Home Trending Latest threads New posts Developer diaries

Want a more determined Germany AI? - Try this!

Jump to latest Follow Reply
  • We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.
Show only dev responses
Luxor said:
What I mean is that in this way much probably whatever of good is in this "mod", won't be in 1.2
Click to expand...
The mod is also a lot about how Spocky and I think the AI should act. It may differ from the standard AI Paradox wants, and our interpretation is not meant as an offence, DAIM has its own weaknesses.
We'll see about the AI in 1.2, I'm confident that Lothos will come up with his own solutions to improve the AI.
 
I already went through the DAIM project. Many of the things they are doing I had already setup back in November in the beta before gold and we put the changes in the events to make the AI more accurate on his attacks. For example not launching offensives in the winter for Germany. What I did like in DAIM was the ignore techs which in 1.1 I did not have the time to put in as other things took precedence. Several things they did in DAIM to make Japan better was

1) Remove the passitivity on Japan
2) left the passitivity on China (which is fine and guarantees Japan will win every time) if this is removed as well Japan will loose in China 100% of the time if they are AI controlled.
3) Removed the month events from Germany AI events so he is aggressive like in HOI one and goes on an attack spree. This does have side effects as it increases US entry faster but thats how they wanted to design it they have that right as it is their mod.
4) They had Japan immediately load an invade Phillipenes AI file which is fine but has side effects cause if a human picks up on this and is playing the USA all he has to do is mass troops in the Phillipenes and the Japanese will go nowhere. The original setup had a flow for Japan to follow and if the USA did something like that he would move on to another target.

Granted I am not saying Japan was great in 1.1 but things are more complicated that what they appear and it takes 100s of hours of balance testing to find out what works. Many of these things have already been solved but I can not give anyone a list cause of the NDA. I appreciate what DAIM is doing as it keeps people playing HoI 2 with some interest but they should also be aware that the AI for Vanialla HoI 2 will continue to be developed for a very long time.
 
Lothos said:
*snip*
Several things they did in DAIM to make Japan better was

1) Remove the passitivity on Japan
2) left the passitivity on China (which is fine and guarantees Japan will win every time) if this is removed as well Japan will loose in China 100% of the time if they are AI controlled.
Click to expand...
Did you test your statement? I just did, I removed the passivity (including the "handicap") of Nationalist China in all files and ran two 1936 handsoffs with DAIM. The result: China annexed or decisively beaten until 1939.
Sorry, but your conclusion is plain wrong.
 
Just downloaded your MOD again and yup I am 100% correct you forgot to remove in the CHI_Japan.ai file the handicap that China has against japan. Once you remove that Japan will get slaughtered.
 
Lothos said:
Just downloaded your MOD again and yup I am 100% correct you forgot to remove in the CHI_Japan.ai file the handicap that China has against japan. Once you remove that Japan will get slaughtered.
Click to expand...
As described above, I removed it for the handsoffs and it didn't bother AI Japan much...
 
umm it has huge effects!!! Passitivity of 95 with a 300 day delay. For the first 300 days of the invasion China will not launch any counter attacks. It has a HUGE effect! With a 95% passitivity and your saying it does not effect it. Please rerun your sims!
 
Lothos said:
umm it has huge effects!!! Passitivity of 95 with a 300 day delay. For the first 300 days of the invasion China will not launch any counter attacks. It has a HUGE effect! With a 95% passitivity and your saying it does not effect it. Please rerun your sims!
Click to expand...
Please reread my last two posts, there seems to be a misunderstanding.
I know those numbers, but I removed them for China and it had no significant effect on the Japanese advance. Perhaps passivity doesn't work as it should, but the same test can be made by anyone to confirm.
 
Remove them again and run another sim that is what I am telling you considering I have way over 1000 hours of simming and put that in there for a reason. It obviously does something cause you removed it out of the Japanese but did not from the Chinese so what your saying is the command only works for Japan and not China. Please do not insult my intelligence here! Remove it from China and watch Japan get crushed.
 
Lothos said:
Remove them again and run another sim that is what I am telling you considering I have way over 1000 hours of simming and put that in there for a reason. It obviously does something cause you removed it out of the Japanese but did not from the Chinese so what your saying is the command only works for Japan and not China. Please do not insult my intelligence here! Remove it from China and watch Japan get crushed.
Click to expand...
What I'm saying is that DAIM Japan wins against Nationalist China even when NatChina is not kept back from attacking, nothing more, nothing less. I'm confident there was no mistake in my test, I ran a second handsoff after the first one and doublechecked in the savegame that there was no passivity for NatChina. I don't think running another couple of handoffs will make my case stronger...
 
