K-Mod: Far Beyond the Sword

[Tigranes]

Other than that, if you so strongly feel like K-Mod doesn't bring you any benefits... why come here with post after argumentative post instead of just not playing the mod? I don't see the need to defend my/our opinion that the AI is better. If you don't think that... fine. Up to you.

Why come to forums at all? To share opinions, to inform each other, to explain each other's points. I never hang my words in the thin air -- if I say something, I support it with an example. But you say things in general, without "before" and "after" comparisons. This thread is not titled -- "General Praise for K-mod". Imagine if I tell you in your own way -- if you don't like what I post -- don't read my posts. I mean, come on. We are all grown ups, I hope. You have a point, I have a point. Nothing is wrong with that, unless we make baseless claims. We both want a better AI, I presume.

It is not an easy task. For the original game a big team of artists and graphic designers works on the looks, but AI is written by one-two programmers. Difficulty levels just give AI more free cookies to make it look stronger. And yet people somehow designed a chess machine that can beat a World Champion! So no wonder all us dedicated Civ players are hungry for a better AI. I play civ since Civ I and gradually had to shift to Multiplayer mode to avoid constant disappointments.

People team up for projects like better AI for years, and things still do not work out the way they should. Now after all those years of existence of BTS a new mod claims to be a substantially better AI. I welcome that. All I am saying is -- show me how is it better? I have seen prosperous Empires before, old AI is pretty good at spamming settlers and workers, improving land, beating you on Wonders and Techs with proper discounts. But humans always outwit them in Espionage and War tactics, get themselves those Wonders, and winning one way or another in a pretty straightforward way. AI builds units with discounted prices but how it ends up using them? That's were one expects a better AI. Does K-mod deliver it? I say no -- and prove my way with examples. You say yes -- and period.

 

[Tigranes]

Ok. In Kmod the AI will sometimes DoW on me and invade with a stack of archers, just to pillage my tiles and fortify on a hill - essentially choking me as a human player would in MP. I've never seen the unmodded AI do that. As others have said, the major differences with Kmod, though, aren't tactical, they are strategic. Kmod AI will defend itself better because it will have more military units. It also deals with war much better. In Kmod, if you DoW and march towards the capital that is defended by 4 units and it takes you 6 turns to get there, by the time you get there the capital has 16 units, from whipping and reinforcements. Again, unmodded AI doesn't handle surprise attacks very well. Another example - the AI in Kmod expands very aggressively. I've seen Zara forward build his cities right up against my capital and then whip in city walls. I'm not sure if this was Karadoc's doing or if it's from BBAI but unmodded AI wasn't that cut-throat.

Basically, the AI in Kmod isn't as much of a pushover. In the unmodded game Monty and Shaka would rush me with a million swordsmen, which would kill me if I didn't prepare for it. If I did prepare for it I would just end up getting a bunch of level 3 axemen and a GG out of it. Ghandi would get 10 techs ahead of me but still only have 4 elephants defending his cities and I'd wipe him out, etc. In Kmod no one will suicide their entire army on you longbow/wall/hill city. Everyone will have a large standing army and everyone techs much more quickly as well.

Thank you. At least now I see some specific points. What level are you playing, if I may ask? If you play on Nobel and somehow AI gets ahead in techs because of smart choices, and not accumulated discounts this would be quite impressive.

From what I see so far some capitals start on K-mod map with buffed up food resources. Like Monty's capital had 3 Corns and 1 Clam in BFC.

 

[Habitus]

m what I see so far some capitals start on K-mod map with buffed up food resources. Like Monty's capital had 3 Corns and 1 Clam in BFC.

That happens due to map scripts, K-Mod as far as i know has only changed how the program evaluates starting locations by stopping 3 resource tiles such as plain cows and plain hill sheep counting as food resources (possible making them .5 food resources like flood plains).

The AI uses Espionage points alot better than BTS AI, since the only thing it does then is poison wells, now if its actually good or not I'm not totally sure but 'better'. War tactics are 90% of what K-Mod does, and what most the patches have tried to improve. While noble still isn't a challenge for most people, they generally have to play at 1 or 2 levels lower than BTS due to the better warring.

 

[karadoc]

From what I see so far some capitals start on K-mod map with buffed up food resources. Like Monty's capital had 3 Corns and 1 Clam in BFC.

