Civ 4 AI Survivor: Season One Alternate Histories

[El Grillo]

I remember looking into passive espionage, including demographics, when I was modding Civ4 several years ago, and I believe it's accessible via python, where it would be possible to add a condition such that the human player skips the espionage point check and always gets full visibility. I think that would solve the issue without affecting the AIs, but I can't guarantee it.

[Eauxps I. Fourgott]

Now that Season 8 of AI Survivor is done, I am planning to resume posting these. With interest in this competition at Realms Beyond so dead, though, I don't know if there's an audience here!

So if you are interested in having them here, speak up now. Otherwise, I'll just post them in my sister thread at CivFanatics.

[haphazard1]

Hi Eauxps! I would be interested in seeing the alternate histories here. Been a while since I last went to CivFanatics, I will have to remember to see what is going on there occasionally.

[LKendter]

I enjoy reading the alternate histories.

[Hesmyrr]

I think it's more of the fact that General Discussion is generally quiet in the first place rather than the fact that there's lower interest in RB (though it's definitely less compared to Civfanatics for sure!)

Anything that would be cause of activity would be nice to have, will be sure to comment on your analysis if you do post them here!

[Eauxps I. Fourgott]

Thanks folks, now that I've been corrected, I'll certainly keep posting them here.  I'm thinking I'll leave a week for cooldown from the Season 8 Championship, then post my next teaser next Saturday ahead of the full post on Monday. I have eight more writeups stocked up at this point (was on a huge AH kick a few months ago) so it will be going on for a while.

Earlier, the whole espionage points deal was a big point of discussion here. I made the decision at that time to go ahead and keep doing these as I was already doing them, with no espionage points invested against the AIs. I can't say there was any noticeable difference in the game dynamics, but in any case, there's that disclaimer. I intend to do AHs for the Season 2 games once I'm done with these, and at that time will start doing the Great Spy thing after doing my best to figure out how many Sullla was using against each civ that season.

More notably, I only recently realized that I was making the mistake of failing to give Archery to the barbs in all of the replays. I'll fix that in all future ones... but that leaves much of this season of AHs run with barbarians who had only warriors until partway into the game when they popped Archery on their own. I can't be certain but I think the biggest effect was to allow a few more outlier results when an AI got some easy barb cities and was stronger than normal, but that should mean that the core dynamics in these are faithful to the original conditions.

It's also been noted that a downside to the AHs is that, since WorldBuilder editing is necessary to enable the "new random seed on reload" option, the AHs are run without the same original rolls (such as peaceweight modifier) as the original game had. Since AHs are normally run from a regular savegame file, this means that they're also all run with the same, different starting PW roll, and that will be the same for most of my AHs. For my most recent one, though, I determined that I could load from the WorldBuilder file instead with no ill consequences, and thus get new rolls each replay. I'll be doing that moving forward.

Finally, I should note that it was recently brought to my attention that my inclusion of an AI observer civ reduces the land threshold for Domination by 2% pre-United Nations (at which point I delete the observer). This has probably resulted in a few narrow Domination wins that otherwise would not have occurred, but again the core dynamics of the map are preserved, and it's a worthwhile tradeoff to be able to capture the accurate dynamics of the maps pre-AP removal.

Now that those caveats are out of the way, I'll be back in a week!

[haphazard1]

Thanks for doing all of this, Eauxps! And thanks for clarifying the espionage stuff. I know Sulla does not like using mods, so the effects of the great spies is just part of the AI Survivor environmental effects. It does impact some leaders more than others. It always makes me cringe to see AIs running 20% espionage slider and burning huge amounts of precious commerce on spy points. But there is not much to be done about it, any more than the way the post-BtS AI loves to run spy specialists rather than something more useful. That entire aspect of the expansion was poorly implemented, with the AI vastly over-valuing espionage points. (Yes, I know I am still complaining about an expansion that came out many years ago. )

[Eauxps I. Fourgott]

After a season-long delay, it's finally time to start these up again with Game 4. The original version of this game saw Huayna Capac's start of glory as he came out on top of a bunch of warmongers to win a Domination victory, while on the other side of the map Suleiman took advantage of a strong start to cruise to second place. Full writeup: https://sullla.com/Civ4/survivor4.html

Here's the map for this game: 



And the teaser results:



Full results coming Monday.

