Civ4 AI Survivor: Season Four

[haphazard1]

Heh. Yeah, don't see those off-shore aluminum sources every game. I am guessing seafood was being added and there was a mis-click. The map for the championship is mirrored, right? So it should be possible to look at the matching areas and see what is in the corresponding spot.

[RefSteel]

I am guessing seafood was being added and there was a mis-click. The map for the championship is mirrored, right? So it should be possible to look at the matching areas and see what is in the corresponding spot.

Indeed. It looks like the spot where Kublai has Underwater Aluminum is one north of a spot that corresponds to an ocean tile where (only!) Willem and Chuck have Fish. I wonder if there are any other accidental breaks in the mirror....

[Sullla]

Here's the link to Friday's First to Die Fiesta game on YouTube. Not going to write this one up because it was a side project not part of the main competition, but it was still fun to watch.

Community predictions have been added to the preview page for the Championship! Here's a quick summary:






Unlike most of our previous games, the Championship field was evenly divided between different options in the picking contest. While this was assuredly due to the mirrored terrain on the custom map, it was still exciting to see such a wide open competition going into the last match. Willem, Kublai Khan, and Stalin all had roughly equivalent support to win the game, with Charlemagne dragging behind in last place. The runner up contest generally held the same pattern with slightly more support for Gilgamesh. First to Die had a clearer favorite in the form of Darius, the choice of about 40% of the contest entrants. These people were betting on the high peace weight of the Persian leader making him unpopular with his nearby neighbors. Finally, we had an even division between Spaceship and Domination as the victory conditions for the Championship, with a slight preference for the former. No one had chosen a Diplomatic victory at the time of writing even though the First to Die Fiesta had ended in that fashion earlier this week. Once again, the picking contest seemed to suggest that the game and the overall crown were open for the taking.

Finally, here are some of the best/craziest written predictions about what would take place during the game. There were many other excellent entries but I had to pick and choose my favorites to keep this from running on too long. Thanks again for the submissions!

Warclam: I think this is pretty clearly a game where Willem's financial nature and neutral peace weight will let him get ahead in tech and stay there, while Darius does stupid stuff and then dies. So if Willem goes to space (or culture), who's his running mate? Not Burger King, who's going to be fending off attacks from the warmongers. Kublai's sufficiently non-aggressive that he might be able to get strong while Joe and Gil slam their heads against BK's protective walls. On the other hand, either of them could take enough territory off Darius and Burger King to get out in front, leaving Kublai in third. It's a risk, but I have to believe Kublai has the heart of a champion. Third time to the championship, that has to mean something. So I say Clam King Willem goes to space, while Kublai becomes the War King and consolidates a strong medium-sized empire taken from his share of the Burger King and the Dumb King. Gil and Joe get to duke it out for the bronze.

Takuan: Kublia conquers Kublai, Stalin conquers Darius. Both of them snowball the game, but neither come out on top of the other - or they're friends and the game continues to space. In any case, i'm rooting for Kublai, since he deserves to win a championship game, given his record in past seasons especially compared to the other AIs in this game. Go Kublai !

LinkMarioSamus: First of all, Kublai Khan has the lowest unit emphasis rating of these leaders, and he has a fairly high wonder preference. Oh, and his nearest neighbor is Charlemagne, who has a diametrically opposed peace weight at 6 vs. Khan's 1, and who just beasted all over his playoff field. Oh, and Khan's the only leader here who hasn't outright won a game this season. Oh, and he was FTD in the previous two championship games. Oh, and he's in roughly the same position he was in those games. Uh oh. For the winner I'm picking Gilgamesh, since his AI ratings have the most potential of this bunch. He'll build a lot of wonders and he's also very willing to wage war on his rivals, without being too aggressive about it. For 2nd place I debated between Charlemagne and Willem, ultimately going with the former because he has tied Gilgamesh for the most points earned this season thus far at 11, he should be able to snowball off a conquest of Khan (also he's the only leader here who doesn't have a high wonder rating, so he won't waste time with them as much and will go straight for units like a good Templar Knight), and Willem might just sit in his corner and build until he gets double-teamed by Gilgamesh and Charlemagne. For a reflection on Gilgamesh, if you put a lot of stock into past results his potential zero to hero performance this season may look like a surprise, but it probably shouldn't. He had horrendous draws in Seasons 2 & 3, getting thrown into games with lots of high-peace weight leaders. At least he got his revenge on Mansa Musa and Gandhi this season! Also, we had only done three seasons, so it isn't all that illogical that a strong AI leader was hiding in plain sight due to consistent bad luck.

