March Madness: Civ4 AI Edition

[Sullla]

I'm not sure that we have to put every comment under the spoilers tag. While the idea is nice, it does make it hard to read the thread.

Anyway, I thought we'd look at some of the predictions from earlier in the thread along with the results. It's always interesting to see everyone's thought process in action, although of course there's no obligation to write out a longer entry. For example, von Adlercreutz had a fairly accurate read on the game:

Winner: Montezuma
Runner Up: Suleiman
First to Die: Sitting Bull
Number of Civs Surviving: 3
Victory Type: Domination
Victory Date: T290

To me it seems that wonder focus has been very bad for AIs as it leads to slower growth and neighbours getting more land. The wonderbuilder here is Huana so I think Monty will benefit from having him as a neighbour by being able to expand more. Monty will probably attack either Huana or SB early as warmongers hate peaceful AIs. For his sake I hope he chooses Huana. I expect Sitting Bull to be a target of a major dogpile, since he is neighbour with Genghis, Shaka and Monty. Even his über archers can't deal with that even though I think it will be bloody for the attackers.

Suleiman has imp which I think is verystrong for AIs. I believe that will somewhat hinder Shakas growth into his direction. I thnk Gilga and Sully will concentrate on killing each other on their corner of the world while the big battles are fought elsewhere.

Almost everything worked out this way, with the one exception being that Huayna won the middle of the map duel against Montezuma. Swap that, and this is otherwise very close to what took place.

poor Giggles ... 1 logical city plant from Sulei (to the northeast of him) and he's pretty much cut off from all the action and all the land

Yep, pretty much.

Winner: Huayna
Runner Up: Gilgamesh
First to Die: Sitting Duck Bull
Civs remaining: 4
Victory: Spaceship t303

Everyone will hate Sitting Bull, he's a certainty to die. Monty probably dooms himself by adopting a different religion to Huayna and getting killed post-SB. The Gilgamesh-Suleiman dynamic will be very interesting for how the game turns out. If Sulieman wins that he could get into a position to beat Huayna, whereas Gilgamesh would probably only have enough land for 2nd place.

Azza nailed everything except the outcome of the Gilgamesh vs. Suleiman situation. If Gilgamesh hadn't had stone and wasted so much time on wonders, then this very likely could have been the final result.

Montezuma won't get killed. He builds way too many units, and with that power rating the other AIs won't declare on him. He's never competitive enough in tech to win though.

I say Huayna wins easily. He's got an enormous economic advantage here. Only way he doesn't is if he goes off putzing around with Liberalism and Economics while Montezuma swarms him with maces. Huayna reaching Rifling is a lock.

As Parkin has pointed out, T-Hawk nailed half of the prediction, and the more important half of the two. The Montezuma part we'll just ignore.

Winner: Huayna Capac
Runner Up: Suleiman (score leader)
First to Die: Sitting Bull
Number of Civs Surviving: 5
Victory Type: Culture (!)
Victory Date: T339

Huayna would normally be squished, but his early game is saved by Sitting Bull taking all the heat -- and putting up a stubborn fight. Midgame is surprisingly calm due to Huayna sweeping the religions. They and all his wonders make him switch to culture, and he wins with plenty of turns to spare despite a midsized empire.

Becko nails half of an outstanding prediction here as well. Everything was good up until the part where the prediction suggested a peaceful midgame. With this collection of leaders, that was probably never going to happen.

OK, now for the results table:



Our table is starting to get rather large. There's been 48 entries total thus far, and 44 entries just for Game Four. It's been great to see so much interest in this thing. As far as individual scoring categories:

- We had 15 out of 44 entries pick Huayna as the winner, almost exactly one third. Shaka and Montezuma also had a lot of support. No one picked Sitting Bull or Gilgamesh to emerge on top.

- Suleiman as runner up was considerably less popular, only 8 correct picks. Huayna was the most popular selection here, almost everyone who didn't take him as the winner had him in second place.

- Sitting Bull was quite correctly picked as the first to die by almost half of the entries. Nice work sniffing that one out from so many of you.

