Brutal AI

[Forsti]

Also, maybe the Babylonians simply rushed some advanced tech before getting the ones left from the previous eras ?
I know I sometime go for Archeology, which you can get with simply 1 tech per column in the last 2, and only a few before; then the old techs go very quickly, because of the higher science/turn and the bonus because everybody researched them.

Before Babylon's tech spike starts, they have 38 techs. This means that the tech they rush can't be further than late industrial. If that is the case, they would have 4 first column renaissance techs, 2 second column renaissance techs, 1 first column industrial tech and 1 second column tech when they get their 39th tech and the spike starts. Had they filled the tech tree as evenly as possible, they would have 5 first column reneissance techs and 3 second column reneissance techs. Even with maximum rush on the tech tree, they can only have substituted 2 reneissance techs to 2 industrial techs. It can't explain the pace in which Babylon managed to get 28 techs or in other words, advanced 3 eras.

If I'm looking at the graph correctly, it seems that Babylon got 28 techs in 31 or 32 turns. The x-axis is split to 40 parts and as there were just less than 300 turns in the game according to the OP, each part consists of around 7.5 turns. Babylon's tech spike took just over 4 parts and they got from 38 to 66 techs during that time. For me it seems like something is not working. I edited OP's screen shot a little to show what I'm talking about.

Spoiler :

 

[albie_123]

Zulu and Japan seem to go through a similar period, albeit a little later.

 

[IcyAngel]

Actually it was correct, you follow this path and go to DifficultyMod.xml
At the bottom you have a line
"<AIPerEraModifier>-10</AIPerEraModifier>" which lower every yield, include science I guess, and stack with eras.

Why ignore my previous post? actually it was question/suggestion more than complain, but I'm not native english speaker, sorry if it feel like that.
If I'm still playing Civ V, it's because of this mod, I just want try help improve it

or maybe it's just one subject per thread and I should PM or something?

Your English is good! I'm a mere foreigner too, anyway. Sometimes I don't read every post, don't take it personally.

And I'm not convinced the link is correct. There used to be a file with the precise numbers per difficulty level (on food growth, science, culture and so on), and I'm pretty certain it wasn't 10% across the board!

 

[Aristos]

You want a specific answer, as if this is a hard-coded 'do x at 40 tech' situation. It isn't. There's no hard-coded answer. WC + Rationalism + Golden Ages + Meeting all Civs + Handicap.

G

Come on man, you know me better than that. That would be a stupid question to ask, I wouldn't ask for such an answer.

I am looking for coded logic as an explanation. I want to be defeated by an excellent AI, no doubt... a game were I am assured to win is utterly boring to me. Utterly. But obviously, I also want to "learn" from the defeats, and more or less see where and why I was surpassed by the opponent (AI or human). For example, if/when Deity vanilla AI crushes me, I know why, and it's usually the huge, ridiculous bonuses they get (which I don't like).

This time, I am trying to find an explanation for something that is clearly marked by the graph I posted. I just want to know that this AI really became that good, and thus have a nice party with it from now on (defeats included). I also want to "learn" from it if that is even possible.

Right now though, I still cannot explain how Babylon (and similarly other AIs in same game) jumped up to 25+ techs in as few as 30+ turns. I am not finding a way for me to do something similar, thus my curiosity. I know I am not a mediocre player or anything, yet I am failing at finding such explanation.

That's all. Not looking for specific answer, nor lame "if/then" hard code blocks.

 

[Aristos]

Before Babylon's tech spike starts, they have 38 techs. This means that the tech they rush can't be further than late industrial. If that is the case, they would have 4 first column renaissance techs, 2 second column renaissance techs, 1 first column industrial tech and 1 second column tech when they get their 39th tech and the spike starts. Had they filled the tech tree as evenly as possible, they would have 5 first column reneissance techs and 3 second column reneissance techs. Even with maximum rush on the tech tree, they can only have substituted 2 reneissance techs to 2 industrial techs. It can't explain the pace in which Babylon managed to get 28 techs or in other words, advanced 3 eras.