Well your wrong and like I said if the command did not work then why would you remove it from Japan and not China.

The problem with Japan AI is very simple

Not enough units. When you remove the passitivy off China which will then have about 120 or so divisions to Japans roughly 30-40 "INVADING" divisions the Japanese get some initial advances but then China starts to get into the holes by simply over whelming Japan and cutting their supply lines off. Seen this happen a million times and I do not see anything that will change that in the BIGPACK!

As I said before how can the command have such an effect on Japan that you removed it but none on China. That statement makes absolutely no sense. If what you are saying is true and has no effect then why not release the BIGPACK with Chinas handicaps removed hehe.
 
Lothos said:
Well your wrong and like I said if the command did not work then why would you remove it from Japan and not China.

The problem with Japan AI is very simple

Not enough units. When you remove the passitivy off China which will then have about 120 or so divisions to Japans roughly 30-40 "INVADING" divisions the Japanese get some initial advances but then China starts to get into the holes by simply over whelming Japan and cutting their supply lines off. Seen this happen a million times and I do not see anything that will change that in the BIGPACK!

As I said before how can the command have such an effect on Japan that you removed it but none on China. That statement makes absolutely no sense. If what you are saying is true and has no effect then why not release the BIGPACK with Chinas handicaps removed hehe.
Click to expand...
That's exactly what is going to happen in the next release.
As for now, I don't see a different possibility to convince you than to suggest that you repeat the test I did: a handsoff with DAIM with all handicaps edited out for NatChina manually. Remember: You said China would win in the first place. I don't think you ran a handsoff with DAIM yet.
 
Good luck you will find out real fast what I am saying!

Don't have to run a handsoff game with DAIM like I said everything you guys have done I had in at one point or another with the exception of the ignore techs. Look I am not trying to be-little your work I am simply trying to point out to you why Japan wins with the DAIM pack and things that you guys have not even thought of or run into yet. I applaud the fact your are taking time to learn the very complicated system I have in place for the AI but there are allot more things you have to consider and simply changing builds etc.... or remove a handicap on one side may not necesarily grant you the results you want. Everything has a balance and you have to check the situation from multiple areas and then try and understand why it was done that way in the first place.

Perfect example, you removed all the controlling of his attacks for Germany so he is more aggressive but I bet you did not take the time to see the speed difference on when France gets attacked and falls in the summer compared to the middle of winter. In the summer time Germany can take out France in about 1 month maybe 2 and in the middle of winter it takes him 3+ months on average. In edition take into account the casualties Germany suffers when he does this. What good does it do Germany to attack in the middle of Winter and suffer heavy losses just to then get axed by a human Russian player because the German AI does not have the manpower to go up against the Russians. This same thing applies to Yugoslavia etc.... Sure your Germany takes everything at warp nine but you did not tweek the rest of the world to compensate for it. I bet you also did not take into account the long term ramifacations of things like this.

In all honesty keep going with your work. Modding the AI is by far no easy task and at some point you will start looking at the entire picture as it stands right now that is a bit complicated as you probably still do not understand the entire system in place and when and what things get going yet but as you dig deaper I am sure you will.



Edited: My last post :( Luxor said I can't play with you anymore!
 
Last edited:
Well, I'll be a sport and run some 3rd party handsoff w/o CHI passivity towards JAP and report back.

I agree on the 'Warp 9' issue in that historically Germany didn't want that winter offensive. I wouldn't either. 1944 was a different, desperate story of surprise. 1939/40 was not. Still, I am going to create a 'French Decision' event for Germany once Poland falls where there is a small chance Hitler goes apesh*t and orders the Wehrmacht into Benelux in the late fall/winter. Just so the Sitzkrieg isn't hardcoded. I hate the hardcoding stuff.

Still, I am interested to see how it performs. Of most interest to me is the FRONT section. I have something I want to try with all AIs from this point on but will talk about it later after verification. I do believe in the case of the China aggressive/passive thing, numerous things can be done to keep China from attacking in all but the most advantageous situations. This, IMO, is preferable than just locking China out from counter-offensives via passivity. That is something else I want to look into. Again, a huge weakness of the game engine (as in HoI1) - one front section for ALL fronts. Very weak.

