March Madness: Civ4 AI Edition

[Jkaen]

Winner: Pacal - I think the central location will let him spread his religion quickly as well as grabbing land, meaning diplomacy wise he should make up for his 'evil' rating. If the religion doesnt spread right however he is screwed, especially when his evil friend dies, so this is a bit of a gamble.

Runner up: Caesar - I think he will gain the most from the first war (see below) and snowball on from that

First out: DeGaule - Aggressive good rome will attack evil demanding degaule, who will then also get piled on by pericles from the south

Surviving: 3 - Because it always is!

Victory: Diplomatic T308

[Jkaen]

doh missed this one too

[Sullla]

Alright, once again thanks to everyone who submitted predictions for this game. Before getting to the tallies of the results, let's look at some of the more interesting written predictions.

Winning Leader: Elizabeth
Second Place: Pericles
First to Die: Frederick
Surviving Civs: 3
Victory Type/Date: Domination t310

Pacal and Julius will double team Frederick and kill him off pretty quickly. But this won't give either of them much land to work with. Meanwhile, De Gaulle will butt heads with Pericles and fail miserably, and Elizabeth and Darius will take advantage of the large amount of land to the west and expand far and wide. Liz and Darius will then double team Pacal, Liz taking the lion's share of the land, while Pericles with the help of Julius kills De Gaulle. Liz will continue up through Rome with Pericles support, and win by domination.

Azza successfully picks the two leaders moving on to the next round (in backwards order), although misses on almost everything else.

Winner: Elizabeth

Runner-up: Pericles

First to die: Frederick

Surviving Civs: 3 - is a magic number.

Victory type: Space

Victory date: T309

Elizabeth wants to be left alone, and that she shall. Something will happen with the middle crowd - Pacal will beat up frederick, Jules will stab him, Darius will pile on ze Germans. Meanwhile the outer rim civs - will tech to an advantageous position. Especially Liz, who will be able to field redcoats when the going gets tough. Not for her, that is.

Pericles will be behind a lot, but will slip into runner-up place because someone will decide that they don't really want to finish second and that they would rather do or die instead. They will die. Is this you, Jules? Perhaps even Pacal, depending on how the early stab-backstabbing turns out.

Molach saw things playing out in nearly identical fashion. The two of you must have been pretty disappointed in Lizzy building the United Nations, otherwise these predictions would have been almost perfect!

Winner: Pacal
Runner-up: Elizabeth
First to die: De Gaulle
Surviving civs: 4
Victory Type: Space
Victory Date: T309

Caesar's the most likely fighter and the strongest in the early game, so guessing where he'll clash first is key. Frederick has that wall of mountains limiting access between the two of them, which should prevent a long border from forming, and probably won't make for great results if Caesar does attack that way. I'm guessing Caesar winds up at war with De Gaulle instead, due to some combination of De Gaulle pushing west and stupid demands, Pacal takes the chance to get a large enough glob of the center for his traits to carry him to victory, and Elizabeth techs away happily on the more peaceful west coast.

Outside of Pacal winning the game, everything else was spot-on in this prediction. Katon already had the lead in overall score coming into this game, and the field didn't make up much ground in this one.

Winning Leader: Pericles
Second Place: Julius
First to Die: De Gaulle
Surviving Civs: 4
Victory Type/Date: Culture T306

My thoughts:

Hmm, an interesting one. In my own tests Elizabeth and Darius have never done very well, but here they're in the backlines so they should have a better chance. Pacal usually does very well and often wins with a smaller empire, but here he's in the middle of the map where his religion can't spread to all his neighbours, and he could easily crumble to an attack. Frederick and De Gaulle are the only ones who seem to consistently do badly. Pericles tends to either do very well or very poorly, here I think with his position he's got a good chance. Julius usually pulls a solid but not brilliant game, here he might have an opportunity to rise from the middle of the pack due to multiple weaker neighbours.

Also not far off, a possible outcome where Caesar didn't stupidly pick a fight with Pericles.

