March Madness: Civ4 AI Edition

[Lord Parkin]

If you can be bothered, I'd be interested to see the rankings for "average score" (total score / # game predictions). Seems it would be a better indication of accuracy given that we haven't all bet on the same number of games.

[Ceiliazul]

If you can be bothered, I'd be interested to see the rankings for "average score" (total score / # game predictions). Seems it would be a better indication of accuracy given that we haven't all bet on the same number of games.

I know we're emulating college sports here, but I don't think splitting the standings into two different scoring methods would help. Split championship? This is March, not January!

I did scan the board tho to see who's doing well despite missing a game. Definitely some stand outs there!

[Ituralde]

A big thank you to you as well, Sullla! Those writeups probably don't write themselves.

Already looking forward to the next round of candidates.

[Hashoosh]

That puts the people who have smaller sample sizes to choose from at an advantage than people who predict every game, not that I care either way.

Just objectively speaking, if someone has 30 points from 2 games, they'd have 15 points per, while someone who has 60 points from 4 would be tied with them, the fact that they have double the sample size is an indicator of better performance.

[pindicator]

Except some of us have actually rolled zeroes for certain games...

[Sullla]

I'm not a huge fan of the per-game suggestion of tracking scores. For one thing, it would require even more work on my part, and keeping the current spreadsheet is a fair amount of effort already, not to mention playing out the games, and writing the reports, and doing the leader previews. Beyond that, Hashoosh correctly points out that per-game averages tend to work against participants who have a large number of entries. It's easier to get a high per-game average with 2 entries than it is with 6 entries. We could see a situation where people intentionally refuse to submit predictions, in order to preserve a high per-game average. Keeping a running tally does the opposite: participants want to get in more submissions so that they can score more points, and the whole system is more fun when we have more people to compare between. All in all, I think it works better this way.

Now the leader draw for Game Six:


Darius

This game is heavy on leaders with the Financial trait, and Darius is the first of the bunch. Darius has Financial and Organized traits, possibly the best pure economic traits in the game. This pairing is one of the most popular for our Multiplayer games, and is reigned in only by having no advantages in the early game. As a leader of the Persians, Darius brings the Immortal unique unit and the Apothecary unique building, neither one considered to be especially good. Then again, the immortal might actually be useful here given the number of archers that the AI likes to train. Darius the AI has gold and growth flavors for his research. He has a heavy wonder focus for a non-Industrious leader (8/10), along with an above average preference given to units. Darius AI is otherwise fairly average across the board, with an aggression rating in the middle of the scale (5.2 out of 10). He's considered to be a "Good" leader in peace weight. Darius is pretty easy to keep on good terms as a neighbor, and generally wants to be left alone to tech in peace. If he would manage to acquire a lot of land somehow, his traits make him a serious danger to run away with the game.


De Gaulle

De Gaulle has a number of oddities in the way that his AI is programmed. He brings Industrious and Charismatic traits, making him unpopular for everyone other than Parkin in our Multiplayer events. De Gaulle's French civilization, on the other hand, is an outstanding choice, easily one of the best in the game with the Musketeer and Salon unique items. De Gaulle the AI has production and growth flavors. You'd think that he would have a wonder focus with his Industrious trait, but no, De Gaulle AI barely looks to build them at all. Only 2/10 rating for wonder construction. Instead, De Gaulle AI will make lots of tribute demands (10/10) and civic demands (8/10) on other leaders. While his aggression rating is only average at 5.7 out of 10, De Gaulle gets classified as an "Evil" leader via peace weight. He's actually as low as it's possible to go on the scale, sitting in the same position as leaders like Ragnar and Shaka. This suggests that someone at Firaxis was having some fun at De Gaulle's expense, especially when you consider how De Gaulle gets a rating of 0/10 in the "Resists Capitulating" category. Anyway, De Gaulle AI will be seriously unpopular with many of these other leaders, and that should produce some weird stuff before we're done.


Elizabeth

Elizabeth is another Financial leader, one who leans towards peaceful tendencies. She has Financial and Philosophical traits, a pairing that was more popular in the past for our group, only to fall into some disuse of late. Financial and Philosophical are traits that are seen as having some anti-synergy, between working cottages and working specialists. This is the first time that we've seen any of the three English leaders; the civilization has the Redcoat for a unique unit and the Stock Exchange for a unique building. Both are well above average, and the one weakness of the English civ is the starting techs, which won't be a factor here. Elizabeth AI has gold and culture flavors, befitting her peaceful nature. She has a very low unit emphasis (2/10), an extremely high peace weight as a "Good" leader, and one of the lowest aggression ratings in the game (1.9 out of 10). That's 49th out of the 52 leaders in likeliness to declare war. Elizabeth AI is another leader who's easy to befriend as a neighbor, and she rarely even makes demands. Even more so than Darius, she wants to be left alone in peace.


Frederick

Frederick is another leader similar to Darius and Elizabeth, only without the advantage of being Financial. His trait pairing of Philosophical and Organized makes him another leader with strong economic abilities. Frederick's German civilization has a lategame focus, with the Panzer unique unit and the Assembly Plant unique building, which rarely have much of an impact. In all of the Civilization games, it's better to have unique things that come earlier on the tech tree. Frederick AI has only one flavor: PRODUCTION. That's kind of an oddity, and it will be interesting to see how that affects his tech choices. Frederick AI is very average across the board, with middling scores in most everything. His aggression rating is below average at 4/10, and he can sometimes get himself into trouble with his low unit emphasis (only 2/10). Like Darius and Elizabeth, he has a high peace weight as another "Good" leader. In short, Frederick AI looks very similar to Darius AI and Elizabeth AI, only with slightly weaker traits.