If I'm looking at the graph correctly, it seems that Babylon got 28 techs in 31 or 32 turns. The x-axis is split to 40 parts and as there were just less than 300 turns in the game according to the OP, each part consists of around 7.5 turns. Babylon's tech spike took just over 4 parts and they got from 38 to 66 techs during that time. For me it seems like something is not working. I edited OP's screen shot a little to show what I'm talking about.

Spoiler :

Thanks a lot, that is exactly what I was trying to express with words, but a good, edited graph beats a thousand words.

I am trying to find a way to do something similar, with the same civ/leader to be able to understand such a jump (and learn from such a nasty AI ), but so far I couldn't.

Perhaps this could be a nice challenge? Any ONE player that can show a similar jump of 25+ techs in 30+ turns will have found a way, or the way, and kudos to both him and Gazebo's AI.

 

[Moi Magnus]

Thanks a lot, that is exactly what I was trying to express with words, but a good, edited graph beats a thousand words.

I am trying to find a way to do something similar, with the same civ/leader to be able to understand such a jump (and learn from such a nasty AI ), but so far I couldn't.

Perhaps this could be a nice challenge? Any ONE player that can show a similar jump of 25+ techs in 30+ turns will have found a way, or the way, and kudos to both him and Gazebo's AI.

Do you remember if Babylon was friend with Japan and Greece (even few turns) ?
The first part of their "jump" is synchronised with a jump for Japan, and the second one with Greece.
If that's the case, it really look like a research agreement bug.
(Either the value of the research agreement was stupidly hight, either the research agreement was duplicated a higth number of times)
The England jump is understandable : They just need to have very efficiant spies, and maybe a research agreement.

Having 15 techs in 30 turns is doable if you stack a lot of bonuses, but 25+ techs when you already lead seems too much to be possible...

 

[Aristos]

Do you remember if Babylon was friend with Japan and Greece (even few turns) ?
The first part of their "jump" is synchronised with a jump for Japan, and the second one with Greece.
If that's the case, it really look like a research agreement bug.
(Either the value of the research agreement was stupidly hight, either the research agreement was duplicated a higth number of times)
The England jump is understandable : They just need to have very efficiant spies, and maybe a research agreement.

Having 15 techs in 30 turns is doable if you stack a lot of bonuses, but 25+ techs when you already lead seems too much to be possible...

Nope, as I said from the very beginning, no RAs where in place because it was a strangely unfriendly game and no single DoF was signed.

 

[Aristos]

OK, so I just did a quick Np++ compare between Difficultymod.xml files from 1.1.23 and 1.7.7 versions (as 1.1.23 is as far as I remember a version being ostensibly easier than 7.7), and the files are almost identical (bonuses almost identical). The only difference between the versions is a DifficultyBonus value which I don't know what it does, but I don't think for a minute it has any influence in explaining the supertecher AI.

Bottom line: handicap bonuses are not the explanatory variable, exactly as Gazebo said. Which, to me at least, is a huge relief as I dislike super-boosted AI with bonuses.

 

[Flamen]

Deity AI in 1.7.7 seems to be a little bit op.I have played 2 deity games on 1.7.7 lost both due to AI beeing 2 or more eras ahead.(Never happened to me on privious versions)
Might have been just unlucky games but i severly doubt it.

 

[Funak]

Deity AI in 1.7.7 seems to be a little bit op.I have played 2 deity games on 1.7.7 lost both due to AI beeing 2 or more eras ahead.(Never happened to me on privious versions)
Might have been just unlucky games but i severly doubt it.

Is that a bad thing?
Having a top level that's almost unreachable is preferable to having a toplevel that you know you can beat every time.