Commodore: There are two ways this game shakes out; either Willem or Darius reaches rifles, or neither makes it. I don't see Charlamagne giving up much early on, but his nasty neighborhood won't let him win, either. Kublai and Stalin might get gummed up in his Protective works, but I think Darius' low peace weight will tempt them more profitably north, abetted by sharing the Holy Roman religion. I can't wait to see how Mr. Disappointment of Persia messes up this call.

Eauxps I. Fourgott: Kublai Khan is a complete fraud, who gets by on lucky breaks time and time again to continue advancing and has deserved none of his championship appearances. I expect him to properly be first to die here yet again. In fact, almost everybody here has played poorly and/or had lucky breaks to get this far: Gilgamesh, Kublai, and (to a lesser extent) Charlemagne were lucky to advance past the opening round. Stalin was lucky to survive the playoff round and get chosen by Gilgamesh to advance over Shaka. Darius prospered from long years of peace and threw away a certain victory in the playoff round. Out of these six leaders, Willem has by far the most impressive resume this season, with a second place finish with the top score in the opening round, and a win while beating back a 2v1 war in the playoff round. He's needed the least lucky breaks of any of the leaders, and to make his situation even better he has a nice middling peace weight that will prevent any of the other leaders from instinctively hating him too much. This is Willem's game to lose. To round out my predictions, I'm picking Charlemagne to have yet another impressive game that still falls short of the win for second place, and a cultural victory as a realistic Willem choice that would continue the pattern for victory conditions in championship games.

Max: Stalin and Kublai team up to attack Darius and Charlie and eat them up. Gilgamesh declares a random war that goes nowhere. William quietly teches but doesn't get to Spaceship before Stalin gets Domination.

Zalson: Gilgamesh continues to demonstrate his excellent "settling" in a straight line technique, right into Darius, crushing his hopes just as Gilgamesh did in my deity play through. Stalin joins in, both of them divide Darius right up and then go on to stomp out Charlemagne and Willem. Or none of this happens, Willem expands his way to power and then fends of a 2v1. Or none of that happens and Stalin gets AP-cheesed to death by Charlemagne. Commodore, this is a beautiful map, BTW.

TMIT: There are a few properties that will make this game deviate from normal maps. First, there are a lot more plains around starting positions than typical. This will cut into cottage setups. Second, distances between AIs are longer than usual, so we're less likely to see early aggression with delayed islandtarget checks (and less likely to see it be effective, should someone go for it anyway). Third, wonders are disproportionately strong on maps like this. As a result, I somewhat favor Stalin here, with a tough choice after him between Gilgamesh and Kublai. I give Kublai an edge due to his location + possibility to conquer Willem, who I have pegged as first to die. For obvious reasons, I don't trust game-thrower Darius. Despite his fantastic traits, him even being in position for a victory was something of a fluke compared to previous games, and he even managed to choke that! Burger King isn't impressive but could easily ruin the game for someone else by going into some religious aggression. His behavior in the early turns is likely to be a wildcard yet again in the finals, though it is significantly less likely to favor him this time...

One More Time!: Willem falls on his face, gets eaten by Gilgamesh. Charlemagne really does get 2v1'd this time. Darius can't survive to the end, Russia wilier than Genghis or Gilgamesh. Cosmonauts end it when the low peace weights stop the offensive.

RefSteel: Kublai Khan might be the most underrated leader in this competition. There was even some discussion in the thread at Realms Beyond about how he's an awful AI because he doesn't build enough units - when in fact he builds just the right number for a low-peace-weight leader whose enemies are most likely to have low unit counts themselves. Once he's planning war, of course his unit count will rise, and his aggression rating is high enough that he's likely to be the one plotting war when his relations with someone are poor. On this beautiful mirrored map, his natural ally is Gilgamesh, who is also very close to him in peace weight, and I'm guessing they're going to roll the higher-peace-weight leaders one by one, with their split of the Dutch - the civ to whose land Russia has the least access - ultimately helping them first and second place. Of course, religious fallout and other random factors could well trample this prediction into the dust ... which will only help Mr. Xanadu fly under the radar again next season!

