[BTS] AI Autoplay Ranking- testing with fair maps

ROUND 59 (no show:Churchill,Hatshepsut,Lincoln,Zara Yaqob)
agg ai
round59.jpg

non agg ai
round59 nonag.jpg
 
SUMMARY 2 for Round(1-65)
Here is the result for these settings are:
Noble difficulty
Option=GAMEOPTION_AGGRESSIVE_AI
Option=GAMEOPTION_NO_TECH_TRADING
Option=GAMEOPTION_NO_TECH_BROKERING
Option=GAMEOPTION_NO_VASSAL_STATES
Option=GAMEOPTION_NO_EVENTS
summary.jpg



Result for these settings are:
Noble difficulty
Option=GAMEOPTION_NO_TECH_TRADING
Option=GAMEOPTION_NO_TECH_BROKERING
Option=GAMEOPTION_NO_VASSAL_STATES
Option=GAMEOPTION_NO_EVENTS
summary non ag.jpg

So here I tested effects of Aggressive AI on and off. And turning it off reallly makes game boring. It is like whoever is smart enough to turn that culture slide %100 will most probably win. I think the tables say it all. Beside there are many undeserving culture victories. The ranking could be better if I have turned culture victory condition off for none aggressive AI games. AIs just sit there doing nothing about random AIs random %100 culture spam.

I would also like to test tech trading on and off but unfortunately my autoplaying AI trades techs with them too. And if you have looked on screenshots of my games in every round you'll see my ice tundra one city empire is at around 200-400 points where when I enable tech trading this goes up to 1700-1800. I need to unmeet the AIs as long as possible and this commodore map is too small for that to happen. And even then it would be unfair for those southern civs that cannot meet me.

And finally here is the combined ranking:
comparison.jpg



About the ranking method, I think kill points turned out to be a good factor instead of just pure luck factor here, which also tells AIs interest in building army and conquering their worst enemies.
There are only 3 games where a leader got 4 kills in a game and those are
Round 50-Game3 Julius Caesar
Round 56-Game1 Hannibal
Round 34-Game7-non ag Justinian

From now on I think I want to try something else for ranking. Maybe a game of 12 leaders perhaps,or I try to remake this map of 6 with more y-axis tiles where I can hide myself from AIs meeting me until someone builds the UN.
 

Attachments

  • civ4 commodoremap.xlsx
    920.6 KB · Views: 128
This thread is hugely interesting for playing Hall of Fame Conquest and Domination victories.

And it' s done well, because Human still can make better choices like when comparing for a Spacerace i. e.
 
This thread is hugely interesting for playing Hall of Fame Conquest and Domination victories.

And it' s done well, because Human still can make better choices like when comparing for a Spacerace i. e.

Thanks, I guess one can select leaders with most kills as their enemy for more challenge.
I left continue doing this because I was annoyed by the map so I had to remake it for better.
-----------------------------------------------------------------------------------------------------------------------------------------

Anyway it sucks I feel like I have to restart for a new ranking only to see how tech trading option affects AI which I may do it in near future out of lockdown boredom. So I would like to share my remake by attaching this map for future AI ranking. Maybe weeks later I share more :)

Error: Capital 2 had its flood plain removed next to river, its plain desert right now, make sure you add that.
 

Attachments

  • 8commodore.CivBeyondSwordWBSave
    262.9 KB · Views: 53
Last edited:
Thank you for proving definitively that I am right to turn culture victory off in all my games.
 
Hello again, I decided that I wanted to do a ranking again since I have actually bothered to edit worldbuilding file to make a fair map of 8 leaders so why not use it. The map Commodore made for Sulla's final game had lots of errors when fresh downlaoded from the zip files and one of the mistakes I forgot to fix is capital 4 had two wines while other starting positions had a wine and a cow.
Beside I couldn't unmet and hide from AIs for testing a tech trading on ranking other than just passing turns one by one instead of AI Autoplay mode. So there I made my own map stealing their design and a bit changing, adding all the missing resources or tiles such as incense or oasis and a few more to the map.

Why 8 is pretty obvious due to grind system and by default you just create maps with tiles multiple of 4 only. I think 8 is better than 6 for ranking. Beside I really don't know how to change map sizes but no need. So with 8 leaders in a game, I could make the map fair.