Winner: Caesar
Runner-up: Elizabeth
First to die: De Gaulle
Surviving civs: 3
Victory Type: Domination
Victory Date: T300

I think that Caesar will kill De Gaulle early on since both are somewhat antisocial. Caesar tends to do well in my experience. The map also gave him some back lines which is nice but at the same time surrounds him with 3 other AIs. It seems likely then that he'll invade and kill someone. I'm not sure how well Lizzie will do, her land kinda sucks and I'm not sure if she'll be able to outsettle/conquest towards all that grassland. I have no idea about Pacal as he might have terminal wonder whoring, he'll probably win in a spaceship on turn 312 or something if he gets rolling.

Another good entry derailed by Caesar's overaggressive play. What is it with the AI and foolish war declarations?

Winner: Julius Caesar (4 of 6 in alphabetical order)
Runner-up: Charles de Gaulle (2 of 6)
First to die: Lizzie (3 of 6 and god hates the queen don't'cha know)
Surviving Civs: 5
Victory type: Domination (3 out of 6 alphabetical order, Conquest out due to the no. surviving civs. I'd actually like to see this happen with five civs. I included religious diplo as a seperate win condition to UN, because frankly it is.)
Victory date: T389 (I set a minimum of 200 turns and a max of 500)

We'll see what happens here. Hopefully the blindfolded monkey will beat the market again!

The all random prediction in action! Does generating random numbers work better than trying to make an informed guess? Let's find out in the next post.

[Sullla]

Our results from Game Six:



- We had 39 entries this time, another strong showing. I added Jkaen's predictions when I was tallying them, since there didn't seem to be any major reason not to do so.

- Opinion was fairly split on the game winner. There were quite a few picks for Caesar, Pacal, and Elizabeth. Pericles only landed 6 votes. That was better than Darius, who had only 1 vote, and Frederick/De Gaulle, who had zero.

- Elizabeth was oddly popular as the runner up selection. A lot of people thought that she would wind up behind the game winner, but still sneak through, and that turned out to be completely correct. Nice collective work from the field. Caesar was also a popular choice here.

- Frederick was the consensus favorite as First to Die, with 21 picks (54%). I don't think this was a bad choice either, as he was a likely target for Caesar. De Gaulle was the next most popular option, and turned out to be correct. The two of them together dominated this category.

- Somehow, despite every previous game ending with 3 leaders surviving, the most popular pick in the next category was 4 survivors. And that was correct! Yeah, it was only by a single choice (17 picks to 16 picks) but that's still amazing. Nearly everyone had either 3 or 4 leaders to survive here.

- Spaceship was the clear choice for victory type, with 20 entries favoring that victory path. Again, we nearly saw the game end that way, so it wasn't exactly a bad choice. Domination was next most popular (8), then Diplomatic (6), and then Culture (5).

- We saw the last couple games ending around Turn 310 or so, and that guided the picks for victory date. The early Diplomatic victory on Turn 289 resulted in the entire field overestimating the date. Not even one choice was "too low" this time. Gavagai was one turn off with his Turn 290 submission, and they all escalated upwards from there. Over two-thirds of the picks were in the 10-25 turns off range, highly concentrated around roughly Turn 310.

Now for the individual submissions:

* Katon, Hesmyrr, and Lord Parkin tied for the highest scoring game with 16 points each. This was the same as our top scores from Game Five. Congratulations to each of you.

* Jowy was the only one to get both of the leaders advancing to the playoffs correct. Azza and Molach had them in the backwards order, our special case worth half points. Five entries had De Gaulle as First to Die along with one of the other two leaders to advance.

* This was the tightest concentration of victory dates we've had yet, with only 69 turns separating Gavagai's Turn 290 prediction from Becko's Turn 359 prediction. I'm not counting the RNG created Turn 389 number here.

* Speaking of RNG, randomized scoring did not seem to work, as Brian Shanahan's entry produced exactly zero points. Maybe we have a little longer until the cyborgs take over.

* Katon tightened control over first place position, by having the highest score coming into this game and then also scoring the most points. Hesmyrr is only two points behind, and then there's a large concentration of scores following behind von Adlercreutz and Ceiliazul.

* Per-game scoring average seems to be roughly 8 points from what I can tell. It's hard to track exactly due to the missed entries from a lot of contestants.

Tomorrow, we'll do the draw for Game Seven and get ready for the final rounds before the playoffs begin. Thanks again.

[antisocialmunky]

Well to be fair, I thought Caesar would have gone after Pacal or Fredrick rather than Pericles.

