation. Poor Alex was stuck as a low peace weight aggressor who lacked the strategic resources needed to train axes, spears, and swords. He still made a game attempt at fighting anyway, as indicated by the 24 war declarations that he initiated despite a horrendously bad strategic situation. (Keep in mind that strong AIs are more likely to start wars and weak AIs are less likely to start wars, making Alex's 24/37 balance pretty incredible for his position.) It wasn't enough, of course, as Alexander was eliminated in fully 90% of these games. He was also First to Die in more than half of the matches, another sign that he wasn't getting unlucky with late-developing events and instead suffered from a fundamentally flawed setup. Amazingly, Alexander still managed to score 4 kills across these games despite his hopeless situation. He was the opposite of Saladin or Louis, a leader who had a maniacal desire to keep fighting and never let an opponent off the hook even when it would have been a good idea to do so. Alexander actually did manage to acquire a native source of metals in two games, once getting an iron resource down by Charlemagne and another time getting a copper resource even further to the south. This could only happen when Washington and Charlemagne were caught up in an early war and therefore it was exceedingly rare. Still, I did see it happen and Alex was vastly more dangerous on the rare occasions where he could solve his strategic resource shortage. There's not too much more to say here: Alexander was screwed over by the map setup and he was crushed as a result in almost every game. I'm sorry bud, that was my bad.

Conclusions

There were some peculiarities here and there but we ended up seeing a fairly typical outcome on this map for Playoff Game One. Washington was the best choice for Winner, Alexander was the best choice for First to Die, and it was a coinflip for Charlemage/Saladin in terms of Runner Up. All of that was exactly how the playoff game played out. Spaceship was even the most common victory condition and number of wars was close to what we observed (average = 11.50). This game was a perfect example of how an individual leader doesn't have to be great overall to be a strong choice on a specific map. Everything was set up perfectly for Washington here and he was repeatedly able to take advantage of it. After seeing several low-odds results from some of the other alternate history scenarios, it was reassuring to confirm here that there are indeed predictable trends that play out over and over again. We're aren't tossing darts at a target blindfolded; there really are "correct" answers to the picks on many of these maps. As always, I hope that you enjoyed this data dive - thanks for following along with Civ4 AI Survivor!