WANTED: feedback on new AI tweaks in April 28th patch

[unmerged(6777)]

The title says it all. We need to hear as much feedback as possible about the tweaks that Johan has made to the AI in this week's patch.

• AI learned to mobilize more of its realms troops at need.
• AI is now more likely to honor mobilization requests, especially from other ai rulers.
• Improved some aspects of the military AI for it to fight some battles more focused.

So does it seem to be doing what Johan has been coded to do? Is whatever it's doing actually helping it behave somewhat better in prosecuting wars -- both AI vs AI as well as AI vs human? Does it seem to be handling vassal troops somewhat better than before? Etc.

Your feedback is both greatly appreciated AND will have an impact on future AI development/tweaks (or lack thereof).

Cheers!

C.

 

[Duuk]

Posted this in the other thread.

AI mobilizes like mad. Though it does tend to "spam request" if its been declined, so this could be an issue.

AI doesn't seem much better at actually using the troops though. BYZ emperor mobilized my troops (after I turned him down 7 times) and then they just sat in my province for over a year doing not much.

 

[Duuk]

Having played quite a bit now...

The AI needs to decide when to give up on mobilizing otherwise it will cause itself civil wars.

The AI needs to better be able to handle "liege to liege" wars. If the AI has a target province, it can nail it. If it's fighting "the Seljuks" it gets all confused as to where to attack. Have the AI try to pick off vassals.

I like the new events though. Big time.

 

[Grosshaus]

My vassals kept on revolting (my king being excommunicated and for his last years also heretic might have had something to do with that) and every time they mobilized vassals and used all their troops relatively wisely. Not that they had any chances but still.

I only fought a single major war against Croatia, I was Hungary myself. I ended up crushing them quite fast although they did mobilize quite a lot of troops. They used the troops piecemeal after a loss in an original major (6000+ troops on both sides) battle against my invading army, it was too easy to pick <1000 armies with a couple of 2000+ armies.

 

[Spruce]

I fought a long exhaustive war with the Seljuks - they were quite a challenge. Same goes for battles with king of Burgundy,

pf, the AI brings in enough troops - that's for sure.

In one case I invaded an enclave with 16.000 troops - as I was sure to have it - I was surprise the Burgundian king had massed 30.000 troops to hold me off...

just thinking = can we change the enemy army counter to =

below 1000 = small,
1000-5000 = medium,
5000-10000 = large,
10000-20000 = huge,

that would make the war even more fun to play,

 

[Duuk]

I fought a long exhaustive war with the Seljuks - they were quite a challenge. Same goes for battles with king of Burgundy,

pf, the AI brings in enough troops - that's for sure.

In one case I invaded an enclave with 16.000 troops - as I was sure to have it - I was surprise the Burgundian king had massed 30.000 troops to hold me off...

just thinking = can we change the enemy army counter to =

below 1000 = small,
1000-5000 = medium,
5000-10000 = large,
10000-20000 = huge,

that would make the war even more fun to play,

Or use slightly different graphics for them possibly? Like below 1000 is the old "peasant with a pitchfork" from EU2 fame, 1000-5000 is the current icon, 5000-10000 is a mounted armoured knight, 10000+ is infantry with a siege tower?

Something so you can look and say, 'Uh oh. Big scary army coming'.

 

[Spruce]

Or use slightly different graphics for them possibly? Like below 1000 is the old "peasant with a pitchfork" from EU2 fame, 1000-5000 is the current icon, 5000-10000 is a mounted armoured knight, 10000+ is infantry with a siege tower?

Something so you can look and say, 'Uh oh. Big scary army coming'.

yeah - exactly,

knowing the exact size is just so louzy ! Often I just check the martial rating of the enemy commander and then mobilise enough troops to match the numbers and march them off -

In my last game it was even more worse - Burgundy king had 5 ! armies in the South of Italy, and I just picked them all one by one to have the biggest number advantage...

 

[unmerged(21937)]

While having sprites vary like that is definetly CK2 stuff, having troop strengths obfuscated like that is somewhat more reasonable idea.

 

[past caring]

Hmmm - Castile normal/normal....

It would appear that the AI tweaks are quite effective.

Mobilised - gathered all regiments in Soria.
DoWed Molina. Molina has no allies.
I'm DoWed by Zaragoza. Zaragoza has no allies.
I take Molina and move on Zaragoza.
I take Zaragoza.
I'm DoWed by Toledo and Sevilla (allies)
I get my arse kicked, losing Molina, Soria, Valladolid and Burgos in short order. Seems like the marshall recruitment bonus has been tweaked down again - can't recruit quickly enough. AI mobilises well - and all regiments target same province, including allies.