[Hashoosh]

Judging by the AI's performance in these games, I'd say we have a very long time indeed before the cyborgs take control.

[Jowy]

If you can get over your hate for Civ5, it could be interesting to see how that AI performs in these AI vs AI games. There's actually an observation mode built-in :P

[TheSunIsDark]

No, AIs don't diplo vote for spite. They just vote for their highest rated diplomatic pal, minimum +8 including hidden modifiers like iBasePeaceWeight (this is the good vs evil "warmonger scale" to which Sullla keeps referring.)

I seem to remember (reading your reports from your site) that you forgot about the hidden diplo modifiers in your Adventure Fourteen report. The post here was correct though.

[superjm]

If you can get over your hate for Civ5, it could be interesting to see how that AI performs in these AI vs AI games. There's actually an observation mode built-in :P

I've actually done this already a few times.

Spoilers: The AI is not very good in that game either!

[Sullla]

Alright, our draw for Game Seven is underway. Make your picks from the following group of leaders.


Catherine

Catherine has one of the most recognizable AI personalities in Civ4. She is the flirt, the tease, the constantly-demanding-something high maintenance girlfriend. Catherine's traits are Creative and Imperialistic, one of the best pairs at claiming land and spreading out rapidly across the map. Her Russian civilization brings the Cossack unique unit and the Research Institute unique building. Catherine was significantly stronger back in pre-expansion days (when she was Financial / Creative and before the cossack nerfs), but she's still a reasonably strong leader to this date. Catherine the AI has culture and military flavors, as well as the "flirty" personality that I mentioned above. Catherine loves to ask for help (10/10 rating) and issue demands for tribute (8/10), and if you refuse any of these things, you receive a special -2 diplo penalty not present for any other leader. She's always badgering her neighbors about something. Catherine AI otherwise has mostly middling scores, although her aggression rating is slightly higher than average at 6.7 out of 10. She's considered an "Evil" leader in peace weight, the goodly AIs will not like Catherine. Finally, there's one other thing that makes her unique: Catherine AI is the only leader in the game who can be bought into a war against someone with whom she has "Friendly" relations. She can be bribed into a war against anyone, regardless of diplomatic rating. All in all, she's quite the package.


Charlemagne

Charlemagne has one of the worst trait pairings in the game, combining Protective trait with Imperialistic trait. It's just as terrible as it sounds. His civilization is quite a bit better; as the only leader of the Holy Roman Empire, Charlemagne has access to the mediocre Landknecht and the amazing Rathaus unique building. That super courthouse is the main reason why the HRE civ ever gets played online, despite its atrocious starting techs (Mysticism and Hunting). Charlemagne the AI has military and religious flavors, and he can be quite the aggressor when provoked, with a rating of 7/10. Charlemagne quite commonly founds his own religion, and he feels very strongly about it, attaching a major penalty to the leaders who follow other faiths. His scores are otherwise average across the board, and Charlemagne can be an excellent ally if you share his faith. He's rated in the middle as a "Neutral" leader on peace weight. Chalemagne's fate in this game will likely be heavily tied to how the religious diplomacy shakes out.


Gandhi

Gandhi is arguably the most unique leader in the entire game, and his AI personality differs in several respects from everyone else. Gandhi has Philosophical and Spiritual traits, a decent second-tier pairing with a pretty good economic boost. He gets the Indian civilization, with the single best unique unit in the entire game in the Fast Worker, although the AI doesn't know how to use them properly. India's unique building is the Mausoleum, a jail replacement which is rather good itself. Gandhi the AI plays the game as a hyper-pacifist, as befitting his historical personality. Gandhi is the AI leader least likely to declare war in the whole game, with a normalized aggression rating of zero out of ten. He's the benchmark upon which all other peacenicks are judged. Gandhi only has one flavor for his research: CULTURE. He can often ignore military techs to an alarming degree. Gandhi's other ratings are a similar story in contrasts. He places no emphasis upon building units (0/10 rating), or espionage spending (0/10), or demanding tribute (0/10). He will ask for help (8/10) and demand civic changes (10/10) at a high rate, but even in those cases, Gandhi has a unique diplomatic mechanic where refusing his demands does not lower his opinion of you. However, Gandhi also has a unique mechanic whereby he places a -2 diplo penalty on anyone who declares war against someone he regards with "Pleased" or "Friendly" ratings. He has the highest peace weight in the entire game, a perfect 10/10, indicating quasi-divine morality. No one else has the same peace weight in the whole game. Overall, Gandhi is about as extreme as it's possible to get on the pacifist scale. He's extremely good at pursuing Cultural victory... but he has to survive to get to that point, and that's quite often his problem.