OK, so I just did a quick Np++ compare between Difficultymod.xml files from 1.1.23 and 1.7.7 versions (as 1.1.23 is as far as I remember a version being ostensibly easier than 7.7), and the files are almost identical (bonuses almost identical). The only difference between the versions is a DifficultyBonus value which I don't know what it does, but I don't think for a minute it has any influence in explaining the supertecher AI.

Bottom line: handicap bonuses are not the explanatory variable, exactly as Gazebo said. Which, to me at least, is a huge relief as I dislike super-boosted AI with bonuses.

You didn't know? G fixed a bunch of stupid bugs including one that actually made the AI completely ignore buildings.
We did a few AI-matches on the old version where we went in and checked the building-situation in AI cities and 90% of the cities didn't even have all ancient-era buildings done. I think 2 cities out of 50 on the map actually had libraries, and one of those libraries were from building the great library.

Seems logical that the AI would perform better if it could actually construct buildings.

 

[Gazebo]

You didn't know? G fixed a bunch of stupid bugs including one that actually made the AI completely ignore buildings.

Wasn't a 'stupid' bug - was an extremely complex issue that took me hours and hours to figure out. Besides, the AI's been gaining steam for months now on the city production front thanks to my rewrite of construction logic back in March.

G

 

[Aristos]

You didn't know? G fixed a bunch of stupid bugs including one that actually made the AI completely ignore buildings.
We did a few AI-matches on the old version where we went in and checked the building-situation in AI cities and 90% of the cities didn't even have all ancient-era buildings done. I think 2 cities out of 50 on the map actually had libraries, and one of those libraries were from building the great library.

Seems logical that the AI would perform better if it could actually construct buildings.

Of course I did know, I followed notque's videos... what has that do with the post about bonuses not being the culprit?

 

[albie_123]

We did a few AI-matches on the old version where we went in and checked the building-situation in AI cities and 90% of the cities didn't even have all ancient-era buildings done. I think 2 cities out of 50 on the map actually had libraries, and one of those libraries were from building the great library.

Out of curiosity, how do you go about doing this (seeing what an AI has built in their city)?

 

[Funak]

Out of curiosity, how do you go about doing this (seeing what an AI has built in their city)?

We used InGame Editor to switch ownership and check the buildings out.

 

[tu_79]

It seems like a bug to anyone less polite than Aristos. If there's no explanation of how it is possible to have such jump within the rules, then it's a bug.

Is it possible to check the log, and see what happened?

 

[Aristos]

It seems like a bug to anyone less polite than Aristos. If there's no explanation of how it is possible to have such jump within the rules, then it's a bug.

Is it possible to check the log, and see what happened?

This is the first time in this subforum that someone, anyone, refers to me as "polite".

Now THAT must be a bug.

I was trying to replicate the behavior, at least to some extent, by playing the Babylonians myself, but now G came up with a new version and I am torn apart... do I continue this wonderful game I had with 1.7.7, or do I play with the new toy?

Curse you foul demon G, as notque likes to say.

EDIT: there is a note in the new 1.7.15 version about some bonuses for the AI when conquering cities that were "snowballing", but did not ask G about the meaning... was that perhaps some "pseudo-confession" about this issue? I don't want to ask him, enjoying my sure-to-be-very-temporary "polite" reputation here and now...

 

[tu_79]

If you want to replicate, play a game with same civs run by the AI, and let the log record everything.

Thanks for the compliment!

 

[Aristos]

If you want to replicate, play a game with same civs run by the AI, and let the log record everything.

No, I meant "replicate" as in "find how the AI did it by doing the same myself", not as in duplicating the result.

Por nada, hermano! (literalmente, no veo ningun halago de mi parte que agradecer..)

 

[tu_79]

No, I meant "replicate" as in "find how the AI did it by doing the same myself", not as in duplicating the result.

But if it happens to be a bug with AI, you won't be able to replicate it being the player.

 

[Aristos]

But if it happens to be a bug with AI, you won't be able to replicate it being the player.

EXACTLY. But if me or anyone else, just one person, can do something similar with the babs, then it's not and we have Skynet.