On Cheftain IIRC they have a big cut in culture points needed for social policies. About one third. AI also has a 60% bonus against barbs, and even with it they perform quite bad.
Well, really wanted to say TRUE PRINCE stating that even if on Prince they have no extra production/maintenance/gold bonuses, they play on Cheftain.
As developer I have a good feeling of what can be done and what cannot. There are lots of shortcuts to get the desired effects without going complicated reflexive processes. After all, the AI can manage much more accurate information and than the player, as they can effectively cheat and take into account things that are not visible to the player.
I don't expect a clever AI, but as I said before there are some workarounds to make the AI feel smarter. Anyways I feel many aspects could be improved. There are very stupid behaviours that could be avoided, and they are there for long time.
Yeah, but because they get ridiculous bonuses, they need a lot of help to perform decent. I usually play on Immortal BTW. I always wanted to make a reverse difficulty mod, It shouldn't be hard. Starting with Deity techs, units, and 50% production cut while AI has prince settings, that should be funny to try.
Yeah, that's a great idea, a hidden UA only for AI to help them perform better with their strong points, also making AI to carefully choose wonders instead of trying to build them all at some point, would make them perform better, much like players that focus on basic infrastructure, that on the long run usually provide better performance than building one wonder after another.
BTW when looking at flavors on the game DB tables, they also appear in buildings, and were a lot of different flavors, like growth, I don't know if they modify flavours as you build those, or the flavours are there to calculate the AI build priorities, or both.
Many thanks for the insight on the AI behaviour, BTW.
Hmm. Well one of us is making a mistake of interpretation or in understanding. Everything I have found shows that the AI only gets the happiness benefit, the barb bonus and some minor gold bonus'. There is no culture bonus ????? The XML has no handicap for culture that I am aware of, and from perusing the C++ I can find no indication of an AI culture bonus. Is this something you heard, or is there something in the code that I missed ?
Code:
<Type>HANDICAP_PRINCE</Type><Description>TXT_KEY_HANDICAP_PRINCE</Description><Help>TXT_KEY_HANDICAP_PRINCE_HELP</Help><StartingLocPercent>50</StartingLocPercent><AdvancedStartPointsMod>100</AdvancedStartPointsMod><HappinessDefault>9</HappinessDefault><NumCitiesUnhappinessMod>100</NumCitiesUnhappinessMod><PopulationUnhappinessMod>100</PopulationUnhappinessMod><Gold>0</Gold><GoldFreeUnits>0</GoldFreeUnits><ProductionFreeUnits>5</ProductionFreeUnits><ProductionFreeUnitsPopulationPercent>50</ProductionFreeUnitsPopulationPercent><ProductionFreeUnitsPerCity>2</ProductionFreeUnitsPerCity><RouteCostPercent>100</RouteCostPercent><UnitCostPercent>100</UnitCostPercent><BuildingCostPercent>100</BuildingCostPercent><ResearchPercent>100</ResearchPercent><PolicyPercent>100</PolicyPercent><ImprovementCostPercent>100</ImprovementCostPercent><CityProductionNumOptionsConsidered>2</CityProductionNumOptionsConsidered><TechNumOptionsConsidered>2</TechNumOptionsConsidered><PolicyNumOptionsConsidered>2</PolicyNumOptionsConsidered>
<!--<InflationPercent>90</InflationPercent>-->
<AttitudeChange>-1</AttitudeChange><NoTechTradeModifier>70</NoTechTradeModifier><BarbCampGold>25</BarbCampGold><BarbSpawnMod>0</BarbSpawnMod><BarbarianBonus>33</BarbarianBonus><AIBarbarianBonus>60</AIBarbarianBonus><EarliestBarbarianReleaseTurn>0</EarliestBarbarianReleaseTurn><BarbarianLandTargetRange>5</BarbarianLandTargetRange><BarbarianSeaTargetRange>10</BarbarianSeaTargetRange><AIDeclareWarProb>100</AIDeclareWarProb><AIUnhappinessPercent>100</AIUnhappinessPercent><AIGrowthPercent>100</AIGrowthPercent><AITrainPercent>100</AITrainPercent><AIWorldTrainPercent>100</AIWorldTrainPercent><AIConstructPercent>100</AIConstructPercent><AIWorldConstructPercent>100</AIWorldConstructPercent><AICreatePercent>100</AICreatePercent><AIWorldCreatePercent>100</AIWorldCreatePercent><AIBuildingCostPercent>100</AIBuildingCostPercent><AIUnitCostPercent>85</AIUnitCostPercent><AIUnitSupplyPercent>20</AIUnitSupplyPercent><AIUnitUpgradePercent>50</AIUnitUpgradePercent>
<!--<AIInflationPercent>80</AIInflationPercent>-->
<AIAdvancedStartPercent>100</AIAdvancedStartPercent><IconAtlas>DIFFICULTY_ATLAS</IconAtlas><PortraitIndex>3</PortraitIndex></Row>-<Row>
I guess I am still not following your TRUE PRINCE line of thought ? Hey look this could just be me, its late and I am tired and not too bright to begin with
But there are minimal gold bonus' on PRINCE just some for unit upgrades and unit construction, which it needs as the tactical side it really struggles on. The rest of it is pretty much all the same, check for yourself.