That's not due to K-Mod changes (unless there is a bug). Habitus is right that K-Mod has slightly different rules for the minimum food of starting positions, but the example you gave (ie. 3 corn + 1 clam) is well above the minimum. In that example, the city would still be above the minimum if one of the food resources was removed - and so we can be certain that food was not added to reach the minimum requirements.

3 corn + 1 clam is possible in the unmodded game as well. The changes K-Mod makes to resource placement are very minor. Aside from the slightly modified minimum food, I can't think of anything which would make more resources appear more than in the standard rules. And even the minimum food rule will only make a difference in some fringe cases.

 

[BandC]

In case no one else noticed, new version!

 

[noto2]

Karadoc I might be imagining things but I've noticed lately that I get a lot of starts where the settler is not beside fresh water. I seem to remember starting beside fresh water almost all of the time in the past, has something changed?

 

[sjodster]

I notice in CvPlayerAI::AI_civicValue on line 13310 that you infer whether the method call is for adding or removing the civic:

int iS = isCivic(eCivic)?-1 :1;// K-Mod, sign for whether we should be considering gaining a bonus, or losing a bonus

The variable is then used in a number of places to invert the value of some of the civic's effects.

However, AI_civicValue is typically called to compare the existing civic value with an alternative. Consider CvPlayerAI::AI_bestCivic near line 13261:

if (canDoCivics((CivicTypes)iI))
{
	int iValue = AI_civicValue((CivicTypes)iI);
	if (iValue > iBestValue)
	{
		iBestValue = iValue;
		eBestCivic = (CivicTypes)iI;
	}
}

The problem is when this code asks for the value of the player's current civic, it will have an altered/negated value returned for it. It won't be comparing apples to apples. I believe that the AI's current civic selections will always be undervalued as a result of this, and in a worst case this could lead to constant civic toggling.

It's possible I'm misunderstanding something here, and if so please correct me.

Thanks!

-Josh

 

[Habitus]

Not looked at the code for along time but I recall that karadoc made it so that the new civic had the value of the old civic removed from it so the AI realised it was losing the old civic bonuses. This allows it to better determine if the new civic is really better (before this it only saw what it gained not what it lost).

*edit*

So I mean is that what your seeing in this code, and thinking it makes the old civic value lower rather than taking it from the new civic.

 

[damerell]

Here is another example. K-Mod hint states that occasionally rookies will defend experienced units during the stack attack, so that best attacker will not always meet best defender, which makes sense. However when I was attacking unit by unit it happened that K-mod picked 0.9 strength enemy Longbow defender to face my full force Elephant, while they had better defending units (and yet still inferior to my attacker). Now this a very artificial and unfair approach, feels like a new AI cheat.

Or it could be bad luck. Isn't this just "I lost a 95% attack - the AI must be cheating" in a different hat?

 

[Fluffy_Rabbit]

@Karadoc still greatly appreciating your mod and the continuing work.

3 separate points

Point 1: AI moving huge stack out of conquered city allowing recapture

Played a game recently as Greece on my Earth18 map. Russians attacked with an absolutely huge stack and took Byzantion with 65+ surviving rifles and other units (Cossacks). Your "crush strategy" in action and omg it is difficult to deal with.

However . . . I had nationalism and infantry and was drafting enough units to contain the threat. I had a small stack of about 10 infantry outside Byzantion on the hill and quite surprisingly the AI moved most of its army out of the city heading E into Anatolia (tautology) leaving the standard garrison of 4 rifles.

Now this is not a good idea with 10 infantry at the walls

Result was: I retook Byzantion, got some more units in there (machine guns) and next turn Russia used its 65+ rifles to take it back - fine, I lost the city immediately, however you can imagine the kill-rate of rifles v infantry and machine guns - 4:1 or 5:1, something like that.

So the Russians lost most of their 65+ rifles when, had they simply fortified in Byzantion, would have taken me ages to beat them down.

Turned out Russia gave me the city back for peace, though at this stage I was building artillery, still drafting infantry as much as possible, and sure I was going to take the city back.

Now perhaps this is too specific an example, your crush strategy is extremely strong - beats me why AI would move offensive army when superior army is at the walls (well, superior after huge garrison vacates).