[Hesmyrr]

I forgot completely about season 1 games, but wow poor Gilgamesh. His position would have never flied on recent seasons where maps are round in general. I suppose he is 6th place just ahead of poor Sitting Bull at last with 0 points, and Monty is clearly 5th from just looking at the survival. Also obviously HC is 1st and Suleiman 2nd just from virtue of their start. I suspect these rankings will stay the same even after full round of games.

High runner up and FtD count from 3rd place is interesting, I assume it's GK? Wonder if full result will differentiate him and Shaka more.

[Eauxps I. Fourgott]

Game 4




In contrast to the previous couple of games, which were fairly open-ended affairs, Game 4 had a pretty limited range of possibilities. I’d guessed before running these games that it would end up as a shootout between Huayna Capac and Suleiman, who had come in the top two in the real game and looked to have much better starting positions than the rest of the field.

I was half right. This game was a shootout, all right, and Huayna was strong on the map, but I was surprised to find out that his big rival was not Suleiman, but Shaka! Huayna and Shaka were the clear dominant pair on this map, combining for all but two of the wins and two-thirds of the game’s kills. Both were routinely major players, and any game where one of the two did not win was an extremely unlikely result. The secondary pairing of leaders in this game consisted of Suleiman and a surprising Gilgamesh; while they were almost always lesser contenders and neither was at all likely to win the game, most games saw at least one of them as a player of note for at least the first part of the match, and whenever Shaka and Huayna fought and one was knocked out, it would be one of these two taking the loser’s place in second. Finally, the remaining three leaders were also-rans; Montezuma frequently got out to strong starts before everything inevitably came crashing down midway through, Genghis was too weak to contend but managed to scrape by as often as not, and Sitting Bull was the game’s punching bag and utterly doomed.

A typical game on this map would see some combination of Monty, Huayna, Shaka, and sometimes Suleiman take a leading position early, followed by either Huayna, Shaka, or both consolidating their position via conquest to become a strong game leader (as Monty collapsed somewhere along the way and Suleiman, if significant, stagnated). Sometimes the big two would take out all the other leaders on the map to go to the end together, or one would be eliminated without the other’s involvement, but there were also plenty of games where they ended up squaring off in a climactic showdown, with the result determined by the game’s earlier events. Shaka tended to emerge victorious if he’d started a successful snowball earlier, while Huayna was more reliant on peacefully pulling ahead while others stagnated. Peaceful endings to the game were very rare, as whoever got the advantage would keep on conquering or culturally encroach on the remaining rivals until crossing the Domination threshold – a full 80% of games ended with that victory condition! Second place was usually a matter of which of the four top leaders didn’t get all beat up over the course of the game.

As could be expected from the lineup of competitors, this was a very violent map, with 16 wars per game on average and very few non-Domination finishes. Suleiman and Gilgamesh were able to stay out of trouble with some regularity, but the eastern side was a total bloodbath. It made for some exciting viewing at times, with things regularly in motion and multiple ‘clash of the titans’ scenarios across the the set. It’s no wonder much of the fanbase prefers the warmongers! It also made tech paths quite important at times, with several results decided by military tech discrepancies. Further adding excitement to my run through these was a very tight race for both the first and third-place positions in the scoring; Shaka and Gilgamesh seemed like shoo-ins for these spots for a long time, but Huayna and Suleiman both made determined runs towards the end, causing there to be TWO ties in score in the final results! I broke the tie in favor of the leader with the most finish points in these cases.

Now for a look at the individual leaders:

Shaka of the Zulu

Wars declared: 86
Wars declared upon: 25
Survival percentage: 75%
Finishes: 9 firsts, 5 seconds (55 points)
Kills: 25
Overall score: 80 points

Shaka put on a very impressive performance here to tie Huayna for the win in final score, officialy placing first thanks to scoring one more second-place finish. He was two points behind this map’s kill lead, boasted its best survival rate, and scored a top-two finish in all but one of those games. This had been quite unexpected before the series began, as his position had seemed quite tenuous – starting in a central position, with both Montezuma and Genghis as neighbors, would surely result in a lot of damaging stalemates and/or dogpiles, right? Not so much, in fact, as Shaka faced one of the lowest war declaration counts of any leader on the map, usually able to choose his battles and rarely caught in multi-front conflicts. It turned out that the rest of the field divided into three pairs of natural rivals (Genghis and Sitting Bull, Monty and Huayna, Suleiman and Gilgamesh), and that meant that everybody else had at least one other leader they were likely to target above Shaka. This turned his central position into an advantage, allowing him several different potential avenues for conquest, paths by which he could start to snowball into a winning position.

