Version 1.00 discussion thread

jdog5000 · May 17, 2010

(You can always get the latest stable release from the download thread or sourceforge. Development versions and updated source code can also be checked out from sourceforge.)

[party]

Version 1.00 [party]

This is version 1.00 because it has a compilation of several big new systems that I've been working on for a while, and a giant leap forward for the BTS AI. Here's just a couple of highlights:

New victory strategy system, so AI actually goes for conquest and space victories
Major overhaul of AI city attack stack logic and ability to defend itself
AI now weighs having stacks walk or sail in transports
Lead from Behind 1.2 added to help AI preserve key units
And a lot more ...

BBAI 1.00 is significantly faster than any previous BBAI release. It plays turns on a Huge Earth scenario with 18 civs in the year 1980 in just under 30 seconds, a 15% improvement over version 0.8x and a hair faster than the UP and plain BTS. Finally, there are also a couple key bug fixes, plus UP 1.5 of course.

I'm not sure if work on the AI could every really be finished, and certainly there are some big things I'd still like to do (centralized wonder selection, nuke decision making, tech selection system). However, usually when I package up a release I immediately start thinking about the next important piece. This release has come far enough that I think I'll pause and play through a couple whole games myself ... that was very hard to do when all I would see were AI weaknesses and deficiencies, but the project really has made immense progress and I think I might actually be able to play through a whole game or two.

Since this is a big milestone, I'd also like to thank and acknowledge a few fellow modders who helped make this a reality one way or another. Much credit goes to Blake and Iustus who really pioneered fan-developed AI, Blake's work made a huge impact on the Civ4 AI leading up to BTS and his example was one inspiration for this project. Also special thanks to Yakk and glider1 for batting around ideas to fix the original closeness bug which really got this all started. I would also like to thank the others who one way or another helped me get into modding, taught me some useful tricks, or otherwise helped make the last four years of Civ modding such a great experience: Kael, Solver, TheLopez, DrElmerGiggles, Dom Pedro II, EmperorFool, glider1, phungus420. And, of course, this mod would not be where it is today without the great community here on the forum and everyone who has posted bugs, ideas, and engaged in our discussions. Thank you!

Enjoy!

New in Better BTS AI 1.00

Spoiler :

LunarMongoose · May 17, 2010

Yes, I know how you feel. Daryn and I have yet to get past the middle of the tech tree in a game now for a very long time, because new versions of BBAI kept getting released and because I kept wanting to make major upgrades to my full mod on top of that. Believe it or not, MongooseMod 3.5... which I'm going to release soon, I swear... feels very much the same: not yet finished, but getting there; all major weaknesses in Civ4 as a whole, modding-standpoint-wise, finally tackled. Obviously without BBAI I wouldn't have stuck with this nearly as long as I have (and even then I got completely dragged away by other games the last 3 months heh), so... thank you as well dood.

Now the moment I've been dreading... I have the daunting task of merging my codebase with yours once again, a process which takes 4 hours when you've posted a relatively SMALL update... *gulp*

Ninja2 · May 17, 2010

Wow. That is one huge list, jdog!! Congrats for finally finishing 1.00. You deserve a couple of games, that's for sure!

Now onto the great task of merging this in with current mods... *dons gloves and wielding helmet*

NotSoGood · May 17, 2010

Do I need to make some changes to my makefile that are needed for this version?

EDIT: Compiling but got error about CvTeam::declareWar. In CvTeam.h it is defined like this:

Code:

	DllExport void declareWar(TeamTypes eTeam, bool bNewDiplo, WarPlanTypes eWarPlan);

But in CvTeam.cpp the function looks like this:

Code:

void CvTeam::declareWar(TeamTypes eTeam, bool bNewDiplo, WarPlanTypes eWarPlan, [B]bool bCancelPacts[/B])

bool bCancelPacts isn't defined in the header file. :nono:

MaxCiv · May 17, 2010

Great job Jdog! Thanks to you an everybody who helped you creating this.

Fimbulvetr · May 17, 2010

NotSoGood said:
bool bCancelPacts isn't defined in the header file.

Are you sure about that?
CvTeamAI.h r560 changes

NotSoGood · May 17, 2010

Fuyu said:
Are you sure about that?
CvTeamAI.h r560 changes

Interesting, because I downloaded mine from our database. It says it's version 1.00 and and in team.h bCancelPacts is missing.