Now to make ranking more fair I decided to use a seeding system just like Sulla but mine is based on peaceweight.
Here are the pots:

Pot 1 of High peaceweight leaders:
1 Gandhi 10
2 Elizabeth 9
3 Hatshepsut 9
4 Lincoln 9
5 Mansa Musa 9
6 Asoka 8
7 Darius I 8
8 Frederick 8
9 Hammurabi 8
10 Roosevelt 8
11 Sitting Bull 8
12 Augustus Caesar 8
13 Wang Kon 8

Pot 2 of Above avg peaceweight leaders:
1 Victoria 8
2 Washington 8
3 Bismarck 6
4 Charlemagne 6
5 Churchill 6
6 Isabella 6
7 Joao II 6
8 Pericles 6
9 Ramesses II 6
10 Zara Yaqob 6
11 Willem van Oranje 4
12 Saladin 4
13 Suleiman 4

Pot 3 of Below avg peaceweight leaders:
1 Julius Caesar 4
2 Justinian I 4
3 Cyrus 3
4 Boudica 2
5 Catherina 2
6 Gilgamesh 2
7 Hannibal 2
8 Huayna Capac 2
9 Mehmed II 2
10 Pacal II 2
11 Qin Shi Huang 2
12 Shaka 2
13 Stalin 2

Pot 4 of Low peaceweight leaders:
1 Kublai Khan 1
2 Louis XIV 1
3 Mao Zedong 1
4 Peter 1
5 Suryavarman II 1
6 Tokugawa 1
7 Brennus 0
8 De Gaulle 0
9 Genghis Khan 0
10 Montezuma 0
11 Napoleon 0
12 Ragnar 0
13 Alexander 0

So basically in every game there will be 2 leaders drawn of all four pots to same game so leaders in same pot have a chance to appear on same game too. I objectively put Willem to pot 2 since he totally sits down and build culture'n stuff and Julius caesar to pot 3 since he is a total warmonger. I decided that 3 american leaders shouldn't be in same pot so I moved washington to pot 2 from 1 since he seems less into teching or cultural winning. And I subjectively decided Victoria and Justinian to be in lower pots, I think they like fighting more than other same peaceweight leaders which is making them more evil.

Now their starting positions are also going to be same.
Leaders in Pot 1 will always start in capital 1 or 2.
Leaders in Pot 2 will always start in capital 3 or 4.
Leaders in Pot 3 will always start in capital 5 or 6.
Leaders in Pot 4 will always start in capital 7 or 8.

So this means everyone will have a warmonger next to them, there is no hiding from fight, surrounding themselves with allies. And there won't be any unlucky loser in the ranking consistently ending up a game full of warmongers etc, being worst enemy of everyone since turn one.

The high peaceweight leaders have the advantage of turn order for finding religions or building wonders at same turn but I don't think this will be a very big issue.

For every round a leader from every pot cannot appear on any games so to complete the cycle I make 13 rounds where every leader appears in 12 games.
ff.jpg


I will make multiple different series, replaying same games with different settings once I am done with 13 rounds 78 games of a certain settings for a comparison later. It's the same story as what I have done before but with more settings this time and only 13 rounds for now. I set my excel file ready. And the very first game's world building file.

I will try to record the first 78 games for the very first settings no tech trading&aggressive AI games but not the others. And I definitely can't stream their tech trading on versions as I am suppose to not meet them for a very long time until UN. So don't expect much.
I will give 5 points to winner, 2 for runner up and 1 for kill credits. Stuff that you know.

Ok, I hope this won't be a terrible idea. And more to track other than Kjotleik's and Sulla's very nice content :)
I will post once I am done with every 6 game (1 round) to here as results.

And more thing, I disabled barbarians, they disgust me and another luck factor in the game against fairness.
 

Attachments

  • 8 Ranking.xlsx
    74.8 KB · Views: 64
  • R1 G1.CivBeyondSwordWBSave
    262.8 KB · Views: 39
Seeding based on peaceweight is a new one. I don't think I've ever seen that before. It will be interesting to see how things shape up now that you also pre-arrange the peaceweight-groups in fixed positions on the map.