And snowball Shaka did, as only one of his eight wins came via the spaceship and the rest were all Domination. Shaka’s general path to victory in these games was through conquest, but the specifics were up in the air and there were a lot of ways that he could pull this off. Different winning games saw him get the ball rolling against Genghis, Sitting Bull, Monty, and Suleiman – he just needed somebody that he could fight, alone or with a friend, and take the land from, then go from there. Gilgamesh was the only leader that Shaka never really conquered here; while Huayna was never his first mark, Shaka often ended up fighting him eventually, and he served as more or less the final boss in several of Shaka’s wins, including one where Shaka was behind in military tech, but brute-forced his way through anyway with about 5 times the units that the Inca had. Another game saw Shaka clearly too far behind Huayna in power, only to sidestep him by conquering Genghis instead to get over the Domination threshold. Shaka’s games varied in dominance and pattern, but he proved versatile and found a lot of ways to get the job done.

Of course, there were also plenty of games where he couldn’t get it done, as a warmonger’s life is never easy. Plenty of these games saw Shaka still perform fairly well, but get outscaled by a peaceful Huayna Capac (or Gilgamesh, in one unusual game) and either finish in second or lose a climactic war. Some others, around a quarter of all the games, saw him simply fail to really get things started, getting bogged down in early wars instead of conquering, or in two instances getting knocked down early and dying first. Shaka’s performance in the real Game 4 was clearly one of those weaker ones, as he never got the snowball rolling and was relegated to the status of minor leader as a result. Still, a major warmonger only logging dud performances in about 25% of games is much better than you usually see, and overall I think this was a very good performance by Shaka. We’ve seen him outperform his peers time and time again, and this set of alternate histories only bolsters the idea that he really has something figured out in this competition.

Huayna Capac of the Inca

Wars declared: 52
Wars declared upon: 57
Survival percentage: 70%
Finishes: 9 firsts, 4 seconds (53 points)
Kills: 27
Overall score: 80 points

Right up there with Shaka as the other titan of this map, with virtually identical results, was the ever-dangerous Huayna. In this field of leaders who lacked economic focus, Capac was by far the best techer, and he leveraged that skill to pull ahead in many of the contests here. He also benefited from what was probably the best starting location on the map, in the corner but with very spacious backlines to expand into; this limited the number of other leaders who could attack him while also allowing him to exit the landgrab phase as a top leader in game after game. From that point, Huayna’s path to victory was to steadily strengthen himself until nobody else could stand up to him; this could happen quickly via a snowball of conquest, or more slowly via partial conquests or out-teching his competition. Huayna also benefited on this map from both of his neighbors being popular dogpile targets; he very often strengthened himself at Monty or Sitting Bull’s expense, and while he sometimes did it on his own, there were also plenty of games where he worked with one or more other leaders, while taking the best spoils for himself.

Huayna was also typically strong even in the games where he didn’t win, or at least a major player in the early game. Only a couple of times did he simply get overwhelmed and knocked out of contention early on (he was never First to Die), and more often he was one of the strongest empires before getting knocked down or edged out by a Shaka (or, twice, another leader) who had successfully snowballed. The reasons for Huayna’s later failures varied; sometimes he was stalled out for too long fighting Monty or Genghis, not being allowed to really tech up. Sometimes he was just not as successful. Sometimes he seemed like he should win, but Shaka found a way to edge him out anyway. Most tellingly, though, there were several games where Huayna should have had the win locked up, only to fall into the classic economic-leader trap and avoid teching Rifling or Assembly Line for ages on end, thus leaving an opening for his rivals and eventually getting at least knocked out of the winner’s circle.