Fimbulvetr · May 17, 2010

Does it say 1.00e? Then that change should be there. Unless jdog forgot to pack the latest code into that zip, in which case you'll have to get the sources from svn.
Anyway..
Thank you, jdog, and keep up the good work. But playing is good too, I haven't done much of that myself either lately -_-

Ninja2 said:
Now onto the great task of merging this in with current mods... *dons gloves and wielding helmet*

The obvious merge is already done:

LunarMongoose · May 17, 2010

NotSoGood said:
Interesting, because I downloaded mine from our database. It says it's version 1.00 and and in team.h bCancelPacts is missing.

Yep, I got mine from CivFanatics and my header was missing the 4th argument as well, however all but one of the many places that call the function only pass 3 args anyway so I just removed the extra one and used the value used in the one place that does call it with 4 args. Hopefully nothing else about the files was non-current.

Two other notes... JDog you snarfed my getNumActiveWorldWonders() function in CvCity (changing iValue to iCount lol), and my multi-Holy functionality in CvPlayerAI but without the function. Minor rewrite to code an extra bonus for a state-religion-holy-city, but otherwise that's exactly my AITargetCityValueFix component. C'mon, don't be a llama.

Secondly, you have a BBAI ToDo comment, not sure how new it is but my first time seeing it I think, expressing confusion about the vanilla civic upkeep code in CvPlayer::getSingleCivicUpkeep(). If you read the description for my CivicUpkeepMod component, that explains what the vanilla code is doing... it is pretty stupid imo, but for what it's trying to do the code is correct heh.

LunarMongoose · May 17, 2010

I compared the zip with the SVN and it looks like CvTeam.h is the only non-current file, with this one line being the only difference. I WAS going to do exactly what it does to try and fix it, but I couldn't remember how to define an optional function argument... I feel very stupid right now. At least what I did will still work lol, but talk about doing it the hard way.

NotSoGood · May 17, 2010

LunarMongoose said:
I compared the zip with the SVN and it looks like CvTeam.h is the only non-current file, with this one line being the only difference. I WAS going to do exactly what it does to try and fix it, but I couldn't remember how to define an optional function argument... I feel very stupid right now. At least what I did will still work lol, but talk about doing it the hard way.

Well that's good to know. Thanks for checking it.
By looking at the SVN it seems that jdog meant the bCancelPacts to be true in most cases. So if you want to add the 4th argument without adding it to all other calls, you can set a value to it in CvTeam.h where you define it:

Code:

	DllExport void declareWar(TeamTypes eTeam, bool bNewDiplo, WarPlanTypes eWarPlan, [B]bool bCancelPacts = true[/B]);

Fimbulvetr · May 17, 2010

LunarMongoose said:
(...) that's exactly my AITargetCityValueFix component. C'mon, don't be a llama.

Yes that's the component. Why is he llama now, your fixes have had the destiny to become part of the Unofficial Patch since the beginning of time (as long as they don't break savegames - otherwise your GPP fix would be included too by now).
This time, I was to one to suggest them but I'm confident they would have found their way into the UP without me just as well. So, when are you releasing 3.5 ?

jdog5000 · May 17, 2010

NotSoGood said:
Do I need to make some changes to my makefile that are needed for this version?

EDIT: Compiling but got error about CvTeam::declareWar. In CvTeam.h it is defined like this:

Code:

DllExport void declareWar(TeamTypes eTeam, bool bNewDiplo, WarPlanTypes eWarPlan);

But in CvTeam.cpp the function looks like this:

Code:

void CvTeam::declareWar(TeamTypes eTeam, bool bNewDiplo, WarPlanTypes eWarPlan, [B]bool bCancelPacts[/B])

bool bCancelPacts isn't defined in the header file.

Sorry, that one file slipped through when I zipped the source code. Fixed.

LunarMongoose said:
Two other notes... JDog you snarfed my getNumActiveWorldWonders() function in CvCity (changing iValue to iCount lol), and my multi-Holy functionality in CvPlayerAI but without the function. Minor rewrite to code an extra bonus for a state-religion-holy-city, but otherwise that's exactly my AITargetCityValueFix component. C'mon, don't be a llama.

Secondly, you have a BBAI ToDo comment, not sure how new it is but my first time seeing it I think, expressing confusion about the vanilla civic upkeep code in CvPlayer::getSingleCivicUpkeep(). If you read the description for my CivicUpkeepMod component, that explains what the vanilla code is doing... it is pretty stupid imo, but for what it's trying to do the code is correct heh.

