(August 3rd, 2020, 17:58)Kuro Wrote: Here is one of the resources I used when looking for how neighboring AI affects likelihood of war, if you or anyone else wants to take a look (since for all I know I messed up XD): https://forums.civfanatics.com/threads/i...st-7190899
I'm sure there's other or better resources but hey.
That's... pretty convoluted stuff.
But yeah, the takeaway here is that it looks like amount of shared border is a bigger factor in war plotting than we'd originally thought.
(SPOILER FOR PLAYOFF 1 SINCE I CAN'T REMEMBER HOW TO TAG IT)
Just eyeballed the map at the time Louis declared war and it does still look like he would've been more likely to attack Sal, but even when adjusting for relations, given this new info, an attack of Washington looks no more likely than one on Pacal. So what actually happened was still not the most likely outcome, but it isn't quite as out-of-nowhere as it seemed at the time.
That's definitely an important insight for future games though. Thanks for sharing this info with us!
The more detailed information about how the AI chooses war declarations is fascinating stuff. I'm amazed that I'm still learning more about how Civ4's gameplay works after all of this time. Anyway, I wanted to mention that the written report for Playoff Game One is now finished and posted to the website. This is another contribution from Eauxps I. Fourgott Wrote so please give him the credit for putting it together: http://www.sullla.com/Civ4/civ4survivor5-10.html
Some other stuff worth noting, from me opening up the zip file in that post and looking around some.
Like the post says, every leader has a different Dogpile War chance (iDogpileWarRand) which affects the odds of wanting to join a war just due to a dogpile. This stat might actually shed some light on some AI personalities we've seen. The way it works, as I understand it, is that it is 100/iDogpileWarRand. De Gaulle, interestingly, has the highest likelihood of dogpiling! With only a 20 score, his odds of simply wanting to dogpile at a base chance is a massive 5% per turn! This means that De Gaulle is more than twice as likely to dogpile someone if the AI is checking for Dogpile War than Montezuma is to just declare war normally (5% vs. 2%). This might also explain why the AI tends to be good at dogpiling or at "sniffing out" weak competition: The Dogpile War chance is checked even if normal war fails and is at a higher rate, combined with the AI performing a defense power check.
Similarly, it has often been said that Catherine plays "opportunistically": One reason for that might be the Dogpile War chance! After De Gaulle's lone 20, she is in a large tie for 2nd with 25, giving her a 4% chance per turn base to want to dogpile someone. On top of that, Catherine's -2 diplo penalty makes her more likely to dislike people and have a larger multiplier for relations / avoid Friendly / then proceed to be willing to dogpile when they are attacked. It also just in general means she is more likely to take on a two front war rather than a random 1v1.
Interestingly, there are multiple "peaceful" AI who have a high iDogpileWarRand! In particular, Roosevelt and Elizabeth have a 25 Dogpile War value, giving them the same 4% chance as Catherine. I can't remember all of Elizabeth's wars right now, but it might explain some of her more random war declarations in some games. The lowest value is Lincoln, not Gandhi, whose lone 150 value gives him only a 0.67% base chance per turn (which, it should be noted, is still 3x Gandhi's base war chance of 0.2%). Willem has a unique 80 value and Charlemagne has a unique 75 value, as well. The only values are 20, 25, 50, 75, 80, 100 and 150: 20 is only with De Gaulle, 75 is only Charlemagne, 80 is only Willem and 150 is only Lincoln.
Also, the AI does calculate dogpile power into account, by adding the sum of the power of all people at war with the potential victim to their own power. This means even a weak person at war with a dogpile victim can matter a lot to both the total war and dogpile war chances: Total War does 3/2*OP(OwnPower)*MaxNearPR or MaxDistPR depending on how close or far the enemy is. Something also worth noting is that MaxNearPR is essentially always higher than MaxDistPR! For the relevancy of this game, Louis' MaxNearPR is 110 while MaxDistPR is 70, so the effect of adding a dogpile attacker's power to their own is further multiplied by the higher MaxNearPR compared to other options. (As a note, this also helps explain Washington joining the war: Louis AND Alexander's power must have ballooned that OP score quite a bit when it comes to rolling for if they want to declare war!)