This is most assuredly not a terrible idea. Looking forward to see if the settings leads to predictable results or not. If they do... there must be something we can learn from this.
 
Spreading out peace weight like that is very sensible. Looking forward to seeing how it turns out.
 
I will keep spamming here these graphs for every round.
So I ended up not to post after every round as it is waste of time to do that. Instead you can check the 4 times speed videos anyway if you want to see what happened in the games.

I am done with all 78 games so far and I want to share with you right now.

Results for A Series (Round 1-13)
where settings are like this:

NOBLE DIFFICULTY
Era=ERA_ANCIENT
Speed=GAMESPEED_NORMAL
Calendar=CALENDAR_DEFAULT
Option=GAMEOPTION_NO_BARBARIANS
Option=GAMEOPTION_AGGRESSIVE_AI
Option=GAMEOPTION_NO_TECH_TRADING
Option=GAMEOPTION_NO_VASSAL_STATES
Option=GAMEOPTION_NO_GOODY_HUTS
Option=GAMEOPTION_NO_EVENTS
Victory=VICTORY_TIME
Victory=VICTORY_CONQUEST
Victory=VICTORY_DOMINATION
Victory=VICTORY_CULTURAL
Victory=VICTORY_SPACE_RACE
Victory=VICTORY_DIPLOMATIC
GameTurn=0
MaxTurns=500
MaxCityElimination=0
NumAdvancedStartPoints=600
TargetScore=0
StartYear=-4000
Description=
ModPath=

A-Ratings.png


This settings favored unit spammer warmonger AIs and completely destroyed high pw leaders so far. Before blaming on map or fixed positions I'm going to replay the same 78 games 3 more times, B,C,D series where aggressive AI turned on and off and tech trading enabled. I suspect faster paced games favor infrastructure building techers while slower games favor unit traning warriors. I am not going to try right now maybe more series E,F? but probably increasing difficulty helps high peaceweight leaders more as well as increasing game speed. Quick and immortal or deity favors cultural or techer leaders while settler chieftain marathon games favor warmongers? I use normal speed noble difficulty
I am most certainly doing 3 more alternative series but if I plan to do another 13 rounds new games (14-26) I will definitely change the way I made their fixed starting positions to something else. Or just try a new map etc.

As for results I sorted them as their win rates, number of games they have won. After all winning matters only. The power ranking has flaws. Around half or one third of the kill credits are stolen or not deserving, same goes for runner up points too.
Therefore I sorted them as number of victories, then power ranking, then survival rates such as how many games they died vs survived and then if all still same then the average turn number they are eliminated. So at least this can compensate unlucky with the power ranking leaders a little.
 

Attachments

  • A series of 8(1-13).xlsx
    153.7 KB · Views: 50
Last edited:
Well done finishing the series. I'm so far behind watching these on YouTube at the moment. I'll catch up eventually.

Fourteen leaders with no wins at all, none of which survived more than 50% of their games. With Shaka, Montezuma and De Gaulle among those, it is clear not all warmongers have what it takes to win despite being helped by peaceful neighbours.

My biggest surprise is Joao II being on zero wins. He has done well in previous AI vs AI rankings. In your early one (the Adjusted Average Overall Ranking [48/60 games]) he was in 12th, and has been consistently towards the top in your other ranking-lists as well.

Any insights learned as to why he has done so poorly this time around?
 
Well done finishing the series. I'm so far behind watching these on YouTube at the moment. I'll catch up eventually.

Fourteen leaders with no wins at all, none of which survived more than 50% of their games. With Shaka, Montezuma and De Gaulle among those, it is clear not all warmongers have what it takes to win despite being helped by peaceful neighbours.

My biggest surprise is Joao II being on zero wins. He has done well in previous AI vs AI rankings. In your early one (the Adjusted Average Overall Ranking [48/60 games]) he was in 12th, and has been consistently towards the top in your other ranking-lists as well.

Any insights learned as to why he has done so poorly this time around?