This game showed pretty well that Huayna is definitely a more economic leader. While his peaceweight and aggression aren’t typical (and do serve to make him more effective), he does follow the same overall patterns that make him stronger at tech, but leave him with some weaknesses when it comes to military. Those showed up in this game, where I think it’s fair to say that he should have been even more successful than he was, losing several games in rather silly ways. Still, as is, he only had one credible rival on this map, and if Shaka himself hadn’t performed so successfully then this would have been another case of Capac steamrolling all comers. These games showed clearly that Huayna’s strong start to his career was no accident, and reinforced that he’s one of the competition’s best leaders.

Suleiman of the Ottomans

Wars declared: 31
Wars declared upon: 46
Survival percentage: 45%
Finishes: 0 firsts, 8 seconds (16 points)
Kills: 4
Overall score: 20 points

Suleiman narrowly tied for third place at the end of the alternate histories, winning the tie thanks to being this map’s most prolific second-place leader. He came in second in nearly every game that he survived, backing up his real-life finish by duplicating it 8 times. The Ottoman leader had a spacious starting position that allowed him to expand out to a strong size and be a notable leader after the landgrab in many games; it was further a bit sheltered from the bloodbath in the east, making him a bit less likely to get caught in the crossfire. He was never able to compete directly with Huayna or Shaka, but in the games where they targeted each other (or one was killed early on), Suleiman would usually be over there, the next-biggest leader after their (and Monty’s) demise, winning second place by default – much like what happened in the real Game 4.

However, this performance was severely disappointing when you consider the amount of space Suleiman had, and the fact that he had a cramped neighbor to his north whom he could easily squeeze if he settled well. A position this favorable should have translated into a strong performance with several wins – yet Suleiman couldn’t manage even a single victory (coming close to a spaceship win only once), was eliminated in a majority of games, and scored only a quartet of kills across the entire set – he should have scored more than that against Gilgamesh alone! The lack of kills really tells the story of Suleiman’s performance – while he was able to peacefully expand to a respectable size, he was totally unable to leverage what he had, managing a single solo conquest each of Genghis and Gilgamesh, and nothing else across the set except for participating in some dogpiles. As such, he never managed to put himself into a winning position. Perhaps the most surprising aspect of his performance was that Suleiman was by far the most likely leader to die first aside from Sitting Bull, suffering that fate in seven different matches. These would usually come from dogpiles from a combination of Gilgamesh, Monty, and Shaka, a situation where he did not fare well at all, but there were also a couple of games where despite his better starting location, Suleiman would straight-up lose to Shaka in a 1v1. Unsurprisingly, those games went well for Shaka.

Overall, this was a very telling set of alternate histories that reinforces that Suleiman is not a very good leader for AI Survivor purposes – a lesson that’s also been taught well in his post-Season 1 performances. I’m not sure what exactly his problem is, but he’s simply not that effective, and that’s perhaps best illustrated here, where he could only barely muster the same score as his neighbor who had a much worse starting position.

Gilgamesh of Sumeria

Wars declared: 39
Wars declared upon: 23
Survival percentage: 65%
Finishes: 1 first, 3 seconds (11 points)
Kills: 9
Overall score: 20 points

In direct contrast to his neighbor Suleiman, Gilgamesh outperformed expectations in a big way here, putting on a respectable performance despite what had appeared to be a terrible starting location. Gilgamesh was stuck in a far corner of the continent, and in the real Game 4 had been quickly trapped there by Suleiman’s settling, stuck on just four cities and a total non-entity throughout the game. This happened sometimes in the alternate histories… but not that often. Only 25% of games saw him thoroughly boxed in, and less than half saw him really restricted; in the rest he was able to break out and settle several cities out to the east, getting out to a decent size if not being as reliably strong after the landgrab as some.

Once at that point (or even when he failed to reach it), it turned out that Gilgamesh had one significant advantage over the other four lesser leaders in this match: his position was very far away from Huayna and Shaka, and thus he was far less likely to get run over by them over the course of the match. This gave him nearly as good of a survival rate as the two big dogs, allowing him to lurk in the wings and occassionally poach kills and second-place finishes to build up his point total. In fact, Gilgamesh never once got conquered by Huayna over all the alternate histories – he was just too far away! Shaka did wipe him out several times and he got dogpiled out once, but on the other hand, he was never the first to die and was only solo-conquered by the much more privileged Suleiman a single time over all twenty games. On the other hand, even with this good survival rate, Gilgamesh was still pretty limited in what he could do from his spot and came up empty more often than not. For the most part, he was completely unable to compete with Huayna, Shaka, or even Monty, and only occassionally able to get the upper hand over Suleiman. While he would stay competitive for a while thanks to a strong economy (I think he built the Great Lighthouse a lot), ultimately his lack of good conquest opportunities would result in him getting outscaled as the game went on. Sometimes he could do nothing about this; otherwise it really did seem to be his fault, as he just sat back doing not much of anything for a long time, failing to press any advantage he had on any neighbors. It wasn’t a consistent performance.