In fact, something I noticed is some of the inconsistent or mid tier leaders have a high dogpile war rating compared to their aggression. For example, Sury is "only" a 7/10 aggression rating, but he has the 25 Dogpile War and so has the near-highest dogpile war chance. Some other examples include Brennus (7/10 aggression, 25 Dogpile War), the aforementioned Catherine, Tokugawa (7.3 Aggression, 25 Dogpile War), the aforementioned Elizabeth and Qin Shi Huang (3.9 Aggression, 25 Dogpile War). Interestingly, Huanya and Justinian both have the 100 Dogpile War rating that means they only have a 1% chance, sae with Cyrus.
For fun, given he has the highest chance, I went back to the game De Gaulle won. In that game, he only declared war on Frederick after Napoleon did (dogpile), only joined the war on Hatshepsut after Napoleon + Shaka declared war (dogpile) and only declared war on Boudica after she declared on Napoleon (dogpile). He did not declare a single war on his own the entire game! This also might explain why he seemed to enter wars in that game only to pick up spoils or at opportune times: He had a lot higher chance to declare war after people started, at 5% base, and as the power of the dogpile target goes down the chance they become a dogpile target goes up due to the reduction in MaxNearPR or MaxDistPR. This might mean an ideal map for De Gaulle to win on is one where there are likely to be many wars against other targets that his AI will then join in on due to his high chance of dogpiling. This actually almost perfectly describes the game he won, too! He was surrounded by friends unlikely to attack him (Boudica is inclined to be friends with him and won't declare at Pleased, but is still a warmonger) while bordering a peaceful civ (Frederick) who is surrounded by aggressive civs who can start a war (Napoleon in particular is VERY close and aggressive!).
A few more random notes: I originally opened the zip file to see what Louis' was, which is 50. This means his base chance is 2% at dogpiling, the same liklihood of Montezuma declaring war normally. Mansa Musa actually has a 50 dogpile chance as well, making him a teching leader more likely to dogpile than most: This may explain why Mansa seems to go to war more than you'd think for the peaceful guy ("because he will plot war at "Pleased" and isn't afraid to lay the smack down despite his 1.6/10 aggression rating. " The fact he will declare at Pleased and has a very high dogpile rating is likely why he lays the smackdown!). Asoka also has a 50 dogpile rating, making him another very low aggression leader more willing to war with dogpiles.
Charlemagne has a large difference between his MaxNearPR and MaxDistPR, with 100 NearPR but only 30 DistPR. This means Charlemagne is significantly more likely to go after smart, nearby targets and may explain stuff like picking off Alexander a little. Napoleon, on the other hand, apparently did not learn from his war with Russia, because his MaxNearPR and MaxDistPR are almost exactly the same, 120/100. This means Napoleon is a lot more likely to make bizarre, long distance wars which might explain some issues he has had.
Some other quick examples (using NearPR/DistPR): Tokugawa has a large difference (100/30), Qin has a very large difference (120/40), Roosevelt has almost no difference (100/80), Alexander has almost no difference (90/70), Gilgamesh has a ridiculously large difference (120/30), Joao II has almost no difference (130/110)...and, most extremely, Sitting Bull has a RIDICULOUSLY huge gap between NearPR and DistPR! Because he weighs NearPR at 130 (which is actually quite high), but he has ZERO for DistPR! What an absurd difference.
Anyway, that's enough for now, but I might look into this more later. Anyone who is better at this, the zip in the CivFanatics post I had earlier still works and lets you dig deep into a lot of these stats as you wish!