Sorry, should have given you credit for those. FWIW, I think I wrote myself a note to handle that in the UP, then when it came time to implement didn't remember that you'd already done it and I then reinvented the wheel. I just added the appropriate comments.

Thanks for the heads up about CvPlayer::getSingleCivicUpkeep, that code looks truly bizarre ... I'm glad you figured out that it's working properly, might have to take another look at that later.

Maniac · May 17, 2010

Impressive changelog!

jdog5000 said:
I'm not sure if work on the AI could every really be finished, and certainly there are some big things I'd still like to do (centralized wonder selection, nuke decision making, tech selection system). However, usually when I package up a release I immediately start thinking about the next important piece. This release has come far enough that I think I'll pause and play through a couple whole games myself ... that was very hard to do when all I would see were AI weaknesses and deficiencies, but the project really has made immense progress and I think I might actually be able to play through a whole game or two.

Suggestion for the future: :mischief:

- AttackSea have pretty low attack odds threshold ... how to coordinate buildup of naval strength to take out strong enemy units?

Roland Johansen · May 17, 2010

Wow, this is huge jdog. Have some fun with your reinvented game. :goodjob:

LunarMongoose · May 17, 2010

Fuyu said:
Yes that's the component. Why is he llama now, your fixes have had the destiny to become part of the Unofficial Patch since the beginning of time (as long as they don't break savegames - otherwise your GPP fix would be included too by now).

He was a llama cuz he didn't give me any credit for it like he usually does. Which he now has. So now he's a giant flaming demon llama! ... That's a good thing, means he can go out and conquer the universe. *ahem* Nevermind.

Fuyu said:
So, when are you releasing 3.5 ?

Ooh, didn't know anyone cared enough to post about it.

Well I'm still working on it, have been for about the last two weeks, I'm hoping to be finished enough to be able to start a new game with Daryn either late tonight or tomorrow sometime, then I have a few days of work to do on things I can do post-start without harming anything, then I'll need a day or two to make a public version (gotta strip out all the copyrighted music I'm using in the era soundtracks in my private version for example) and finish writing the descriptions in my thread... And it wouldn't hurt to test everything for a while in this new game we're starting; BBAI 1.0 + all my code was no simple merge, it went pretty smoothly overall but not perfectly (and did take around 6 hours, as I feared lol). So yeah, hopefully within 2-3 weeks, we'll see.

jdog5000 said:
Sorry, should have given you credit for those. FWIW, I think I wrote myself a note to handle that in the UP, then when it came time to implement didn't remember that you'd already done it and I then reinvented the wheel. I just added the appropriate comments.

Yeah I figured it was something like that... or else that someone else had snarfed it in their mod and you took it directly from there without knowing. Thanks dood! I wuv yew. *hugzors*

jdog5000 said:
Thanks for the heads up about CvPlayer::getSingleCivicUpkeep, that code looks truly bizarre ... I'm glad you figured out that it's working properly, might have to take another look at that later.

I found that mess a number of months ago (with much the same reaction lol), and if you analyze their code you can eventually figure out what it's doing... Basically it makes it so the further left in the Civics Advisor a civic column is, the more expensive all civics in that column are relative to other further-right categories. So that Government civics, even the Low Upkeep ones, probably cost more than High Upkeep Religion ones, for example. That's the main effect, it may be doing something else too I forget. But yeah my little component just took all that out basically, and made X Upkeep cost the same regardless of which civic type it was in, b/c playing with it the vanilla way was driving me nuts. It also made balancing the vanilla civics, and my mod's additional civics, a big pain because I would've had to balance their effects against gold costs that varied over a much wider range than the differences in the relative power of the civics do.

Edit: Lemme add another example cuz I wanna be extremely clear about this. I would describe the actual cost variance across columns with the vanilla code as "moderate", however assume for a moment that it was a more extreme difference. Every civic and thus group of civics has a range of gold costs for its upkeep levels that could be considered appropriate for the effects those civics have on your empire. Say you balance a far-right column first, giving it approximately-right gold costs to make players consider whether to use them or not depending on their situation. Now, if you do that, with the more extreme variance the far-left columns would instantly become prohibitively-expensive, and no one would ever use them even if they wanted to b/c they just couldn't afford it. Alternatively, say you balance a far-left category first... now the far-right ones are nearly free to use, everyone will use them automatically in every game, and people may even rush to them in the tech tree just so they can get their free empire boosts sooner. The effect isn't this drastic with vanilla's moderate variation in cost but it IS the same effect, which is bad. QED.

