New Version - December 1st (12/1)

[CrazyG]

Military power is cut by the number of cities you have. This is to counter the fact that larger empires get more troops and automatically win strength contests like that. Also helps with AI calculations.

G

That explains it. Its still really weird to me though

 

[IcyAngel]

I am a little bit confused now. In my current game as Ethopia (Standard Map Size, Standard Speed, Continents, Emperor), there are not just 2 big landmasses, I have my own with one city state, the next to it connected via one coastline access has Persia and the Maya on it and the Maya has a connection via coast tiles to a landmass with 3 CS and one natural wonder.

It is turn 179 now, I am second behind the maya, but the map is not yet completely discovered and one civ is still missing for me. But the point is, that the Maya popped 3 admirals at that point, he warred with persia, but on his own landmass, so I doubt they had big naval warfare, even with the mayan UA, the choice of Artistry (the wonders I have to check, but pisa tower wasnt build yet) it seems a bit weird to me that he already had 3 admirals ...

(... at least it shows that he focused a lot)

You get Great Admiral points for hitting naval units with ranged land units.

Probably a big factor here.

 

[randomnub]

I'm in the middle in my first game with this version (Deity). On average, AI's are not as good as the 11/26 patch, but there is a huge difference between them. The best civ (Arabia) had a 20 tech advantage (+2 policies) and the worst (2 civs) had a 10 tech disadvantage (and one less policy tree). Most were on par with me. I'm in the renaissance Era.

I will try another game without Arabia (or Egypt for that matter).. maybe they wonder-spam too much and it generates historic events that gives them this huge advantage.

Same feeling here.

In my recent game Polinesia spawned landlocked (I count a grand total of 4 moais on the map on some expos, all connected by land, so it's pretty much a civ w/o any bonus) and had an average early game, warring a bit with Shaka in what resulted to a stalemate until Kame took the lead after spamming some wonders. The lead became substantial, I found myself as his only competitor for wonders and started get some just to deny him easy historical events.

Look at the situation, I was doing okay (+400gpt, great production, 9 CS allies + some friends and ready for Germany prime time) and on par with other AIs except Austria that is my vassal and Kame that in the last few turns got a honestly broken lead. When my best return is sending trade routes to him instead than city states it's time to be worried. Turn 168 Deity standard. For the record, Napoleon is in a lot stronger position in-game, with better dirt/monopolies/city placement and waging some successful wars against Maria/Shaka but he's right there with me playing a fair game.

Then the last couple turns simply made me ragequit.

I have Events OFF, so no GE points from anywhere. The spy in Honolulu shows me that Kame is pumping out wonders in 1 turn in the same city, with excess production as well. Goes without saying he's also influential with 6 civs and the snowballing is real, I blame too many bonus yields from historical events.

 

[randomnub]

This is also worth show, Kame food surplus and Honolulu been growing +1/turn for some time now. He has not Apostolic Tradition, no events or any other food source I can think about.

Look at Berlin, it's a way better city... in theory , I'm 10 techs behind, duh.

 

[randomnub]

I played just one game so the sample is limited, but I don't feel like starting another to witness even better tall AIs taking such lead. Historical Events are already a reward by themselves: while the earlier game was indeed fun and balanced and the idea of less frontloaded boni is commendable, late game could use some more fine tuning and less carrots for wonderwhores.

 

[Moi Magnus]

I played just one game so the sample is limited, but I don't feel like starting another to witness even better tall AIs taking such lead. Historical Events are already a reward by themselves: while the earlier game was indeed fun and balanced and the idea of less frontloaded boni is commendable, late game could use some more fine tuning and less carrots for wonderwhores.

The question is "Should we nerf tall AI or should we nerf tall play ?". Because the starting points of those changes was that Tall AI had almost no bonuses while Wide AIs had a lot of bonuses, with the consequence that Tall was easier for player than for AI, making the balance difficult.

 

[LukaSlovenia29]

In my two games with the new version, my experience is similar to that mentioned above, i.e. one AI pulling ahead massively despite a normal amount of cities. For example, in my current game, Assyria has 5 cities (progress, artistry) and is 10-15 techs ahead of the pack. Not sure what the reason is, just adding my observation that may or may not be a simple aberration.

 

[randomnub]

I think the one and only question is "Should we fix tall AI?". While not perfect, previous patches hit a better balance spot so I just put Gazebo on the naughty list.

