AI Survivor Season Two Alternate Histories

[Eauxps I. Fourgott]

A newer but (to some, at least) no less interesting feature in Sullla's AI Survivor series is the 'alternate histories': looks at sets of 20 replays of each game to determine the map's 'true' dynamics. These have been run in an organized and official capacity, with the results posted to Sullla's website, since Season 5, but that left a gap in coverage in the older seasons. Various people have filled in that gap over time, leaving only Season 2 remaining, and @Genghis the Barbarian and I agreed to "split up" these thirteen alternate histories to finally complete the full coverage with detailed writeups for each game. Sullla has previously declined to undertake the effort to convert the writeups to HTML and post them on his website, so here they will be posted instead, for the few fans who actually are interested to read them. (I would be doing them regardless as they really are that inherently enjoyable for me.)

Genghis and I will be posting the writeups for all of these alternate histories in this thread, for however long it takes to complete them all. Below this will be the index with direct links to each completed writeup, and below that are a few more up-front comments about methodology and such that form the rest of my introduction and may prove useful to some.

Also: if anybody new to the AI Survivor series manages to stumble upon this in the future, I recommend you put these off until catching up through Season 8 in the series, if you care about avoiding spoilers.

[writeups will be indexed here as they are posted]

Now then, a few notes about methodology and game conditions.

• For anybody unfamiliar with the process, the idea is that we take the starting save for the map and replay it 20 times, using debug mode and (for most of us) autoplay commands to watch the game play out rapidly and observe the dynamics on repeated playthroughs.
• Some rely mostly on the endgame replay to glean important bits of information, or watch the whole game through without stopping, while on the other hand Genghis avoids using the replay at all and plays through all replays manually (I don't know how he stands it). I personally go for a midway point, usually autoplaying 50-75 turns at a time, taking a break to observe the game state, then continuing. This is not a very meaningful difference, however.
• Other "vintage" alternate histories, such as the official ones for Seasons 3 and 4 or the slightly different experiments done by other community members, modify the game setup somewhat to reflect later developments in the AI Survivor ruleset: specifically, they often remove the free Deity starting techs and the Apostolic Palace. However, I wanted these games to reflect the dynamics of the games as originally played as closely as possible, so I left these elements in.
• This did require one modification: the addition of an AI observer civ who sat in the ice to prevent AP victories (SITTING_OUT_BULL, for anybody who remembers Season 4). I was pretty careful with this, in particular deleting BULL whenever the UN entered the picture, and so other than some small changes to the Domination thresholds, I don't think it ever impacted the replays. Genghis, on the other hand, was able to leave out SITTING_OUT_BULL due to not using autoplay.
• Unlike in my Season 1 Alternate Histories, this set is being played out WITH the observer civ's customary pool of espionage points (gained by merging Great Spies at the start of the game) added to each AI. This was not necessary for the replays, but was done to make the replays' conditions as close as possible to original game conditions..
• Unlike early in my Season 1 set, Archery HAS been added to the barbarians throughout these replays, aside from a few isolated incidents where I forgot it.
• Finally, while the long Alternate Histories tradition has been to simply run the replays directly from the starting savegame file, it was brought to my attention during my Season 1 set that certain start-of-game dice rolls - most notably peaceweight - could be randomized by making and loading a WorldBuilder file for the start instead. We thus are doing this for the Season 2 sets.

I think that's about all I have as words of introduction.Enjoy the writeups as they come!

 

[Eauxps I. Fourgott]

With all that said, I of course waited to post this thread until I had something more to actually share. Today I finished the replays for Game 1. In the real first match ever played out on Livestream, returning playoff contenders Pericles and Mao worked together to kill off returning finalist Liz and then glide to the playoffs once more, as Zara and Victoria weren't strong enough to change the game's dynamics and Hammurabi just relaxed in his corner. How typical was this?

Original game: Video Writeup

Below is my customary "teaser" image showing the final results, but with names removed. If you're so inclined, have fun guessing over the next however many days it takes me to produce the writeup!

 

[Thrasybulos]

Well, looks like your set, in spite of the different settings, was pretty damn similar to what I got when I ran these : "Season 2 AAH".
(And I'll have to remember to veto this map if it gets drawn for Pool 1 in my League...)


By the way, I don't know whether it might be of interest to you or not, but I found (when looking for something else) a way to check the actual peaceweight values from the savefile.
I'm pretty sure there's a Python console command which would allow to get those in-game, but when I asked, no one answered, and I'm not familiar enough with the SDK (actually: barely at all) to find out by myself.

Using the tool found here (there was a post about on this forum, but can't find it right now), you can "translate" a savefile into text.

Then, using a text editor, fist look for "Begin CvPlayer":

This will tell you for which leader you're about to get the value (first one will be the observer civ).

Then look for "PeaceWeight":

Repeat for each leader.

 

[a pen-dragon]

Mao - Zara - Liz - Pericles - Victoria - Hammurabi

I do not have much map knowledge, not having watched this game (yet) and no resource being visible on the maps.

Overall, Mao, Liz and maybe Pericles and Zara seem to have the best starting positions with relatively good land. Hammu and Victoria seem to have significantly worse starts. Since I think Hammu is weaker and has a weaker position, I place him last and Victoria fifth.

Mao first, due to the starting position and being a leader unlikely to be killed on this map, even with the bad PW situation. Liz and Pericles are not big warmongers, the two northern ones are weak, leaving Zara to be a threat. Zara has not that high a peaceweight and I think he also gets warmonger respect, so the will not be enemies. Also Mao does not care about religion, which is probably a plus here, with all others likely to found early religions, sinking turns into it and worsening relations.

I think Zara is quite likely to win lots of wars, so I put him second. the low survival matches with lots of conflict with Mao, who will kill him frequently, but when Zara survives he has an easier access to the north and is likely to win these games.

Third Liz, who also suffers from being close to Mao but is the best techer in this field. Pericles fourth, likely to go culture, but because of spread-out religions I do not think he will be fast, winning less frequently than Liz. Those two could well be reversed.

After looking again at the results of @Thrasybulos , maybe Peri - Mao - Liz - Zara is more likely?

(there was a post about on this forum, but can't find it right now)

here it is

 

[Genghis the Barbarian]

Hammurabi's land is garbage and we can already consider him an irrelevant. 0 finish points but 65% survival rate is pretty obvious.

I think the top 3 are, in order, Zara, Mao, and Liz. Mainly because Mao's spot has been identified as ridiculously strong, but Mao will also get dogpiled by Zara and Liz (and others) to be FTD in some games, and the higher win rates of the two leaders who bookend him are a result of them getting his territory. Something to note is Mao isn't completely screwed - Zara in particular is a great potential ally, and I think many of his 2nd places come in Mao's wins (and vice versa). Kill counts really tell me where Zara and Liz are, and me thinking Mao 2nd is partially based on kill count. I wonder if Mao gets screwed out of multiple playoff spots by the UN...

4th place is Vicky in games where the southwest trio stalemate and go irrelevant and she doesn't go crazy with wonders. 5th place is Pericles - I have two issues with him. First, I don't trust T400+ actual game winners - I learned that from Mao's S1O8 set. A game going that ridiculously long is a sign that something went haywire. (Maybe in the actual game Pericles helping dogpile Liz was atypical) Pericles with Hammurabi in the game is a questionable combo. He wants to go culture, but doesn't have culture tech preferences, and Hammurabi is literally Gandhi-obsessed with culture and wonders. Pericles is gonna find himself en route to T400 culture wins.