These games are not something we are very familiar since probably we all play as tech trading enabled. I need to have all 4 versions done first. Even then the results would be still based on very small sample and one map only. So consider all of this as a start I will end up completing over a year or so with nex maps and more rounds etc. :)

As for Joao, he is often the next target once the highest 2 peaceweight leaders gone in this setup. He will tech democracy before millitary science or replacable parts while all the warmongers skip philosophy and music to reach rifling or millitary science and porbably always before Joao and take him down. We al witness AIs like Joao have researched entire renaissance entry techs Education,nationalism,astronomy,printing press while loser player me struggling to finish engineering,philosphy and then lose track of teching at that era every time.
Anyways, therefore I need to enable tech trading to find out how the games turn out in C,D series first.
 
Results for B Series (Round 1-13)
where settings are like this:

NOBLE DIFFICULTY
Era=ERA_ANCIENT
Speed=GAMESPEED_NORMAL
Calendar=CALENDAR_DEFAULT
Option=GAMEOPTION_NO_BARBARIANS
Option=GAMEOPTION_NO_TECH_TRADING
Option=GAMEOPTION_NO_VASSAL_STATES
Option=GAMEOPTION_NO_GOODY_HUTS
Option=GAMEOPTION_NO_EVENTS
Victory=VICTORY_TIME
Victory=VICTORY_CONQUEST
Victory=VICTORY_DOMINATION
Victory=VICTORY_CULTURAL
Victory=VICTORY_SPACE_RACE
Victory=VICTORY_DIPLOMATIC
GameTurn=0
MaxTurns=500
MaxCityElimination=0
NumAdvancedStartPoints=600
TargetScore=0
StartYear=-4000
Description=
ModPath=

B-Ratings.png


Hi again. I'd like to share with my results with you, hope this would be fun for you to have here.
This settings made even winning a game is somehow undeserving for some games. Someone not suppose to win still wins the game just by turning culture slide on while the deserving leader gets stuck on space race that'd be longer to achieve because of no tech trading. And they can't attack each other due to pleased relations too. Or that they need to finish space race first to be able to plot war. However it was still interesting to have this for comparison at the very end when I finish with two more series of this. I will make my conclusion then. Anyway, aparently no aggressive AI alone was not enough for higher peace weight leaders to perform better. Maybe tech trading will make the difference. I shall see. Overall same low peace weight leaders that can turn on culture slide on even without wonders in their cities won absurdly too often. This ended up badly unbalanced, going for culture is around 60-70 turns faster than space race, this made leaders with no culture flavor look like a sucker D:
I will be back for C series where finally tech rading will be on.
 

Attachments

  • B series of 8(1-13).xlsx
    143.2 KB · Views: 52
Last edited:
Results for C Series (Round 1-13)
where settings are like this:

NOBLE DIFFICULTY

Era=ERA_ANCIENT
Speed=GAMESPEED_NORMAL
Calendar=CALENDAR_DEFAULT
Option=GAMEOPTION_NO_BARBARIANS
Option=GAMEOPTION_AGGRESSIVE_AI
Option=GAMEOPTION_NO_VASSAL_STATES
Option=GAMEOPTION_NO_GOODY_HUTS
Option=GAMEOPTION_NO_EVENTS
Victory=VICTORY_TIME
Victory=VICTORY_CONQUEST
Victory=VICTORY_DOMINATION
Victory=VICTORY_CULTURAL
Victory=VICTORY_SPACE_RACE
Victory=VICTORY_DIPLOMATIC
GameTurn=0
MaxTurns=500
MaxCityElimination=0
NumAdvancedStartPoints=600
TargetScore=0
StartYear=-4000
Description=
ModPath=


C-Ratings.png


I just finished C series and would like to share. Same 78 games with different options replayed again. Options really effect how well AIs do very much. Even though the only options I mess with are no tech trading and aggressive ai. I just don't know if I should mess with vassal states for an E series yet. I'm going to try to analyze who does better in what settings after finishing D series but you can tell no tech trading option totally killed high peaceweight leaders generally and now that it is gone they started to do better. Expect me to be back here for results of D series and a conclusion and who knows, new maps!
 

Attachments

  • C series of 8(1-13).xlsx
    146.2 KB · Views: 53
Last edited:
Top Bottom