Brian Shanahan: Interesting to note that three of this year's finalists are debutants, none of whom have done that well previously. This will make for an unusual game, I think. The other interesting note I see is that I think the south has the stronger AIs, with Stalin and Kublai (the two leading candidates) situated as neighbours. Hence why I'm picking Dayrush as my winner. As the stronger of the northern candidates I think he can get enough of a lead in the north to translate into something of an underdog victory before the winner of the Russian-Mongolian wars can get around to attempting to taking him down. Willem is first to die because he's just pathetic at expanding. Of course what I really want to happen is Mansa Musa to return from space, kick everybody's behinds and romp home to victory on turn 150.

lightms: I don't know what's going to happen, but Willem appears to be the new Stalin, even more than Stalin

JackDRB: I must admit, this linup very much surprises me (in a good way), considering on the power rankings only two of these leaders are ranked in the top ten! I actually think Willem can win this game, given his relatively low peaceweight, he is likely to be able to tech in peace for a while, at least until Darius and Burger King is gone, and a field of highly aggressive leaders aren't nearly as scary once you zoom past them in technology. For second, Stalin looks in good position to pick up spoils from both leaders I think will definitely die this game, and is highly aggressive himself, so he will get a good amount of territory for himself, even if I don't think he would be able to translate that into a technological lead. First to die I see as being Darius because, well, he's going to be the big fat red target that all the other agressive AIs gravitate towards to.

Y3llowSheep: Who knows? This year has been the year of darkhorses, so anything goes.

Championship Game Picking Contest Entry Form

[Fluffball]

A lot of well deserved Kublai hate in those predictions.

[Reverent]

Another quick reference post for the upcoming game.

• Willem, Kublai Khan, Gilgamesh, and Stalin can all plot war at pleased. Charlemagne and Darius cannot.
  
• Willem and Darius share a favorite civic (Free Religion)
  
• Willem's strategy: Gold (5) & Science (2)
  
• Kublai Khan's strategy: Military (5) & Culture (2)
  
• Gilgamesh's strategy: Military (5) & Culture (2)
  
• Stalin's strategy: Military (5) & Production (2)
  
• Charlemagne's strategy: Military (5) & Religion (2)
  
• Darius's strategy: Gold (5) & Growth (2)
  
• The turn order appears to be Willem - Kublai Khan - Gilgamesh - Stalin - Charlemagne - Darius
  

[Reverent]

Also, since we were talking about this...

• Gilgamesh, Stalin, Charlemagne, and Darius all share a build unit probability of 30. (Max 40)
  
• Willem and Kublai Khan share a build unit probability of 25. (Max 40)
  

In this case, I think the AI's other flavors have a bigger impact on their unit totals than the actual unit build probabilities. Leaders like Willem (25), Darius (30), and Victoria (25) have a lot of problems building units, especially late game. However, their unit build probabilities aren't that different from other leaders. Stalin's certainly no military slouch, and Gilgamesh builds some decent armies himself, yet both of them them share an identical build unit probability with Darius!

I think team dynamics must be quite different, however. I've never noticed Kublai Khan having any trouble building up sizable stacks in single player. He's also built up some truly terrifying stacks in past rounds of AI survivor. (Just look at that 100+ unit stack he sent against Willem in their playoff game!) However, since he's unable to plot war when allied with the player in a team game, maybe he has a lot of trouble triggering that military strategy. I suspect that the build unit probabilities probably play a much bigger role in team games. In single player games, I think the strategies and wonder build probabilities make a bigger impact.

Tied to this, I suspect the overly wonder-prone AIs actually benefit from losing the bonus deity techs this time. For Darius, having to research all those early worker techs keep him from the catnip wonders. If he researches mysticism on turn 40 instead of turn 15, that gives him 25 extra turns of useful production that goes into the critical early expansion phase. He's still not going to be one of the rapid expanders, but it lessens his weakness in expansion. It helps prevent him from settling city #3 while Gilgamesh works on city #6.