Still, Gilgamesh was able to see real success a few times. His three second-place finishes all came via similar fashion, as one of the two big dogs was killed somewhere along the way (once largely due to Gilgamesh himself!), and Gilgamesh himself helped eliminate Suleiman (once completely on his own!), thus leaving him as the second-place leader without the Ottomans around. He also was able to clock in with his own Domination victory once, in a game where he expanded to a bigger size than normal, got the better of multiple dogpiles executed with Shaka, then knocked out Shaka himself in an endgame conflict. Overall, then, it was a much more impressive performance than that of Suleiman, who had mostly just scored his points by having a good start; Gilgamesh fought for all of his, scoring over double the kills and winning when Suleiman could not. While there were some disappointing games from Gilgamesh here and it felt like he could and maybe should have done even better, it still was a surprisingly strong outing, which indicates that Gilgamesh does deserve his fairly strong AI Survivor reputation; he can get a lot more done than some leaders, even from a cruddy position.

Montezuma of the Aztecs

Wars declared: 43
Wars declared upon: 66
Survival percentage: 5%
Finishes: 1 first, 0 seconds (5 points)
Kills: 5
Overall score: 10 points

Monty performed very poorly on this map, but rather than the impotent psycho that we’re used to, in this set of alternate histories he was instead something of a tragic figure! Monty was far from a joke at the outset of these games, with a decent position and near-constant early religion making him one of the early score leader in almost every game. He often was a credible fighting force as well, and plenty of Huayna’s non-wins can be attributed to fighting with Monty in almost every game (thanks to the two usually founding the opening religions), with Monty doing a good enough job to stall out the Inca and prevent them from snowballing (in territory or tech) too early. However, Monty was ultimately doomed; in game after game, he found himself stabbed in the back, dogpiled and knocked out of the game with not even so much as a kill to his credit. He had the most central position on this map, and thus was a frequent target for everybody except Genghis and Sitting Bull – AKA the two least relevant leaders. Remember that Monty is THE most aggressive leader in the game, then look at the stats above – less than 40% of the wars he fought were ones that he started! While he sometimes was asking for this (mainly when Huayna came attacking him back after Monty had attacked him earlier on), there were also plenty of games when Gilgamesh or Shaka or Suleiman attacked him unprovoked. As a result, he was constantly facing multi-front wars that he couldn’t win.

At some point this crossed a line, and Monty’s terrible luck became hilariously ridiculous. He didn’t score a single point until the second half of the alternate histories, and the games at times went to great lengths to keep this the case. Multiple times, he’d finally be getting the upper hand in a 1v1 and knock his opponent down to his last few cities… only for somebody else to slam into him from the other side and eliminate him completely, as he never got to strike the finishing blow. (In one case, he had Genghis down to his last city before getting dogpiled by Huayna and Shaka, never getting to finish the job.) He at least got a single day in the sun, in a Game 14 where he never got backstabbed and actually got to finish all of his wars himself. That game was a resounding success where he eliminated both Huayna and Shaka en route to a Domination win and eight of his ten total points. Aside from that, though, he managed to score a single kill in two games, but died in all nineteen other contests. Even though this performance tracks with Monty’s poor record, I think that at the same time it can’t be counted as representative at all. Normally Monty is digging his own grave, but here he kept getting screwed by forces largely beyond his control. It was odd to see him as the sympathetic underdog, but it was the role he had in these games. Poor guy.