Succesive and repeated attempts to sue for peace get me nowhere - offers of 300+ to Toledo and Sevilla are rejected.

Capital is now Zaragoza.
Toledo takes Zargoza.
My capital doesn't move to either of my surviving vassals (Viscaya or Asturias)
Game over by February 1070.



I didn't do that badly in my very first game of CK.......

I'll give it another go tomorrow.

 

[unmerged(6777)]

Seems like the marshall recruitment bonus has been tweaked down again - can't recruit quickly enough.

AFAIK it hasn't. Could you have lost your marshal somewhere in there (maybe set up shop in one of your conquests, or died via event and wasn't replaced)?

At any rate, in my tests with the new exe the AI seems to have perked up quite a bit when on the offensive but is still a bit lackluster on the defensive until you set foot in its territory. When I attacked England (as France) I was using the nofog cheat and was somewhat surprised at the lack of response. When I DOWed, England was suddenly awash with regiments but they just stood there doing nothing. The second I landed on English soil they all began to converge on me in very short order. I ended up in a desperate struggle that was ruinously expensive and it took quite some time before I finally beat them all down, and I eventually managed to obtain the peace I was after...but the cost was excruciating. (Which is exactly as it should be...and I would have had an even tougher time without the cheat.)

What would improve the English AI in that situation -- and this would apply to almost any potential target that can raise large numbers of regiments -- would be for the initial mobilization to then seek to assemble a series of as many 6000-10000 man armies as possible, distributed roughly evenly through the demesne. Once the invading army is seen, these armies can begin to converge which would have been more than enough to dislodge me from my attack. The reason I prevailed is that I had a couple of really nice large armies and I was only forced to fight a very long succession of uneven-odds battles. If England had brought the same number of forces against me in a less piecemeal fashion I would have been waiving the white flag.

Interestingly, it was a few years later England initiated a tussle with Scotland. In stark contrast to its defensive posture against me, this time it assembled a decent army and made some serious inroads into enemy territory. It still had a lot of regiments standing aroudn doing nothing for quite a while, but it seemed to be a much more coherent attack than I've seen in previous games.

Another note: it seems to take a very long time for the AI to release and/or disband regiments once the war is over. Its own forces were stood dodwn fairly promptly, but it was several years (!) before all of the vassal regiments had finally disappeared form the map. I was checking all the time and there were no wars in which either the vassal or their liege were involved, so this seems like a possible "vassal control release" issue to me.

Once other thing...Germany got involved in a war (started by its ally) with a tribe that was waaaaaaaaaaaaaaay over on the east edge of the map. I guess they couldn't reach the target for some reason, but they remained at war for about 8 years with all of their regiments activated but motionless. Good thing that the AI doesn't have to worry about cash or the German king would have had a debt in the range of about 500,000 gold by the end of it. If the AI determines that there is no path to the enemy then it should be very willing to make a white peace or, at very least, should stand down its forces.

 

[Duuk]

I just lost a war to the AI! On *NORMAL*.

I'm... shocked. And I'm in love.

Though I agree they don't seem to swift on defense. The Byzantines seem to be losing badly to the Seljuks and it appears that's because the Turks have the initiative and the Romans are just sitting around losing provinces.

 

[unmerged(2456)]

I probably wouldn't be so unwilling to no give up my troops to my liege every time he spammed if i knew he wasn't going let them sit their idly, while I, the court, am forced to pay for it.

That's the crux of the matter. Even though the troops are under the liege's command, 100% of the cost is burdened by the count, something that is unhistorical. Generally speaking the first 90 days were payed by the count and afterwords the liege took over responsibility. Since this is likely nigh impossible, perhaps it should be a percentage paid by both parties 90/10 or 80/20. The king, who makes more money and is actually using the troops should have to pay more.

 

[hjarg]

Have to agree with Jinnai here. As Milan, my beloved king called to my service. Well, about 7000 soldiers aren't cheap... and before i noticed, my budget went negative, half of my improvements got sold, prestige all time low, and my armies fighting god-knows-where.

At least let me keep track of my units.

Oh, and loved the new events!!!

 

[past caring]

Could you have lost your marshal somewhere in there (maybe set up shop in one of your conquests, or died via event and wasn't replaced)?

Definitely not. The one thing I always pay particular attention to is making sure that all my sieges are led by my King or an anonymous commander......old El Cid was still fighting gamely on right at the end.

Anyway, on my second game now. More later.....

 

[Spruce]

in my game I was a bit cheating due to some old save file - where it seems I've started a war when I was in a county (so you get -5 bb) - and getting me "gold cheats".

I had enough money and I was intending to dow the Seljuks and take their Croatian Lands - which had become Muslim...