EDIT: Looking at it, I am not actually sure MaxNearPR is used specifically for Dogpile Wars or if just the DogpilePower stat is used. Also, DogpilePower I THINK is not used for Total War but only base OurPower. The observations abotu MaxNearPR going down as units are lost in a war leading to higher dogpiles remains, though. The stat for Dogpile War seems to be 2/3*DogpilePower.
Interesting stuff, Kuro. Thanks for digging into it and posting.
A question/thought: Does the aggressive AI setting being checked have any effect on dog pile war decisions? I believe Sulla described it as causing the AIs to not check power differences the way they normally would -- this helps explains some of the declarations by weak AIs against much stronger AIs. If this also changed the dog pile war decision process with its shared/joint power calculations, then mass dog piles become even more likely to occur.
A lot of people like Julius Caesar but there seems to be plenty of strong leaders in this game.
Mousey_Commander: Asoka is going to get eaten and Augustus doesn't have the iron or the skill to save his high peace weight buddy. Mao has a good backline, good resources, and is likely to benefit most from Asoka's early death. Julius is going to benefit from bullying Augustus and Willem but ultimately will be bogged down by the split directions.
ZincAlloy: Everyone, say hello to double Rome! And then say goodbye, because they're basically going to war immediately. It'll be Julius who unites the Eastern and Western empires, and then takes that combined strength to conquer the world. The only leader with a chance of stopping the Roman legions is Willem, but what's this? Willem left all his rifles back in the opening round? That's fine, you can just research some more, right Willem? Right? Riiiiiight? Oh.
Ochotona: It seems crazy but I like Asoka in this one. I expect Mao to be squeezed between culture from creativity on his left and religion on his right, hamstringing him. And I expect Julius's economy to suck a la Boudica game 6. Ultimately I'm thinking Asoka can stand up to their pressure and whizz to a culture victory.
Bobchillingworth: I'm confident that the two high-peaceweight civs are dooomed; even if the evil leaders get embroiled in stupid early wars with each other, I don't think either Asoka or Augustus will reach key military techs quickly enough to achieve lasting security. Beyond that though, I only marginally prefer Sury and Julius's positions over their fellow ne'er-do-wells, primarily because Mao will probably be squeezed and Willem's land is short on luxuries and rivers. If this season has demonstrated anything, it's that the evil civs are easily driven mad by border overlap, which occurs quickly when large portions of the map are ice and tundra. Anyone expecting an early dogpile on India or Good Rome will probably be disappointed, as the creative leaders cheese off their neighbors.
nabaxo: This is a rough picking contest. So many killers.
SpamBot: There are three two man "peaceweight teams" in this match. Mao and Suryavarman are Team 1, Willem and Julius are Team 4, and Asoka and Augustus are Team 8. Of these "teams", Team 4 are the most likely to infight. Team 1 are in position to dogpile upon this weakness by teaming up with Julius against Willem. Of course, maybe Mao and Julius and Suryavarman will all fight amongs each other, bringing all of them down and possibly killing Willem in the process, thus allowing Asoka to win by culture. But I gotta go with mah boy Mao.
Eauxps I. Fourgott: AD AETURNUM! AVE CAESAR! With the two greatest leaders in Roman history, aided by squadrons of elite Praetorian guards, there is nothing to stop Rome from ruling the entire civilized world! (Ignore the reports of strange barbarians beyond the ice fields. They are none of our concern.)