- We've had four straight games now with 3 surviving civs, and that was the most popular pick. At some point, the streak has to be broken here... right?

- Domination was the most popular choice for victory type, with just under half of the predictions going that way. Spaceship was also a popular pick. (For the curious, Huayna still had a long way to go on his spaceship when this game ended. Quite a few more techs needed.) We also had two picks for Culture, and even one pick for Conquest victory! Nice to see some variety.

- Katon hit the victory date exactly, with a Turn 301 prediction. Well done, even though of course it's largely due to blind luck. (But picking upsets in the NCAA basketball tournament is no different, and everyone enjoys that part of their brackets. I say enjoy it.) We had six other predictions that were a single turn off in victory date, oh so very close!

Now with regards to our individual scores:

* Tyrmith remains in first place with another consistent prediction. 16 points, 12 points, and 13 points - workmanlike performances. Apparently he didn't even need to submit a "First to Die" pick for this game to increase the lead over the rest of the field. Nice work.

* Katon leapfrogged into second place with the highest-score single game that we've seen to date, a massive 26 points. The 15 points from the perfect victory prediction obviously helped, but the rest of the submission was also quite good. This high ranking comes despite not even having a submission for Game Two earlier. Impressive.

* Special shout out as well to Merovech for turning in probably the single best entry we've seen thus far (not one that happened to hit on the exact victory date bonus). Merovech correctly predicted the winner, runner up, first to die, number of surviving civs, and victory date to within 14 turns. That submission was only missing the correct victory type for a perfect entry. I wonder if we'll get one of them before the competition ends, it's virtually impossible but we do have a lot of entries to draw from.

* Kudos as well to Becko and TheHumanHydra for nailing all three of the leader picks, the winner + runner up + first to die. No one had done that yet, and then we had three such entries in one game. Quite a few others had two out of the three.

Then there are the random oddities from some of the entries...

# slowcheetah had the two leaders who moved on to the playoffs correct, but in the opposite order. From previous games, we've established that this special case is worth 5 points, and it was scored accordingly.

# David Corperial was convinced that this game would be a monster bloodbath, predicting only 2 leaders to survive, and for the game to end in Domination on Turn 212. Very early finish there! Even Shaka, Monty, and Temujin weren't quite able to pull that one off.

# Amelia predicted only one leader to survive, and had the latest victory date total on Turn 367. Becko was the opposite, predicting five leaders to survive, and the game to end via culture.

# Jowy had two leaders surviving and the game ending through a diplo victory, which is actually not possible due to the Beyond the Sword game mechanics. (The AIs will always vote for themselves in the UN, and you need someone else to vote for you to win via diplomacy.)

All in all, some great stuff. Thanks everyone. Feel free to discuss the results, I'll post the leader draw and leader profiles for Game Five tomorrow.

[Molach]

Did a bit better this run, at least. Was a bit worried early on but my guy fought it off and rose stronger and betterer than anyone else.

Seeing starting positions sure take a way a bit of the randomness, even if there is a lot of little things that can and will topple a budding Great empire.

I was a bit disappointed by my own predictions in the 3rd game - but luckily people that didn't participate in an earlier round is given 0 points, so I won't stand out that bad on the scoreboard...

Anyway I spent my playtime last evening watching those 3rd game goons having at each other again - in a totally different pangea - to see if I could change the wheels (or tank threads) of fortune.

Note: I probably did some things wrong in the setup, and I watched the actions using great spies and a lot of EP’s, so I think the score leaders were spending quite a lot of points on espionage instead of teching, leading to the slow date. Or perhaps because I failed to set them at deity? Who knows.

Winner: Qin
Runner-up: Louis (I was voting for him 3rd game)
First to die: João
Second to die: Boadica
Third to die: Alexander
Fourth to die: Isabella
Number of surviving Civilizations: 2
Victory type: Spaceship
Victory date: Turn 357

Awards:
* Catfighters: Boudica and Isabella. They had different religions at one point (Isabella founded Buddhism, then founded Judaism, then Judaism spread to miss B. (Insert Jean-Luc facepalm here) but luckily for Isabella miss B. decided to go with Buddhism a bit later after all. No matter, they declared and fought FOUR times, each time ending up worse off.

