I won that game I mentioned. Bismark stalled a bit due to voluntary vassals breaking free and needing to be reabsorbed. He did manage to conquer Qin Shi Huang, and he also got most of his spaceship done. He was only missing one part. -- Anyway, with his stagnating war efforts, as I predicted, the race between me and Gandhi was very close. In fact, I launched my space just
one turn before he launched his. -- So it was a close game.
--
Here are some answers to recent questions:
1) How much did you reduce AI cheating beyond removing knowledge of tiles? Are details in the changelog? (haven't gone through all of it)
2) Which other AI cheats remain besides handicaps?
Here's a bit of background info: The AI is written in such a way that the programmer has to explicitly check whether or not each thing 'should' be known by the AI. For example, to check whether its city is under threat, the AI will check the map data directly to see what units are nearby, and
then it will check stuff like
"do I actually have vision on that plot; is that unit 'invisible'; if so, do I have something in the area which can see invisible units; ..." and so on. In this paradigm it's easy to 'accidentally' make the AI check. It isn't always obvious when or how the AI is cheating; it actually takes effort for the programmer to make the AI not cheat.
So with this in mind it's hard to say exactly what cheats remain, but I can say 'not many'; and (as far as I know) all of the cheats are related to information. ie. there are no cases where the AI gets special combat advantages, and no cases where they can do things which would be impossible for a human player to do. But there are still some cases where the AI uses information that a human wouldn't have direct access to.
In most cases, the AI's vision cheating has been removed. It no longer knows about cities it hasn't seen, and it can't calculate the size of a stack that it can't see (including city defences and so on). But there are still some cases where the AI knows it can attack something which it technically should not know about. This is most noticeable in naval warfare, because in many cases the AI will know it can attack you (and know the odds) without even being able to see you... I don't intend to remove that particular cheat, because without it, the current AI would be completely useless in naval warfare. It doesn't try to predict where units it has seen have moved, and so it would simply never find any of your ships unless you ended turn right next to them...
Aside from the vision cheating; K-Mod has removed a few minor & strange AI cheats - I think some of these cheats were actually created by the 'better AI' mod. For example, AI units on the coast use to be able to jump into adjacent boats without using any movement points. AI spies use to be able to move and then do a mission on the same turn (but in most cases they would choose not to do so anyway).
Some other cheats still remain, but that are pretty minor. The only ones which come to mind are that the AI players know how much of your world map they don't yet have. (They need this cheat to properly evaluate map trading. - In fact, I'm tempted to give similar information to the human players rather than remove the information from the AI.). And in some cases, for example when the AI is evaluating which building to destroy with espionage, the AI acts as though it has internal vision of the city. ie. when it is evaluating a building in an enemy city, it sometimes takes into account stuff like the city's health & happiness, and the state of the commerce slider and stuff like that even if it shouldn't have that information.
Actually, in some sense that espionage stuff is a cheat that I added myself. The original evaluation of espionage missions was so basic that it didn't make full use of that info... ... hmm... now I'm wondering if I should remove the AI's ability to evaluate buildings when it doesn't have vision of the city. It's tricky though, because without that evaluation they'd essentially have no idea whatsoever of what to sabotage, whereas a human player would be able to reliably guess that info anyway. (It would take a huge amount of programming to allow the AI to use such guesses rather than the true numbers in their evaluations - so I'm certainly not going to do that.)
Maybe add some quick info on religion changes on the download page for those checking it out?
I probably should do that. But I'm not really sure what I should write for it. The changes to the spread of religion only have a fairly subtle effect on the game; and most of the details are just technical stuff which probably wouldn't be of much interest to most people. -- But I suppose that if I'm saying it's a big new feature, I should at least say what has been changed...
3) Is there a mod or anything I could do myself to reduce the graphics demands below lowest ingame settings?
Not that I know of. The in-game settings show only single-unit graphics and to not animate the unit are probably helpful; but other than that, I don't know.
4) Have you thought about how less predictable technology progress could work? Some randomized prerequisites with the tech tree being revealed only gradually? Partly blind research similar to Alpha Centauri? Or is this of no interest to you?
I wouldn't say it's of no interest to me, but it is beyond the scope of this mod.
--
A question: I think I have mentioned this before but I am not sure if anyone answered. In several occasions, I saw an AI big stack attacking my city but instead of attacking first with siege units, it started attacking with some other unit(s) then switched to siege units. Isn't it ALWAYS better to start with attacking with siege units first if you are making a committed attack? Is this a small battle AI bug or am I missing something?
Well, it isn't
always better to start with the siege units. In some cases, the non-siege units might have good winning odds while the siege units have losing odds - and so it might be worth skipping the siege units so save them for a different city; or maybe just to let a non-siege unit take out the best defender to see if that gives good odds to the siege unit afterward...
In terms of the AI, two possible reasons for what you saw come to mind: The AI will attack with non-siege units if the odds are very good. (I don't remember the exact number off the top of my head.). Secondly, the AI might attack with non-siege units first if the units are in different groups and for reason they have decided to move the non-siege ones first. (This will be very rare, because units with collateral damage are generally given high priority for the AI to move first.)
What is this error when loading a save game:
"Cannot load version 303, expected 302 or lower."
If I load the mod first, then load the savegame, it's fine.
That's a problem which is difficult to fix. Here's what's happened:
A couple of times in long-past versions of K-Mod, I changed some things in the format of save games. Since older versions of K-Mod cannot load these newer saves without crashing, I decided I might as well make use of the "save version" feature which already exists in the game. I increased the save version from 302 to 303 to explicitly block the loading of the new save games with old versions of K-Mod. That worked fine, but unfortunately it turns out that the version check is done before the Mod check when loading the save – and so, as you have seen, the game will simply refuse to load K-Mod saves rather than suggesting that the mod should be loaded first. In my view, that's a bug in the game. It should check the mod first, and _then_ the save version. But that bug can't be fixed with a mod.
I could decrease the save version back to 302, but if I did that then all existing K-Mod saves would fail to load. If I ever change the save format again, that would be a good time to bring the save version back down. I might do that if I remember.
goodlistener said:
Noticed that I only start paying maintenance due to distance when I have 3 cities.
Is that different in the unmodded game?
I just got outplayed.