For a similar reason, I actually like Willem's odds less this upcoming round than if his capital had been landlocked! Since he starts with fishing, I'm a bit worried that he'll go for an early sailing tech, then tie his capital up on a lighthouse and the Moai statues in the early game. Worse, since the capital has a mere three ocean tiles in its BFC, he won't even benefit that much from either build. Admittedly, the land's so bountiful that I think he'll still prioritize worker techs first, but there's still a decent chance he goes for an early sailing research and then cripples his expansion.

Later game, I think that strategies like gold, culture, and science end up affecting the AI's unit counts more than their unit build probabilities. Leaders like Willem and Darius love to build research and wealth in their core cities. However, I really haven't looked into this enough. Specifically, how often do they reconsider these research and wealth builds? If each research build ties up a city for an average of 20 turns, an AI with the gold or science flavors probably produces far fewer units than an otherwise identical AI with, say, the production flavor instead.

[Fluffball]

I think team dynamics must be quite different, however. I've never noticed Kublai Khan having any trouble building up sizable stacks in single player. He's also built up some truly terrifying stacks in past rounds of AI survivor. (Just look at that 100+ unit stack he sent against Willem in their playoff game!) However, since he's unable to plot war when allied with the player in a team game, maybe he has a lot of trouble triggering that military strategy. I suspect that the build unit probabilities probably play a much bigger role in team games.

The inability to plot war while allied with me noticeably screws up some AIs. 
The weird thing with KK in the game I was referencing earlier was that we were in a non-stop war with someone or other for hundreds of turns in a row because KKs power rating was so low the enemy AIs were basically forced to declare war on us. Mid-wars, he was building intelligence agencies in frontline cities that were defended by less than 4 units.
Contrast that against my experiences with Napoleon, who despite not being able to plot war will just sort of park 80 units on an enemy border and sit their waiting with puppy dog eyes for me to declare war. I think I've said out loud "Oh! Looks like we should attack ___ who I was not even remotely considering attacking."

All that said, team games and AI survivor are wildly different, so I try not to read TOO much into it, especially since KKs peace weight and the benefits of it are just as much a part of his persona as his build preferences. It's hard to overlook the insanely poor performances from Pacal and KK I've noticed in team games though.

Edit: Oh, another fun thing is that in MANY games I've seen Ragnar either eliminated or losing multiple cities to non-raging barbs. In a game I was playing last night Ragnar had his second city razed SIX times by non-raging barbs.

[haphazard1]

Thanks for the info and thoughts, Reverent! Very useful. I also wonder about the AI's tendency to build research, rather than units or infrastructure. There are times when doing so is useful, of course, but the AIs seem to do it a LOT.

Fluffball, I agree that Ragnar for some reason seems to have way more problems with barbarians than he should. He is aggressive and gets free combat 1 promotion on many early game units, plus cheap barracks which the AI tends to build early. And there is also the AI's bonus against barbs at high difficulty levels. Just how does he manage to get himself in barb trouble so often?

[Fluffball]

Fluffball, I agree that Ragnar for some reason seems to have way more problems with barbarians than he should. He is aggressive and gets free combat 1 promotion on many early game units, plus cheap barracks which the AI tends to build early. And there is also the AI's bonus against barbs at high difficulty levels. Just how does he manage to get himself in barb trouble so often?

I suspect there is a value for an AIs willingness to lose units. A long time ago I tried out an Always War game, and I did so on Settler difficulty to get a feel for it. I noticed that putting a single unit on defensive terrain made enemy stacks take massive detours to get around it, where as the regular difficulty AIs would have just plowed through it. In AI survivor it's pretty common to see raiding cavs or tanks attack enemy cities.

My guess is that Ragnar has a huge willingness to lose units, which leads him to suicide against barbs, or loses him units very early in the game when they aren't easily replaced. IIRC Sullla was noticing him attacking out of his cities in the wildcard game, which was certain death versus certain victory to just defend the city.

[haphazard1]

Interesting point. I think you are correct about Ragnar attacking out of his cities and thus losing his defenders (or should have been defenders) at poor odds. I may have to go through the leader XML file and see if there is anything in there that might define this tendency.

I also wonder about the Aggressive AI flag being set for AI Survivor and if i connects to this behavior. Although that would affect all the AIs and not just Ragnar, so probably not. I know it does change whether an AI is willing to declare war on someone more powerful than they are, which helps explain some of the utterly suicidal war declarations we have seen in AI Survivor.