[BTS] Sullla's AI Survivor Season Eight - Post Season Thread

[Noble Zarkon]

Thread opened to discuss anything and everything about this season,

 

[ManiaMuse89]

Well my picking was absolutely awful this season (compared to season 7 when I was in the mix for overall victory until I fluffed it in the final).

Was an interesting season. Definitely a few 'lucky' winners *cough Hammurabi*. Would have hoped for a few more warmongers to make it through to the playoffs/final.

A bit disappointed that Sulla didn't run the mega wildcard game, I was looking forward to that.

I know Sulla tends to get a bit jaded from running these by the end of the season but hopefully he runs another season sometime next year. I would love a bit of a 'twist' next time but I know he likes to keep things consistent.

 

[Keler]

First of all, Congratulations Elizabeth AI!!!

From bizarre outcome of Hammurabi game one victory to large final map, the first think I would like to discuss as it is a hot topic right now is map making and how you would do it.

I personally have no experience but some attempts. In theory I just copy paste the data of regular pangea map from worldbuilder file to excel and group all the units(parameters). Bonus types, Plot types, Terrain Types, Feature Types and number of Rivers. And try to make a map that has similar amount of such values.
Does anyone here have ever created a custom map?

 

[Eauxps I. Fourgott]

I know Sulla tends to get a bit jaded from running these by the end of the season but hopefully he runs another season sometime next year.

Not sure if you saw the end of the stream, but for anybody who didn't, Sullla said that he would love to do Season 9 next year! However, Baby #2 is coming at the end of this year so feasibility of doing it is up in the air.

 

[Plains-Cow]

This season was wild. Lots of great games; I'll be riding the joy of those Sitting Bull and Ramesses games for awhile.

 

[Eauxps I. Fourgott]

Another fun season. As has been observed plenty, this year the defining feature was the rise of the high peaceweights. I'm glad they got their time in the sun for once and it will certainly make this season distinct.

This was also my first time hanging out here on CivFanatics, and I have to say it was a great experience. My welcome was warm, my continued interactions with you all were good, and I proved to be correct that the forum environment was better for the types of analysis/contribution that I like to do. Thanks to all, and I plan to continue to hang around for AI Survivor discussions here.

Alternate histories for this season should be quick to fill out. There's nine games remaining to cover, but five of us who will be working on them, so a light workload per person. I myself will be bringing coverage for the Championship first, and then Game 8 (aka the game with my highest bid leader).

The picking contest... the less said about my performance there, the better. But it's by far the aspect of the competition that least interests me. The most engaging part of that is finding out each week if my written prediction made Sullla's preview. This year was better for that than previous ones.

As for fantasy, the main point of that for me was to have a good time in the fantasy auction. That was a success, so I'm overall happy with it even if my final result was disappointing. (At least I finished in Not Last Place!) While it was fun to get another chance to do that (three years after the first time!), I do hope to see a roster of fresh faces (and perhaps the reigning champ) next season, rather than a bunch of returning coaches again. Some CFC folks should sign up next time!

While I don't think that the AI Survivor formula is perfect, I do think it works very well on the whole, so I'd rather err on the side of keeping it static than mess it up with too many changes. Especially changes to the core format of the games themselves. Fortunately I think Sullla is of the mindset of "no need to change it up a lot", so the danger of this is low.

I'll probably stop doing game writeups for the website next season, as long as there are others like Atesh who want to contribute. It feels like I've about run my course with them and it's the for me to step aside for new blood. One thing that occurred to me as potentially fun is instead writing up play-by-plays, basically a long string of bullet points calling the action and current situation as it comes, rather than a prose summary. I don't know if there's any actual interest in that though.

I updated the "Past Performance" sections of the leader previews on the website for this season, and intend to do so again to reflect the changes after this season's results. What I'm wondering if I should do is also update the personality summaries in those as well. I didn't think it necessary last time, but now I think those could benefit from the increased knowledge of what's significant in AI Survivor, and the culling of any outdated references (like I think Stalin's still implies that he's been playing like a monster, which hasn't been the case for five years now). If anybody's made it this far in my post, I'd like your input: should I update those as well before next season?

I think that's all from me for now. I'll no doubt have more random thoughts that come to mind later.

 

[Noble Zarkon]

If anybody's made it this far in my post, I'd like your input: should I update those as well before next season?

Yes please, if you have the time to ensure everything is up to date that would be great.

 

[The Rook]

I updated the "Past Performance" sections of the leader previews on the website for this season, and intend to do so again to reflect the changes after this season's results. What I'm wondering if I should do is also update the personality summaries in those as well. I didn't think it necessary last time, but now I think those could benefit from the increased knowledge of what's significant in AI Survivor, and the culling of any outdated references (like I think Stalin's still implies that he's been playing like a monster, which hasn't been the case for five years now). If anybody's made it this far in my post, I'd like your input: should I update those as well before next season?

Sounds cool! If you are updating the "personality" section, then one way I think it could be significantly improved is by removing irrelevant details such as: "asks for help", "demands" and "demand penalties". For example, under Cathy it says:

"She has a unique change to her diplomacy in that refusing one of Cathy's demands will give a -2 malus as opposed to the usual -1. This applies to other AI leaders too, and Cathy often ends up hating nearly everybody else in the game because of it. Naturally, she has high likelihood to demand all kinds of tribute, with numbers like 8/10 and 10/10 for various types of demands."