* Lucky number Slevin: Louis. He got bailed out of losing wars twice – he declared on João, who was a little stronger than he was, he lost his foreign legion(s) without doing anything - but Qin attacked from the other end which meant João was too busy to invade the weakened French lands. The second time Boudica decided to invade his wonderful but troop-empty lands, took a city and was poised to take another…and the rest of his empire…then Isabella declares on Boudica. (Buddhist Isabella declared on Buddhist Boudica rather than help tear apart Hindu Louis. Meow.) Boudica makes peace, even liberates the taken city back to Louis, ends up in a long war against Isabella. Louis then declares and destroy Boudica, leaving Isabella winner of the Catfight and few-and-twnety turns before Louis needed her lands for..something.

* Best effort: Alexander. He had the worst ice-ridden start, he never stood a chance. He probably understood this, and did not declare war on his more powerful neighbours. Instead he would declare war more distant civs, happily marching his army through snow and tundra and allied lands, to be slaughtered before the walls of some city his generals never even heard of. Qin took pity and delivered him from the frozen hell around turn 300.

* Best game: Qin. One war to gain the upper hand, One stack to take cities with, One tech path to leave his opponents in the dust and sail off to the stars with. He ended up with Joao’s lands, was friends with lucky Louie all game long (again lucky for him) and could have won 80 turns earlier by both quality and quantity of his armed forces.

So, there it is, Louis is pretty good after all, even if Qin was better. Isabella seems to have a knack for self-(opposite of preservation) that makes for fun times when she is around.

[Becko]

Winner: Huayna Capac
Runner Up: Suleiman (score leader)
First to Die: Sitting Bull
Number of Civs Surviving: 5
Victory Type: Culture (!)
Victory Date: T339

Huayna would normally be squished, but his early game is saved by Sitting Bull taking all the heat -- and putting up a stubborn fight. Midgame is surprisingly calm due to Huayna sweeping the religions. They and all his wonders make him switch to culture, and he wins with plenty of turns to spare despite a midsized empire.

Becko nails half of an outstanding prediction here as well. Everything was good up until the part where the prediction suggested a peaceful midgame. With this collection of leaders, that was probably never going to happen.

Yeah, about that... I've basically built my picks so far on trying to predict how the religious game (and consequently a large part of AI relations) would play out, with varying success. For this game, I remembered that Huayna and Montezuma tend to found religions. I couldn't recall, though, if they tend to found the same religion (i.e. if both tend to prefer Meditation over Polytheism or vice versa). If they do, then Huayna should win that race; and, since he and Monty were close on the map, they might end up sharing the religion and staying peaceful at least for a while.

So what did I do? I started up a test game, of course, with the same settings as you're using. And -- this time at least -- HC did win the race (Poly) and did spread Hinduism to Monty. First to Meditation, instead, was Sitting Bull. And then the fireworks began. And I just had to do one more turn. And one more...

...and well, you all know from my predictions how that particular game ended. Yes, this means that I sandboxed a guessing competition. Einstein was talking nonsense: procrastination, not compound interest, is the most powerful force in the universe.

[Merovech]

Woo! Trying for that perfec one next time.

Thanks for the shout out, Sullla, btw.

[Hesmyrr]

Wow, I'm surprised how well I am actually doing. If I knew about my standing, I probably wouldn't have went for unconventional picks (Mongols over Native Americans etc.) in order to make up for the perceived low performance.

Yes, this means that I sandboxed a guessing competition. Einstein was talking nonsense: procrastination, not compound interest, is the most powerful force in the universe.

Clearly your downfall was not running multiple trials

There's been 48 entries total thus far

47 entries, if you exclude my evil copycat!

[Katon]

Woo! Trying for that perfec one next time.

Thanks for the shout out, Sullla, btw.

A deserved shout out, too. I think predicting which leaders will build strong positions and which won't is probably the least luck-dependent bit of this, and you did that perfectly (as did TheHumanHydra, so props to him). Guessing how the strongest AI will cash in their winning position seems harder, though I did fairly well by assuming that the various lunatics would stir up enough ill will to leave the winner in a military mood, and getting the exact turn they'll line everything up is obviously dumb luck even if you manage to guess the victory type and rough tech pace.