Willem: It is going to be a low peaceweight game filled with strife. I am thinking Sury and Willem are going to settle towards each other and will be involved in early conflict. That is bad for Willem who already has a pretty barren starting position. Sury is also going to be bogged down by this early war. Instead Mao is the one who is going to profit from the war between his neighbors, gobble up some extra space via expansion, and become strong enough to seek and win conflict with Asoka, perhaps helped by Julius, who has a lot of land to expand into and iron. Julius is either going to snow ball off of a successful war with Asoka or Augustus, and mop up Willem at some point. Mao will go to war with Sury at some point and win, but will be faced by a runaway Julius by then. Late-tish, domination win by Julius, with medium to many wars. Only FTD is a real tough one for me, but I am going with Asoka. Yet I really am fearing Willem won't last long.
ljubljana: culture is almost never the right way to play the probabilities...but we do have two contenders in this game, and space and domination seem roughly evenly likely if not, so i guess i might as well? also i'm tired of rooting against willem and being disappointed, so i think i'll try rooting for willem and being disappointed
Wyatan: I'll go with a very specific prediction: Sullla's shown the Iron spot that Julius Caesar'll need to grab for Praetorians. Well, based on his initial scouts deployment, I think he'll go straight for it (river + FP), which should make his supporters very happy. But here's the thing: he can't see the Iron, he can't see the Horses. To him, that's a resourceless spot, picked because there's nothing better in range. And then... his scouts will find the resource cluster to the north: Stone + Cows + Spice + Dyes. At which point he'll redirect his settler to it, like Asoka did in his opening round. So not only will he found a very late second city, he'll actively send his settler away from his only Iron! Hope I'm right, as that should make for some pretty lively chat reactions! :D
dreamyeyed: Asoka doesn't have copper and he starts close to Mao who has some right next to his capital. I doubt he'll survive long enough to get access to iron, especially if Caesar also decides to join the party. But Mao being Mao, I'm sure that he will once again pull off some ridiculously stupid move and give away a certain victory. Caesar should be able to conquer East Rome fairly easily, but where will he go after that? Augustus's territory doesn't look very good, so the snowball will likely stop right there. There are just not enough useful idiots in this game. As for the western coast, I like both of the leaders, but Willem looks like he'll have massive happiness problems with a single gold mine as his only happiness resource. Suryavarman's start looks much better, but unfortunately for him the only elephants are on the other side of the continent so he may never get his ballista elephants.
Vunterslaush: The Roman orgy will be a one-sided slaughter with augustus becoming FTD. Mao will get squeezed early by creative sury & willem plus Asoka w/ an early henge. Sury has no ivory and no early happy resources - no thanks. Julius will dominate with willem riding coattails.
smithy: Things which I think will happen but haven't dared go for: Willem FTD cos JC really doesn't like culture in his face and Sury has nothing better to do. Mao for a runaway win after he steamrollers through Asoka and then anyone else who looks at him funny. Pre T300 win cos this lot are all pretty competent. Things which will probably happen but i'm pretending won't: Willem to be just fine and win again cos he's getting good at this. AC to be scoreboard leader for most of the game due to lots of empty space and GLH. Season5 things: Asoka starts fishing-food techs, leaving the religions to go all over the place (everyone gets early myst for once). Early wars are all started due to border tension leaving Willy & Sury in trouble early. Augustus wins as noone notices he is there all game. 6 wars total as noone ever stops anymore. My lame safe predictions: JC to win with praets and he has ALL the heffalumps. RU Mao as he's generally cool with 2nd. FTD? you know who already.
Commodore: I really wanted to cheer for Mao, but lacking metal I feel like his time is finally up. Julius will prove who's the greatest of the Julio-Claudians by rolling up either Mao or Willem early, moving on to an eventual win, but Asoka will be defended by Augustus long enough to get rifling, probably managing to get in second.
eyser24: If Julius's western neighbor was not Willem I would probably choose him to win, but because Willem is creative I am very nervous that he will not get that iron that is near him. If Willem settles basically anywhere to the east, it seems really bad for Rome. The combination of Julius refusing to tech mysticism until turn 70, and Willem's culture and good expansion will overwhelm his cities and permanently cripple him, especially if he loses the iron for any length of time. I am choosing Mao to win because he has guaranteed copper, a gold nearby, and most importantly Asoka to the east, thus he will be the most likely beneficiary from Asoka's (inevitable) death. He did good with a bad start this season, here he has a good start and someone easy to wipe out closer to him than anyone else. My choice for second place was tough but its either Sury or Augustus, and I am ultimately choosing Sury. I am most worried about Augustus not getting that copper quickly and being really susceptible to Julius, and while he has proved he can really expand quickly and build wonders and create a real empire, I just see him being attacked eventually even if not early on, and eventually being overwhelmed. Sury is the wildcard to me, as I think he is more likely to go after Willem and become strong, but Sury could also attack Mao and completely ruin my predictions, or balloon himself to first place.