Julius Caesar

Now we finally get to a little bit more muscle. Julius Caesar has Imperialistic and Organized traits, a fairly weak pairing which ensures that he never appears in our Pitboss games. His Roman civilization features the Praetorian unique unit (and the "meh" Forum). Caesar's other Roman counterpart Augustus used praetorians to excellent effect in the previous game, we'll see if he can do the same here. Caesar the AI has military and production flavors, the only leader in this particular game with a military flavor. Caesar AI doesn't have an especially high unit emphasis (only 6/10), but he's well above average in aggression rating at 7.6 out of 10. This is the highest by a good margin for this particular match. Caesar also demands tribute frequently (8/10), and he has a lower peace weight number. It's still well above De Gaulle's peace weight of zero, but also a far cry from saintly leaders like Elizabeth. Overall, expect Caesar to be one of the leaders driving the action forward in this game. If he can get an early conquest with his praetorians, he may be able to snowball the game from there.


Pacal

Pacal's trait pairing of Financial and Expansive is widely viewed as the strongest for any leader in Civ4. Pacal is therefore a universal ban for our Multiplayer events, at least the ones played without any mods. He is the only leader for the Mayans, who feature the Holkan unique unit and the Ball Court unique building. This is a rare case where the unique building is the better of the pair, as it adds a nice boost to happiness. Pacal the AI has culture and growth flavors. He places a high priority on religion, with a huge penalty for leaders of rival faiths. Pacal will likely found his own religion in this game, with the Mayans being the only one of these civilizations to start with Mysticism tech. Pacal the AI also likes to build wonders (8/10 rating), and otherwise has mostly average numbers. Pacal has the strange combination of a low aggression rating (2.8 out of 10) along with an "Evil" rating in peace weight. Pacal AI will be predisposed to like De Gaulle, and he won't fit in with the other economic leaders in this game. If Pacal can get an early conquest, his traits will probably make him unstoppable.


Pericles

Pericles is yet another generally peaceful leader with an economic focus. He has perhaps the best trait pairing in the game for pure culture, with Creative and Philosophical traits. That means cheap libraries, universities, theatres, and colosseums. As the other Greek leader paired with Alexander, Pericles has acccess to the Phalanx unique unit and Odeon unique building. There's some nice synergy between Pericles' Creative trait and that odeon, perhaps helpful in pursuing a cultural gameplan. As an AI, Pericles has production and science flavors. He's another AI leader who loves him some wonders, with an 8/10 rating in that category. The other numbers are mostly average across the board, right in the middle. Pericles AI desires peace with his neighbors, having a low aggression rating of 3.3 out of 10. His peace weight sits in the middle of the scale, halfway between the "Good" and "Evil" leaders. Pericles won't be drawn naturally to either side of the alignment scale. Pericles the AI likes to go for Cultural victories, and he'll have a decent chance to pull one off here in a crowd of mostly peaceful competitors.

This game looks unpredictable to me. There are a lot of economic leaders, and several of them (Elizabeth, Frederick, Pericles) will often attempt a Cultural victory if they get left unmolested. With three different Financial leaders, along with a lot of strong trait combinations, we might have the chance to see an accelerated tech pace. However, we also have some major divisions along the spectrum of peace weight, and a couple of aggressive leaders who might like to stir up some mischief. Caesar's praetorians in particular could be a nightmare for one of the more docile AI leaders. Like the previous games, it's hard to know what the AI is going to do. I'm about to hit the half dozen mark on these games, and I still have no clue what's going to happen.

Our map, spoilered for size:

This one was too interesting to pass up. Would that central starting position help Pacal or hurt him? I had no idea. I expect a good number of "Pacal to win" and a lot of "Pacal to be first eliminated" picks. Either could be very viable.

Now our current bracket after five games:



It's starting to shape up now, we're getting a better look at the playoff picture. As before, everyone has two days to get in their predictions for Game Six. The deadline will be Friday evening (my time) in a little over 48 hours. I'm looking forward to seeing what everyone makes of this game. It's pretty wide open.

[Bobchillingworth]

Sort of shooting in the dark here:


Winner: Pacal

Runner-up: Elizabeth

First to die: Frederick

Surviving Civs: 4 (I'm a maverick!)

Victory type: Space

Victory date: T319


Rationale: Pacal is able to expand to encompass much of the lush center of the map, while his uncoordinated and generally peaceable neighbors fail to keep him from running away from the pack. Julius overruns Frederick, but eventually falls behind in tech and is eliminated by a coalition of the "good" civs. DuGalle gets picked off in the late game as well, but otherwise the world ends up being fairly peaceful, as the remaining AI refuse to attack each other and are fairly close to tech parity.


Edit- knocked some turns off the finish date, which I think I pegged too high.

[Sian]

This one could go all ways so lets take an oddball guess

Winning Leader: Elizabeth
Second Place: Pacal (Highest score)
First to Die: Frederick
Surviving Civs: 4 (Elizabeth, Pacal, Julius, De Gaulle)
Victory Type/Date: Cultural t315

[superjm]

Winner: Pacal (Traits are too good, too many peaceful leaders for his central position to stop him)
Runner-up: Julius Caesar (Not betting against Rome, not after last game)
First to die: De Gaulle (Overrun by Rome early)
Surviving civs: 4
Victory Type: Space
Victory Date: T327

[Jowy]

Winner: Pericles
Runner Up: Elizabeth
First to Die: Frederick
Surviving Civs: 3
Victory Type/Date: T302 Space Ship