On the other hand, the easiest thing so far has been guessing that there will be exactly three survivors. Anyone want to guess when the first game comes for which that isn't true?

[Dp101]

When are you thinking of rolling up the next set of AIs?

[Lord Parkin]

# Amelia predicted only one leader to survive, and had the latest victory date total on Turn 367. Becko was the opposite, predicting five leaders to survive, and the game to end via culture.

I don't think it's actually possible for the AI to win conquest with this particular game setup. I've tried a couple of test games myself, and it seems that no-one except Genghis ever razes cities above size 1 on Deity - and even then he only does it rarely. That means to get Conquest without triggering Domination first an AI would have to capture around 40% of the world's cities within the space of just a few turns (before resistance ends in the majority of them). And that's not factoring in the tiles gained by the aggressor's own culture immediately following any border city captures. Probably not really feasible with their programming, even on a relatively small Pangaea map.

[Sullla]

Let's get started with Game Five. We're back to a smaller field again for the fifth game in the series. This time, the leader draw produced a grudge match between a pair of American presidents, along with an interesting division between military despots and passive peaceniks.


Augustus Caesar

Augustus has never attracted that much notoriety as a leader in Civ4, perhaps because of his late inclusion in the second expansion and generally mediocre traits. Augustus brings Imperialistic and Industrious traits, a pair that are generally anti-synergistic with one another. Imperialistic wants a widely spread out empire, while Industrious does better with fewer and smaller cities. As one of the Roman leaders, Augustus is noteworthy for bringing the mighty Praetorian unique unit, along with the unimpressive Forum building. Augustus the AI has production and military flavors, and is best known for heavily emphasizing wonders (8/10 rating). He's mostly average across the board otherwise, below average in aggression rating (4.3 out of 10), but expect to see Augustus focus on building those wonders. He's somewhat surprisingly considered to be a very "Good" leader in terms of peace weight alignment. Augustus will probably hope this game remains peaceful and he can try to win via wonders and culture.


Hammurabi

Hammurabi was one of the few leaders included in the original Civilization game, and I always thought it was unusual that he didn't reappear in Civ4 until the Beyond the Sword expansion. Hammurabi has Aggressive and Organized traits, a decent second-tier pairing, particularly in the hands of the AI. Hammurabi is the only Babylonian leader, and makes use of the Bowman unique unit archer, and the Gardens unique unit colosseum. These are both pretty bad, and the only reason anyone ever picks Babylon in our events is for their awesome starting techs (Agriculture and The Wheel). Hammurabi AI is another one of the rare leaders with only one flavor: CULTURE. He'll heavily emphasize researching techs with a cultural bent to them. Hammurabi's AI otherwise looks like a carbon copy of Augustus, with the same heavy wonder focus (8/10) and a slightly higher aggression rating (5.5 out of 10). He also shares the identical peace weight as a "Good" leader. Basically, Augustus AI and Hammurabi AI will play very similar games and hope to win through peaceful or cultural means.


Kublai Khan

Kublai Khan is the first leader from the other side of the alignment coin, a much more aggressive ruler and someone predisposed to dislike the "Good" leaders. Kublai has Aggressive and Creative traits, a combination that often works well on cramped Pangaea maps. Like Genghis Khan Temujin, he has the Keshik unique unit and the Ger unique building, both of which are above average. Kublai has military and cultural flavors, and at first glance looks very similar to Augustus and Hammurabi. He even has a fondness for wonders (6/10) and an aggression rating that's only a little bit higher (6.4 out of 10). The difference lies in Kublai's peace weight, which ranks him as an "Evil" leader at the other end of the spectrum from Augustus and Hammurabi. This will cause major tensions between the leaders, and make it very difficult for them to get along. So despite having a similar AI personality overall, Kublai may find himself on the other end of a war declaration from Augustus and Hammurabi.