Bernn: You really don't want to be without copper in a game with Caesar and Suryavarman around, and especially not if you neighbor them. I don't think Asoka and Willem will get a chance to flex their economic and cultural muscles this game because of it - they're simply way too vulnerable to early attacks and in a weak position diplomatically. I don't like Mao's starting position at all and I think he'll be a minor power at best this game, leaving Suryavarman and Caesar as the real contenders to me. Between the two, I'm going with Sury. He's a little less prone to crashing his economy than Caesar and Creative does helps enormously in the early stages of the game. Willem is going to get crushed by the Khmer very early (sub T100!) and Sury will smoothly cruise to a win off his Dutch conquests.
(August 5th, 2020, 09:38)haphazard1 Wrote: Interesting stuff, Kuro. Thanks for digging into it and posting.
A question/thought: Does the aggressive AI setting being checked have any effect on dog pile war decisions? I believe Sulla described it as causing the AIs to not check power differences the way they normally would -- this helps explains some of the declarations by weak AIs against much stronger AIs. If this also changed the dog pile war decision process with its shared/joint power calculations, then mass dog piles become even more likely to occur.
From what I can tell, Aggressive AI has at least the following effects (and maybe more I couldn't find).
1. It causes the AI to believe it is more powerful than it actually is, 33% more powerful to be precise (done by multiplying their OwnPower rating by 33%). It does not seem to remove it entirely, but it is likely that this power check increase causes weak AI to almost always be in range to at least possibly declare on someone massively stronger than them, because the main check for base war is 3/2*OP*MaxNearPR or MaxDistPR, so the increased sense of own power then is multiplied again. Since the effect starts with the base of their own power, this also means Aggressive AI should affect AI that build lots of units "more" since they have more to multiply. This might mean AI with high unit build but more middling other aggressive stats are more likely to declare war under this setting. Note that LimitedWar and DogpileWar also take into account OwnPower but at different numbers.
2. A flat +10 is added when rolling for potential victims to each victim. This doesn't affect the AI's desire to start checking for potential war victims, but it makes them more likely to find one (and therefor more likely to start a war) AND more likely to increase the number of potential targets by letting them hit the minimum check. This should make it more likely that low-chance AI wars happen in specific: A war with a high chance of happening probably happens anyway, while a war that is only a 2% chance or whatever might have 10 new numbers that make a war due to +10.
3. I'm not sure if I am reading this correctly, but as I understand it the AI is more likely to declare war while paying for lots of units with Aggressive AI on. From what I can tell, and I again could be wrong, the AI multiplies the base chance to declare total / "normal" war (but not Limited or Dogpile) by it's HUSP score + 1, HUSP being a rating of how much the AI is paying units. Non-Aggressive AI HUSP is Max(0,USP-7)/3. USP = UnitSpendingPercentage, checked by doing 100 * Unit Cost / TotalCost. Agressive AI's HUSP is listed as "*2 + 1), which makes me think it is either multiplying the HUSP value of non-aggressive or flatout just multiplies the base USP value for HUSP. Whatever the specific amount, Aggressive AI makes it more likely for the AI to attack if it is paying a lot for units but not totally broke. Like with the OwnPower check, this should affect AI that make more units more as well.