 

[vuko]

@Gazebo does the AI receive the -10% penalty for every next Wonder? If not, it could explain that one wonder whore that everyone has in their games.

 

[Owlbebach]

It kinda seems to me that the idea of quadratic formula for AI handicap was right, this is more about setting right numbers, which requires more testing with different numbers. For example the exponent can be reduced and we can add a constant there, so AI will be stronger early on and less runaway, but still competitive in the mid/end-game. For example something like iYieldHandicap = 10+iHandicap*(iEra^1,75) (right now it is iYieldHandicap = (iHandicap * iEra^2).

 

[randomnub]

Play a game toying with those settings would you? I'm in a very low mood after 4 aborted games

 

[CrabHelmet]

I think this is reflecting what @CrazyG and @ElliotS were saying - Tall has always been better than Wide, but because the AI only got bonuses for Wide play, it wasn't apparent in @Gazebo's playtesting. Now the AI's bonus are indifferent between Wide and Tall, you can see how good Tall really is.

If you think that Polynesia is bad, try starting games with Arabia. I just quit a Spain game at turn 81 because he had built Stonehenge, the Pyramids, Mausoleum, and the Great Library already.

 

[randomnub]

I know, as I wrote in my situation Poli is in a terrible situation and it's just lucky it's the only one who went full Tall. Any comparison to a successful Arabia/Ethiopia/Korea just exacerbates the issue.

It's not even fair compare the amount of triggers a 'wide' civ gets: amount of settling cities is limited by space, plus more cities lower your scientific/cultural output so sort of balances off, and the amount of GA/Eras trigger is pretty much set for all; only 'winning a war' trigger can be farmed but warring consumes a lot of resources. It's true that pre-patch tall AIs weren't as good as wide ones, with tall being a better strategy for the player only, but wide ones didn't snowball that early and that hard either. Tall AI still played a good role and in some cases could pull out a victory anyway, it wasn't broken (player 'fun factor' wasn't in a bad shape), just not optimal.

Now it's broken, and some numbers have to be tuned down, or have different values for different HE. Great People and Wonders come in big quantity, and once you snowball a category there's nobody stopping you from getting more of the same (see that wonder production overflow and 1 wonder-per-turn), while land-grabbing and warring has heavy diplomatic counters: the amount of runaway Genghis is balanced by the amount of dogpiled Genghis in my experience, while peaceful Haile is usually webbed into a defensive pacts net.

 

[Owlbebach]

Now it's broken, and some numbers have to be tuned down, or have different values for different HE. Great People and Wonders come in big quantity, and once you snowball a category there's nobody stopping you from getting more of the same (see that wonder production overflow and 1 wonder-per-turn).

One thing that i have to say is that i saw even more ridiculous runaways before.

Overall i think we found a very important target in the balancing process and it should be paid a lot of attention. Clearly the previous formula was not good, we need to figure out the new one

 

[randomnub]

One thing that i have to say is that i saw even more ridiculous runaways before.

95% of my games are multiplayer. And out of them 90% are Pangea cause i like the game to be tough

I feel we're getting back to the donut balance dilemma.

 

[CrazyG]

Looks like the problem is food/production rewards for historic events.

Does the AI need these kind of boosts? I don't think the reward for city settling was a good choice either, giving a boost on era advance is fine because everyone gets those, but not all AI settle cities or build wonders, it encourages runaways

 

[Owlbebach]

I feel we're getting back to the donut balance dilemma.

Nope, we are not

 

[randomnub]

Then post some actual game experience, play out your games.

 

[LukaSlovenia29]

If possible, I like Randomnub's idea of different values for different types of historical events. Archeological digs, great persons or wonders should provide significantly less than settling a city, conquering a city, winning a war..., as they're more common and have no inherent downsides (diplomatic repercussions etc.).

 

[CrabHelmet]

So what do we think the precise nature of the problem is? My guess is that: the first AI to reach a Wonder gets a big production bonus that means they're the first AI to reach the next Wonder which means they get a big production bonus which means they're the first AI to reach the next Wonder ad infinitum. Compare this to the other forms: the AI takes a long time to conclusively win wars, so that bonus doesn't snowball, Cities have global diminishing returns, which means that bonus doesn't snowball (the gifted Science doesn't compare to the 8% drag in the long run), and starting a Golden Age is pretty infrequent all things considered.

So it seems to me the problem is relatively clearly that the scaling difficulty of construction on Wonders is insufficient.