Lincoln

Abraham Lincoln is the first of two American leaders that we have making an apperance in this game. The odds suggested we were almost certain to get at least one such pairing; the only surprise is that it took five games to reach that point. Lincoln has Charismatic and Philosophical traits, a pairing that's not particularly impressive. He is also tied to the terrible American civilization, with its too-late-to-matter Navy SEAL and Mall unique items. Maybe this will be the rare game where we get to see them in action. Lincoln the AI has science and growth flavors for his research. You're never going to believe this, but his favorite civic is Emancipation. That one was probably a little too easy! Lincoln AI is extremely peaceful in this game, pacifistic even. His aggression rating stands at 0.8 out of 10, the second lowest in the game behind Gandhi. It's easy to get on Lincoln's good side and hard to get on his bad side. Lincoln AI has a sainty peace weight as another "Good" leader, which will inspire love from some of our competitors and hatred from others. I'd like to see Lincoln do well, but I have a bad feeling that he might run afoul of our next fellow here.


Ragnar

Ragnar is essentially the polar opposite of Lincoln. The Great Emancipator is the second most peaceful leader in the game; Ragnar is the second most warlike, behind Montezuma and only by a fingernail. Ragnar has Aggressive and Financial traits, which has often made him a popular pick in our Pitboss games. His Viking civ has the Berserker unique unit and the Trading Post unique building, both of which are extremely strong on water maps. They probably won't do too much on a Pangaea though. Anyway, Ragnar's traits and civilization are both above average, but what sets him apart is his AI personality. Ragnar AI is another leader in the Montezuma / Temujin / Shaka mold. He has only one flavor: MILITARY. He builds units at a ridiculous pace (10/10 rating). And as mentioned before, Ragnar is the second most likely leader to declare war in the game, with an aggression rating of 9.9 out of 10. That's a normalized score based off of Montezuma's AI, which shows that there's virtually no difference between Ragnar and Monty. Ragnar also sits at the bottom of the peace weight scale, ranked as low as it's possible to go as an "Evil" leader. He's going to be predisposed to hate these peaceful builders, and with such a gigantic focus on military, Ragnar will have a great opportunity to go conquering. I think Ragnar has drawn a fantastic setup, and he's the easy favorite to top this group.


Washington

George Washington is the other American leader in this game, as we've got a presidential grudge match going on. Washington has Charismatic and Expansive traits, which are a bit subpar and a far cry from Washington's original Financial / Organized pairing in the pre-expansion days. Like Lincoln, Washington is stuck with a weak batch of unique stuff by virtue of playing as the Americans. Washington AI has military and growth flavors, and comes off as more militaristic than the pacifist Lincoln. Washington is very average across the board in his various ratings, with the one exception that he likes to carry out espionage missions. His aggression rating is still low (4.3 out of 10), albeit much higher than Lincoln. Washington AI also has a high peace weight, putting him firmly in the camp of the "Good" leaders. He'll need to draw a good start and use the Expansive trait to get off and rolling quickly if he wants to do well in this game.

Overall, this one will almost certainly break down along the lines of the "peace weight" mechanic. We have four leaders who are all rated very high in having "Good" alignment in Augustus, Hammurabi, Lincoln, and Washington. They are thrown into the mix with two leaders who fall very far down on the same scale, with Kublai Khan and especially Ragnar grading out as "Evil" leaders. It will be very difficult for leaders from the two groups to get along with one another diplomatically. I believe that the game will go one of two ways. In one scenario, the peacenik leaders either eliminate the militaristic ones or out-tech them into irrelevance, and then the four of them happily compete for a Cultural or Spaceship win in one giant lovefest. In the other scenario, Kublai and Ragnar mercilessly crush the goody-goodys and are the only ones left standing at the end of the game, with one of the two winning by Domination. I doubt we'll see too much in between those outcomes. Should be fun to see which one pops up.

Our map, spoiled for size:

Here's our current bracket:



Same categories as before for the prediction contest:

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

Predictions will close in two days on Monday evening (my time). Good luck to everyone.

[Jowy]

Winner: Kublai Khan
Runner Up: Washington
First to Die: Hammurabi
Number of Civs Surviving: 3
Victory Type and Victory Date: Space Ship T351