Well we waged war for about 2 years !!! With about the loss on both sides (that Turk had good martial stats !) of about 40.000 soldiers (really it was a slaughterhause)... !

It seems the Seljuks had plenty of money to keep their war effort on - they didn't went bankrupt - event when they had to field their 20.000 men army.

Also is there any means we can give rulers a malus (event) that are overrecruiting their demesne or vassals.

In one game I had so many people killed (around 200.000 men died in total, this is outrageous), that I have the impression you can keep on recruiting without penalty. The current events do hurt - but only financially - so basicly the penalty (rebellion) is hard for "low-cash" rulers and too soft on "high-cash" rulers.

ps = another reason to change code and let rebels from different faith go independent after they take castle in a province,

 

[Winkelried]

in my game I was a bit cheating due to some old save file - where it seems I've started a war when I was in a county (so you get -5 bb) - and getting me "gold cheats".

I had enough money and I was intending to dow the Seljuks and take their Croatian Lands - which had become Muslim...

Well we waged war for about 2 years !!! With about the loss on both sides (that Turk had good martial stats !) of about 40.000 soldiers (really it was a slaughterhause)... !

It seems the Seljuks had plenty of money to keep their war effort on - they didn't went bankrupt - event when they had to field their 20.000 men army.

Also is there any means we can give rulers a malus (event) that are overrecruiting their demesne or vassals.

In one game I had so many people killed (around 200.000 men died in total, this is outrageous), that I have the impression you can keep on recruiting without penalty. The current events do hurt - but only financially - so basicly the penalty (rebellion) is hard for "low-cash" rulers and too soft on "high-cash" rulers.

ps = another reason to change code and let rebels from different faith go independent after they take castle in a province,

AFAIK the AI is never bankrupt. No matter how much money he spends, his account is always at least zero. That's a pity because outlasting my enemies financially/resourcewise is my preferred strategy.

 

[Spruce]

AFAIK the AI is never bankrupt. No matter how much money he spends, his account is always at least zero. That's a pity because outlasting my enemies financially/resourcewise is my preferred strategy.

Yes, I'm not happy about this one too

- it shouldn't be that difficult to teach the AI to hoard some money before going to war - or accept peace after some time when money goes negative.

This is one of the basic concepts of CK - the balance between stewardship - martial rating - and money reserves in case of big armies,

that's a pitty - I didn't notice until now

 

[Ben Music]

Hmmm - Castile normal/normal....

It would appear that the AI tweaks are quite effective.

Mobilised - gathered all regiments in Soria.
DoWed Molina. Molina has no allies.
I'm DoWed by Zaragoza. Zaragoza has no allies.
I take Molina and move on Zaragoza.
I take Zaragoza.
I'm DoWed by Toledo and Sevilla (allies)
I get my arse kicked, losing Molina, Soria, Valladolid and Burgos in short order. Seems like the marshall recruitment bonus has been tweaked down again - can't recruit quickly enough. AI mobilises well - and all regiments target same province, including allies.

Succesive and repeated attempts to sue for peace get me nowhere - offers of 300+ to Toledo and Sevilla are rejected.

Capital is now Zaragoza.
Toledo takes Zargoza.
My capital doesn't move to either of my surviving vassals (Viscaya or Asturias)
Game over by February 1070.



I didn't do that badly in my very first game of CK.......

I'll give it another go tomorrow.

Exact same thing happened to me, but I started with Zaragoza.

 

[unmerged(2456)]

Yes, I'm not happy about this one too

- it shouldn't be that difficult to teach the AI to hoard some money before going to war - or accept peace after some time when money goes negative.

This is one of the basic concepts of CK - the balance between stewardship - martial rating - and money reserves in case of big armies,

that's a pitty - I didn't notice until now

I'm not sure you could teach it to horde money without a strictly hard cap. The same with going negative. But for the latter i'm not sure you could then say how willing it would be to end peace.

Then for going to war you haveto teach it when the appropiate time to spend money on improvents and when is not. That is by far the most difficult because there are times it is worth delaying a war for a new improvement.

 

[Spruce]

I'm not sure you could teach it to horde money without a strictly hard cap. The same with going negative. But for the latter i'm not sure you could then say how willing it would be to end peace.

Then for going to war you haveto teach it when the appropiate time to spend money on improvents and when is not. That is by far the most difficult because there are times it is worth delaying a war for a new improvement.

yes, but war efforts where so costly and it wasn't "total war" - so the endeavours where limited by either money or resources (food),

so an AI with no money - and high martial ratings - able to relentlessly assaulting my realms with massive armies for about 2 years - I can't accept that,

that's a basic rule and it's quite well modeled in CK - but it should also apply to the AI = no money = no war or short war or limited period