I'm 99% certain (and I'm sure someone here will correct me if I'm wrong ) that the AI never "demands" anything from other AIs. I mean when do you ever see modifiers like "you made an arrogant demand" or "you gave us help" in the Survivor diplo screen. AFAIK, the AI will always aim for "fairish" trades between themselves, and if they can't trade fairly then they won't trade. For this reason war bribes are very rare in Survivor, as a large pile of gold is the only way they can make it happen.

 

[Fippy]

War bribes are really expensive if you try them against an AI who isn't at war yet.
(so players should declare first even if they are not ready yet, and get their "ally" in for much less)

But they get pretty cheap once they are..i think ~600 was usually enuf in later stages.
Deity AIs often have that amount ready, they usually keep some back.

Early i have seen them join for stuff like writing + sailing
And god forbid they are at war with 2+ AIs already..you can bribe more in for almost nothing.

Aggressive AI changes a lot of things as well, and we see funny things as result.
They are more likely to sign peace now..which i think is why we see those hilarious "peace for city" deals so often.

 

[Eauxps I. Fourgott]

Conclusions page is up on Sullla's website: https://www.sullla.com/Civ4/civ4survivor8-14.html

 

[lymond]

I was just skimming through that write-up. Most specifically the new rankings. Hatty still has 0 points in the entire Survivor series. She is actually behind the Apostolic Palace, which was somehow awarded a point for a kill. I'm trying to ascertain how the AP could get a kill, and the only possible scenario is that a vote was taken to return a city that happened to be a leader's last remaining city. I did not even know what was possible, but I do not remember that happening. Maybe someone else here remembers it.

 

[AcaMetis]

the only possible scenario is that a vote was taken to return a city that happened to be a leader's last remaining city

Don't remember which game it was, but yes, that is exactly what happened. Chat was having a time pointing out the leader in question didn't defy that resolution.

 

[Thrasybulos]

S4-Playoff 2 iirc. Mao or Toku.

Wasn't it also the game where Shaka won an "oops" AP Diplo victory?

 

[lymond]

Don't remember which game it was, but yes, that is exactly what happened. Chat was having a time pointing out the leader in question didn't defy that resolution.

So the leader voted themself off the island

 

[lymond]

Shout-out to CFC from Sulla, and to My (kinda ):

A special shout-out to the CivFanatics contingent who once again added their own contributions with weekly forum threads and playoff betting odds to spice things up further!

 

[Noble Zarkon]

I'm trying to ascertain how the AP could get a kill, and the only possible scenario is that a vote was taken to return a city that happened to be a leader's last remaining city.

Exactly, I've done this before:-

I'm playing the Large map now and had something really funny happen - have upped the Domination limit to 76% via killing every original AI (only AI left is a former vassal of Hannibal) and setting up 16 vassals. One of them is Asoka but both his cities are under enormous cultural pressure due to Sushi. I have avoided Mass Media so the AP is still active and a vote came up to assign one of his cities to me which succeeded! He was left with one city and yet again the assign city vote came up, I was sure he would Defy but he didn't he simply voted No. The vote was successful so his last city was assigned to me and he's dead!! He's still on the scoreboard so the Dom limit is still 76% - I'm sure @WastinTime or someone can come up with a really evil way to exploit this!!

Challenge-XIX-10

While the general Hall of Fame is an ongoing competition, we are running a series of ten games called the Hall of Fame Challenge Series - "Fall of the Inca II". Standard Hall of Fame rules (*) still apply, but your best submission meeting the settings of one of the games will be counted towards...

forums.civfanatics.com

 

[antimony]

Conclusions page is up on Sullla's website: https://www.sullla.com/Civ4/civ4survivor8-14.html

For the trait rankings, it would probably be better to look at how the median (say, 5th ranked) AI with that trait is doing. Looking at the median, FIN/CRE/IMP are clearly the top traits. SPI, just like IND, is pulled up by its top leaders, whereas on the other end, the mean result for CHA suffers from not having any dominant leader (Hannibal's score feels low, he clearly hasn't had the best luck of the draw). The CHA team is also strange in that the 2 Celt leaders have had decent scores whereas the 2 French leaders (De Gaulle and Napoleon) have underperformed with clearly stronger starting techs.

 

[Thrasybulos]

For the trait rankings, it would probably be better to look at how the median (say, 5th ranked) AI with that trait is doing. Looking at the median, FIN/CRE/IMP are clearly the top traits. SPI, just like IND, is pulled up by its top leaders, whereas on the other end, the mean result for CHA suffers from not having any dominant leader (Hannibal's score feels low, he clearly hasn't had the best luck of the draw). The CHA team is also strange in that the 2 Celt leaders have had decent scores whereas the 2 French leaders (De Gaulle and Napoleon) have underperformed with clearly stronger starting techs.

You're certainly not the only one expressing that kind of opinion.
But it seems to me to boil down to:

• Traits are important for the AI
• Except where they're not

I would suggest an alternative: traits don't really have any meaningful impact on an AI's performance.

 

[Noble Zarkon]

Hannibal's score feels low, he clearly hasn't had the best luck of the draw

He's a leader who looks good on paper but I always struggle playing as him for some reason.

traits don't really have any meaningful impact on an AI's performance.

Wel lthat's a take I guess lol, care to expound your theory?

 

[Thrasybulos]

We assume traits are important, and when the data doesn't support that assumption, we try to make it fit nevertheless ("that trait is weak, but it ranks high because some leaders are strong in spite of it", "that trait is strong, but it ranks lower because leaders with it have been unlucky", etc...).
Just putting forward that maybe, just maybe, what the data shows is that the assumption is wrong.