Roland Johansen · May 17, 2010

LunarMongoose said:
I found that mess a number of months ago (with much the same reaction lol), and if you analyze their code you can eventually figure out what it's doing... Basically it makes it so the further left in the Civics Advisor a civic column is, the more expensive all civics in that column are relative to other further-right categories. So that Government civics, even the Low Upkeep ones, probably cost more than High Upkeep Religion ones, for example. That's the main effect, it may be doing something else too I forget. But yeah my little component just took all that out basically, and made X Upkeep cost the same regardless of which civic type it was in, b/c playing with it the vanilla way was driving me nuts. It also made balancing the vanilla civics, and my mod's additional civics, a big pain because I would've had to balance their effects against gold costs that varied over a much wider range than the differences in the relative power of the civics do.

Edit: Lemme add another example cuz I wanna be extremely clear about this. I would describe the actual cost variance across columns with the vanilla code as "moderate", however assume for a moment that it was a more extreme difference. Every civic and thus group of civics has a range of gold costs for its upkeep levels that could be considered appropriate for the effects those civics have on your empire. Say you balance a far-right column first, giving it approximately-right gold costs to make players consider whether to use them or not depending on their situation. Now, if you do that, with the more extreme variance the far-left columns would instantly become prohibitively-expensive, and no one would ever use them even if they wanted to b/c they just couldn't afford it. Alternatively, say you balance a far-left category first... now the far-right ones are nearly free to use, everyone will use them automatically in every game, and people may even rush to them in the tech tree just so they can get their free empire boosts sooner. The effect isn't this drastic with vanilla's moderate variation in cost but it IS the same effect, which is bad. QED.

I wrote an article on civic upkeep costs a long time ago (see War Academy, Civic Upkeep Explained) when civ4 was on the market just a short while and the SDK wasn't available yet, so the formulas were found by experimentation in the game.

It could be that the formulas have some small errors in them since they aren't derived from the code, but there is a reason that there are small differences between the various cost formulas in the various civic categories. If each of the formulas would be exactly the same, then costs could go up in each category at the same time once one of your cities grows a single population point or you create a single extra city. This would thus result in big jumps in costs as your empire grows followed by long periods of stable costs. That's not something you'd want as a game designer. It's not a good thing when an empire of a single city of size 4 has 0 civic upkeep and an empire of a single city of size 5 has 5 civic upkeep (using the exact same civics) because each civic category increases in cost 1 at the same time (values are just a theoretical example to explain the idea).

More elegant formulas could have been designed by relying less on integer math (using a single decimal in the cost per civic category would have been enough to avoid this problem).

LunarMongoose · May 17, 2010

Very interesting. I just assumed it was based on the fact that Goverment is the left-most category, and that government would be the most expensive institution an empire could run, seeing as how it's, well... government lol. Guess not.

Your point is well taken Roland, however I'd still much prefer that one annoyance over the problems I described with the vanilla system. And it really is just an annoyance, I mean, I experience big jumps or falls in my empire income on a turn-by-turn basis all the time for a variety of reasons: a thunderstorm appeared and made my Pearl Nets temporarily unworkable, a huge major city just got blockaded by an enemy ship, a trade deal giving or receiving a lot of gold-per-turn just got established or terminated, etc... You have to deal with those spikes all the time, and it's just a matter of adjusting the empire spending levels a notch or two, so having civics suddenly add a bit of cost when a city somewhere grows doesn't really bother me. But yes, it's still inelegant. Probably pretty easy to fix, even with integer math, when I have time.

Ninja2 · May 18, 2010

Fuyu said:
The obvious merge is already done:

Nice work, but I can't really work off of that.

NBAfan · May 18, 2010

Thank you for making Better AI. Jdog! :king:

I can't play without it! :clap:

You deserve to be able to play with your project.

Version 1.00 discussion thread

Revolutionary

King

Great Engineer

Emperor

Warlord

Emperor

Emperor

Emperor

King

King

Emperor

Emperor

Revolutionary

Apolyton Sage

Deity

King

Deity

King

Great Engineer

boss

Similar threads