Saladin

Saladin was another leader who was gutted by the Civ4 expansions, when his traits were changed to their current pairing of Spiritual and Protective. That Protective trait is the kiss of death for leaders. He is the only Arabian leader, and brings the Camel Archer and Madrassa unique items. Arabia has arguably the worst pair of starting techs in the game, the combination of Mysticism and The Wheel, neither of which are tied to improving any food resources. Needless to say, Arabia isn't exactly a popular civilization for our Multiplayer events. Saladin the AI has military and religious flavors, and he's another personality who places a lot of focus on religion. Saladin likes to demand that other leaders convert to his religion (8/10), and he places a major penalty on leaders with rival faiths. His favorite civic is also Theocracy. Saladin is otherwise average in aggression rating (5.5 out of 10) and in the middle of the peace weight scales as a "Neutral" figure in alignment. He'll try to found his own religion, then like anyone who has it, and hate anyone who doesn't. Pretty easy to understand.


Stalin

Stalin is the other Russian leader in this game, here to compete against Catherine for top dog. Stalin's traits are Aggressive and Industrious, a bit of an odd pairing without too much in the way of obvious synergy. Stalin also employs the same Russian unique stuff as Catherine, with the Cossack and the Research Institute. As an AI, Stalin predictably has military and production flavors. He is unsurprisingly rated as a rather aggressive leader (7.6 out of 10) and as an "Evil" leader according to peace weight. Stalin and Catherine might turn out to be good friends in this game via peace weight, and certainly neither one will be able to stand the sight of Gandhi. Interestingly, Stalin AI doesn't care much about religion at all, as befitting a good Communist. He gets few benefits from shared faith and little in the way of penalty from differing religions. Otherwise, Stalin's numbers are pretty average across the board, except for his massive love of espionage spending (10/10!) It's a fairly well constructed AI personality based on the historical persona. In a game where so many other AI leaders will be battling over religion, Stalin will opt out of the race altogether and do his own thing.


Victoria

The last leader in this group, Victoria will hope to perform as well as Elizabeth in our previous game. She is the only Financial leader in this match, pairing that alongside Imperialistic trait. This is one of the weaker combos involving Financial, but you can't really go wrong with the game's strongest trait. As an English leader, Victoria has the Redcoat unique unit and the Stock Exchange unique building, both of them well above average. Victoria AI has gold and growth flavors for her research emphasis. Her numbers are middle of the table for virtually everything, including her aggression rating of 5.1 out of 10. Victoria AI doesn't appear to get too high or too low in any one category. She's another leader who doesn't care very much about religious issues, and she's considered to be a "Good" leader according to peace weight, if not so high as Gandhi. Victoria is a fairly bland AI, without much in the way of defining characteristics. There's not too much that can be said about her.

(For the curious, by process of elimination this means that the final game draw has Asoka, Bismarck, Churchill, Cyrus, Hannibal, Mao, and Wang Kon. I will post write-ups for them after we do Game Seven.)

This game should have a very interesting opening set of turns, with three different leaders having Mysticism starts (Charlemagne, Gandhi, Saladin) and likely all pursuing religions. Where and how those religions appear will probably have a huge effect in terms of how this game progresses. In particular, Gandhi is going to need to line up some religious allies to protect him from his more aggressive neighbors. If the game ends up being mostly peaceful, then I'd favor Gandhi to win via culture or Victoria to take this thing via space. If we get a more violent game, then Charlemagne or the Russian duo will be the favorites to take this one. As always with these games, it's almost impossible to predict what the crazy AI logic sequences are going to cause them to do.

Starting map, spoilered for size:

Here is our current bracket after six games played:



Same categories for predictions as before:

Winner
Runner Up
First to Die
Number of Leaders Surviving
Victory Type and Victory Date

Two days to get those entries in, with Game Seven closing on Wednesday evening my time. Good luck to everyone again.