In regards to the AI I guess it depends on what you mean. There are very firm limits on what you can do with todays spec computers (that the average gamer has anyway), particularly when you start talking about 1UPT and tactical AI. For comparison fire up a decent chess programs and see how long it takes between turns on a decent level. Now any truly decent tactical AI would need to devote at least as much a chess program does, considering their are far more variables involved.
When it comes to the non-tactical side (which is where I would more agree with you) you do run into a few more problems :
1. The type of AI you use. Civ uses a flavour base AI where each choice is decided based on the priority of all possible choices. This makes for something that is analogous (given a very long bow) to a static neural net (ie one that cannot alter its self). The strength of this is its ability to act differently each time you play, whilst still being gently nudged along certain desirable paths. It also allows the AI to adapt on the fly more easily, anything that is statically coded becomes very prone to abuse (ie the AI will always do X so we do Y). Now sure Civ V has quite a few deficiencies like this but they are generally the result of static AI programming rather then the flavour based sections I was talking about. The weakness of such a system is it is harder to program and more prone to problems of emergence.
2. Processing Limits. I realise you seem to disagree with me here, but I am unsure why. The reason BNW turn times have increased is because the Ai has more to consider. When you look through the code (and as a developer I would suggest you download it from steam as it is very interesting IMO) you will see the shear amount of code that is being processed by the AI at any given time. The codefiles are enormous, and they are not (in general) bloated by overheads of the engine or language. I would hate to think of the amount of man hours that went into just the Diplomacy codefile its a full MB in size (over 30,000 lines) and that's not looking at the tactical AI, city AI, city state AI etc etc.
3. Allocation of developer resources. Well unfortunately AI does not get the most attention in most games. Its just not sexy enough, and is very hard to market. So if you spend much more on AI (without taking away from anything else) and make your game more expensive how do you mass market that to people ? You can show off the graphics, the sound and the gameplay fairly easily, but the AI would require playing it or hearing from someone who played it.
4. Emergent problems : With a complicated AI, that by its nature is not a simple if-then-else machine, problems of emergence do occur. The ORDER AI issues is likely the result of such a problem. Testing an AI is a time consuming <removed for civility>
Civ doesn't progress very fast even when hijacked through the firetuner, so problems of emergence are generally only found through BETA tests or after release.
You will be happy to know that there are plenty of areas in the game where the AI could effectively cheat (and just look up the info) but it does not (guessing another player GS is one such area). Some other areas it does cheat but for the exact reason you mentioned, its faster and easier to do and really doesn't matter that much.
Unfortunately the shear number of variables that have to be considered by an AI in a game like civ is just too much for a truly challenging AI without passive bonus'.
Now I do agree there is much room for improvement. Especially in the area of making sure AI's do not do things that are detrimental to them, or improving their ability to pursue each GS, etc. Absolutely agree here.
I would suggest adding in DIFICULTY FLAVOURS for the game (or perhaps even each AI) that alters the FLAVOUR values of the AI (and in game ones) based on difficulty. As far as I am aware Civ V does not have such a thing. That way you could favour a more ruthlessly completive playstyle on higher difficulties without upsetting the more casual (or immersive )players. It would make some difference, although how much is pure speculation.
I guess all I am saying is that game AI is an enormous project, confounded by a lack of resources (both man hours and clock cycles) and extremely difficult to advocate for from a marketing perspective. Unfortunately much of it gets thrown out every few years when CIV increments its release (ie CIV V to Civ VI). I really think they need to find a way to keep much of the AI between releases so that the AI can evolve with each new release rather then have to be re-written.
Anyway sorry for rambling on. If you want the reverse difficulty thing just change the XML values and you will get what you want (or close to it anyway), you don't even need to go into the code proper, so no compiling issues.
All the best, good talking with you.
FYI Yes the flavour values are used to increase the AI's priorities to build, train, go defensive, attack etc etc. Most of them are fairly self explanatory but some are not. BIBORS XML spreadhseet I think has an explanation for you that is pretty accurate for the Flavours. I do not believe the flavours are modified by what you build (but could be wrong) they get modified by all sorts of in game conditions for the AI though. For example change GS, and a whole slew of Flavours get changed. Be the Mayans and tech priority to get to THEOLOGY gets changed, if your city is under a certain size then flavours change, if the turn timer (or eras) progress FLAVOURS change etc etc.