Point 2: vulnerable naval units in port/fort/city

Same game, similar point to "caravels in fort" point made above - anyway England was on course to win space. I invaded with huge stack. Should be perfectly clear Aberdeen cannot defend (have already taken Plymouth and London) - and yet a huge amount of naval hardware is left in port.

So it is really the same point about leaving naval units in indefensible positions - a human would gtfoh and save the units. Screenshots attached.

Point 3: idle workers

Same game, England had settled in Brazil. Called city York though in reality Iguatu, Pernambuco, w/e. Anway, because of space-race I had spies everywhere, including here. For moves on end the workers did nothing. Is it perhaps because of barb borders? England not at war with anyone at this time. Terrain in question is absolutely fabulous when developed - to have 4 or 5 idle workers here makes no sense.

I could not see why workers were doing nothing, particularly given horrible unimproved terrain.

Different point, still loving the espionage in K-Mod. Way it works out for me is that I have 2-3 spies fortified in every city and as many spies as possible in the marginal cities of foreign civs (one should put spies in the 1 pop recent conquests of rival civs for maximum effect). This way I can stop a lot of espionage and do the most damage with my EP. Just an observation

 

[Fluffy_Rabbit]

Wish to add a few points:

have downloaded and am testing v 1.39, have read the changelog
gifting great persons? surely nobody, human or AI, would ever do that? not disputing your modifiers, just wondering when/why that would ever happen ...

Forts: forts are immensely important imo, and yet when an enemy civ takes a fort (in enemy culture) they always destroy it. Since once cannot build forts in enemy culture, the capture of a fort in enemy territory is a huge result - should be fortified, not destroyed. Also I do not recall AI building strategic forts *ever*, never seen an AI fort except for resources out of BFC - can this be fixed? I mean it would be really awesome to encounter an AI with forts along its border ...

To deal with your [@karadoc] crush strategy, I tend to build and seriously fortify the normal choke points and forested hills, even at the expense of growth/production. This certainly pays off on most maps - however for the same reasons AI should not destroy - very important tiles.

On a positive point, love your modifications re giving of units, hope to see how it plays out in a real game.

And yes (hopefully) Caesar's greeting is now correct (ht Yudishtira )

 

[noto2]

Diplomacy is very important. Late game a random Gprophet or Gartist spawn might not be worth all that much, so gifting for +1, +2 in diplomacy might be worthwhile, if it gets a civ to friendly, for example, and opens up tech trades, or perhaps you do it to bump someone from annoyed to cautious for open borders, etc

 

[Charles555nc]

version 1.39 is up

I like that Karadoc is giving diplomatic victories, and just influence in general, more options.

And improving first strike promotion is just icing on the cake

 

[damerell]

version 1.39 is up

Github doesn't seem to have noticed yet, but so it is.

Incidentally, I see we're told some bits of BUG have been removed. Apologies if this is a hardy perennial, but... which, please?

 

[noto2]

Yeah the buff to Drill 1 was so badly needed, I'm happy about the change

 

[SyntaxError]

In Sevopedia, is it me or is it when you click on the "Traits" category on the very left side, it returns an empty list on the next column?

 

[satrapper]

In Sevopedia, is it me or is it when you click on the "Traits" category on the very left side, it returns an empty list on the next column?

yep you're right

I mean it would be really awesome to encounter an AI with forts along its border

pretty sure that i've seen the ai do this on occasion. not neccessarily at choke points but i guess it's hard to get the ai to recognise them as such.

Incidentally, I see we're told some bits of BUG have been removed.

where are we told that? github? the changelog issued with 1.39 says nothing about that.

 

[robbyextreme]

How do you make it so the amount of hammers show up when you hover over the whip button? I have to always check the log for overflow amount. And half the time it doesn't even display it in there either.

 

[Charles555nc]

I REALLY want to give thebes to Babylon in this save that I have here. Is any way we can adjust the city gifting rules so I can give it to her?

 

[damerell]

where are we told that? github? the changelog issued with 1.39 says nothing about that.

At http://forums.civfanatics.com/downloads.php?do=file&id=16331

"BUG (heavily modified) [Some new features added, such as Rapid Unit Cycling; and some features removed.]"