4. Without Aggressive AI, the AI has a 25% per turn to simply not be "in the mood" for a war before checking anything else for war if they are not broke and are not pursuing a "Dagger" strategy that only specific AI pursue. This is likely a huge general boost to how often an Aggressive AI goes to war by simply removing a flat 1/4th chance per turn to not even bother checking war.
Barring information I don't know, my own misunderstanding or whatnot, that is all the effects Aggressive AI has on how likely the AI is to declare war.
EDIT: Playoff Game 2 Predictions
Winner: Julius Caeser
Runner Up: Suryavarman II
First to Die: Willem
War Total: 11
End Date: T298
Victory Type: Domination
Comments: Had to submit this very late, so a shorter explanation than normal. Willem and Caeser along with Mao and Asoka will have major border tension, Mao and Asoka seem pretty much destined to fight at some point. I think Caeser will attack Willem despite peaceweights: Willem likely will have a very weak military (no units until iron, low military tech weight, mediocre unit build) while Caeser 100% has copper (meaning he will build his Axes earlier, spike in power earlier and be up for wardeccing earlier) and almost certainly will get iron for Praetorians.
Once Caeser gets Praetorians, Willem crumbles much faster than Asoka. Caeser is MUCH more willing to wardec on Sury and Mao than Mao is on Sury, as well. My first thought was Mao 2nd actually, buuut...Sury has a good chance to join in on the war with Willem and gain shared war bonus with Caeser. I also think Mao's conquest of Asoka might take awhile and leave him behind. Sury also has a good chance to found a religion and spread it to Caeser (even if by conquest of Willem). So I think Caeser turns on Mao (be it before or after Augustus), eats him with a superior empire, and then just rolls over the game. His Imperialistic/Organized also seems really good for this game, good expansion room and econ.
Game of fast conquest, so he wins pretty fast. Most wars fight to the finish, so only 11 wars.
Sneaky alternate thought: Asoka can definitely absorb enough aggression for Augustus to win this game, particularly if Caeser and Mao decide to fight instead which is not unlikely (Caeser's 4 peaceweight is roughly as far from Augustus/Asoka's 8 as from Mao/Sury's 1). Augustus could also conquer someone if they attack him or he attacks with Praetorians. I honestly would not be surprised if he won.
Thanks for checking the aggressive AI stuff, Kuro. Thinking they are stronger than they truly are explains a lot, and your point #4 also seems like it would make a big difference. It is great to still be learning new stuff about this game, even after playing it for years.
(August 7th, 2020, 19:41)Fluffball Wrote: The 3rd playoff game has a stronger field by far than our current championship game.
The second definitely did too, lol. That might've been the strongest playoff game ever, or at least competing for it, based on power rankings. In fact, if you include this Season's points, every leader minus Augustus in Playoff Game Two will be a seeded leader next year.
Playoff Game Two is finished! We had another big turnout with more than 200 viewers on the Livestream, thank you so much to everyone who was able to join the group today. I hope it was an entertaining game to watch! Here's a batch of links for the next game:
Congratulations to Ethan as the sole winner of Playoff Game Two's picking contest with a score of 18 points. Scoring was low across the board for this game with an average score of 5.78 points thanks to some of the community favorites facepalming in various ways. I also need to point out that Random.org tied for 4th place in this game with 16 points. We are continuing with our normal Friday date for next week's game. Next week we continue with the third and final playoff game as our returning winners come back for another round of action. Who do you have this week?
![[Image: survivor5-11-8.jpg]](http://www.sullla.com/Civ4/Survivor/S5/survivor5-11-8.jpg)
![[Image: survivor5-11-9.jpg]](http://www.sullla.com/Civ4/Survivor/S5/survivor5-11-9.jpg)
![[Image: survivor5-11-10.jpg]](http://www.sullla.com/Civ4/Survivor/S5/survivor5-11-10.jpg)
![[Image: survivor5-11-11.jpg]](http://www.sullla.com/Civ4/Survivor/S5/survivor5-11-11.jpg)