Genghis Khan of Mongolia

Wars declared: 71
Wars declared upon: 22
Survival percentage: 50%
Finishes: 0 firsts, 0 seconds (0 points)
Kills: 8
Overall score: 8 points

In direct contrast to Monty, who would start strong but virtually always get killed, fellow doomed warmonger Genghis had a terrible position and routinely terrible starts, but much like in the real Game 4, managed as often as not to get lucky and stick it through to the end. In the original Game 4 writeup, Sullla pegged Genghis’s starting position out as terrible, and he was right. With rather poor local terrain and a location cramped against the corner of the continent, this spot was fairly comparable to Tokugawa’s crappy Game 1 start, and Genghis similarly was playing from behind from the beginning of the game. His expansion was routinely poor and he would only have a few cities when he went off to war; he was never very strong or able to come remotely close to snowballing out of his position.

But much like Tokugawa in Game 1, Genghis was determined to do his best to fight his way out of his corner, putting on a scrappy performance as he kept on fighting and fighting and actually made it through a lot of the time. Mind you, Genghis’s wars were never exactly resounding successes – his kills all came off of weakened enemies or dogpiles – but he was able to stay in it despite his poor size and proximity to both Huayna and Shaka, and declared the second-most wars in these games by a healthy margin in the process. He unsurprisingly fought Sitting Bull the most often (their solo wars generally being stalemates between two weak leaders), and would also go up north to attack Huayna a fair amount of the time, but he actually fought Shaka fairly rarely despite their mutual aggression and shared border – it happened sometimes, but more often they tended to be buddies, which certainly helped Shaka in his successes.

There’s not much else to say about Genghis’s games. He was too weak to really compete, only coming within sniffing distance of second place a single time, but the real result of him surviving to the Wildcard game was absolutely a realistic outcome, duplicated in half of the alternate histories. He had a knack for avoiding hot wars with the major powers that allowed him to scrape by, and his constant attacks in the meantime let him snipe a handful of kills, pretty much the best results that could have been hoped for from this position. Overall this was a performance tied to the position and doesn’t say much about Genghis as an AI, other than that he’ll at least fight and can do a bit more from a terrible position than some leaders.

Sitting Bull of Native America

Wars declared: 7
Wars declared upon: 90
Survival percentage: 0%
Finishes: 0 firsts, 0 seconds (0 points)
Kills: 0
Overall score: 0 points

In the original Game 4 writeup, Sitting Bull was the first explicit example of how peaceweight affects diplomacy, and this was fitting to say the least. He was utterly screwed on this map as the lone high peaceweight leader in a field with six warmongers, and neighboring three of the most aggressive leaders in the game plus Huayna Capac. There was nothing he could do to escape this situation; he was attacked over four times per game despite frequently being eliminated first! Or, put another way, he was attacked on average once every forty turns in this game. While he could sometimes hold off his aggressors for a while, he never got the chance to peacefully develop or made any gains against his opponents, and even on the rare occassions where he survived the first round of wars, he was too far behind and completely impotent, thus just getting picked off later on instead. Sitting Bull clashed most frequently with Genghis, who generally stagnated him more than anything else. He also got attacked by Huayna a good deal, and these wars tended to be much more successful and fuel Capac’s successes. The other four leaders all wanted a piece of Sitting Bull as well and all took swings at him from time to time.

It does bear mentioning that even in his better games, Sitting Bull never showed any signs of pulling ahead in any way or becoming a really viable leader. There were a couple of games where he made it through the early phases relatively unscathed but was still hopelessly behind. While this could be explained as a result of him always having to do at least some fighting, I do think it also speaks to his weakness as an AI, lacking both economic skill and any real ability to dynamically fight. Sitting Bull is very probably one of the very worst leaders for AI Survivor purposes, and he was furthermore stuck in an impossible situation in this match – a perfect storm that led to my first 0 in these games.

Conclusion

The alternate histories revealed that, for the most part, we’d been given a typical map for this match. Huayna was frequently dominant, Suleiman often finished in a distant second place by default, Sitting Bull was frequently dogpiled early, Monty almost always started strong but fell apart later, and Genghis often survived despite a terrible start. Domination was absolutely the expected victory condition, and even the game length and number of wars hewed close to the alternate histories! The only real surprises were the strong performances of Shaka and Gilgamesh, both of whose real performances were near the bottom of their scale of possibilities, and in particular it’s too bad for Shaka that he didn’t get to show any of his strenghth here in the actual game. Still, this has been one of the more straightforward setups in Season One so far. It will be an interesting contrast going from these last couple of games to a couple of more peaceful fields for the next two.