Btw, guys you need ONE thread. I don't mean to sound nasty, but what I am saying is that I am concerned that I am missing out on info from another thread that I haven't found. That is why I suggested a forum for yourselves (posted - of course - in another DAIM thread).
 
Another point I would like to make - take this as an 'option'.

Coding the German AI for CORE v7, I had the Germans fire the BoB AI file when Paris was taken (not when Vichy was created). This cued a different build file but also in the AI load event I had a few transports added to the Germans build que like thus:

action_a = {
command = { type = ai which = core_GER_BoB.ai }
command = { type = build_division which = transports }
command = { type = build_division which = transports }
}

Now you may want to do a few more than two, but I had the Germans slowly building TPs from mid-39 so they had about 6 TPs come the fall of France, with those two in queue just in case a few got roasted in Norway. In fact, here are a few previous AI load event commands:

action_a = {
command = { type = ai which = core_GER_FallWeiss.ai }
command = { type = build_division which = transports }
command = { type = build_division which = transports }
}

action_a = {
command = { type = ai which = core_GER_38.ai }
command = { type = build_division which = transports }
command = { type = build_division which = transports }
}

So that is a total of 6 TPs forced into the queue, besides a few more in the AI build priority list (which, I guess only created a handful). Expect a few to roast over Norway. You just need 4-5 to start England on the road to purgatory.

The reason I mention all of this is because, yes, I think this is the one instance when the AI is given too big a gift - a real cheat. Might as well give Germany 6 Naval bombers. Which you wouldn't. So here is an alternative way of going about it. Make Germany at least honestly build those suckers.

Truth be told, Sealion should be a LOOOOONG shot for occuring.

Ahhh, AI scripting. Love it. Wish I had more time............
 
Wow, this is quite a verbose exchange. Thought I'd have a snoop in one of the countless DAIM threads floating around like flotsam after a convoy ambush and ran into the big scripters having a discussion. For my 2 bits anyways, I found that the original AI behavorial scripts are actually pretty good cause as I make major adjustments to the build priorities, switches, events and bonuses leaving only the behavior untouched, the game plays out much more historically as long as my new events successfully nudge the USA into war around '41 or '42. I got the good results so I'm not gonna mess with the behavior at all. Its real cool.
 
Forgorin, when are you going to email me this package? Last weekend you sent me a bunch of emails but the last few were like "disregard those files". lol
 
watched 2 handsoff games

germany annexed the SU both times (I have all the DAIMs), I think SU should be improved.

also, germany declared war on belgium&netherlands too early (before she sent any troops to her western boarders). this didnt end up with netherlands annexing germany, but still..

also, even after germany occupies them, he attacks through the maginot line. why would it do that? I mean, there were 7 divisions per province in the maginot line-provinces, and 11 per province in others. the german ai stacked ~70 divisions on its original border with france and attacked the maginot line ;p

ps: I think the japanese AI was fine as it is. This gives too much power to japan, as historically japan never annexed nat.china.

ps2: same applies for germany: germany always annexes SU.. VERY easily :)
 
largefort = 10 will prevent the Germans from touching the forts, until like there is one stubborn division there, but they should be in Paris by that time.
 
My mod is still in drydock right now. I'm changing all the brigades to total to 100 instead of 105. And then I need to testplay it a couple of times. The guy who was helping me is too busy humping to be of much use so it'll have to do the testplays myself most probably.
 
Jadelith said:
watched 2 handsoff games

germany annexed the SU both times (I have all the DAIMs), I think SU should be improved.

also, germany declared war on belgium&netherlands too early (before she sent any troops to her western boarders). this didnt end up with netherlands annexing germany, but still..

also, even after germany occupies them, he attacks through the maginot line. why would it do that? I mean, there were 7 divisions per province in the maginot line-provinces, and 11 per province in others. the german ai stacked ~70 divisions on its original border with france and attacked the maginot line ;p

ps: I think the japanese AI was fine as it is. This gives too much power to japan, as historically japan never annexed nat.china.

ps2: same applies for germany: germany always annexes SU.. VERY easily :)
Click to expand...



In my game as England using the Bigpack the Soviets annexed Germany in 1942 without my help so the Soviet AI can be very deadly. I am glad to know that Germany can win thoiugh. JApan fought very well and lasted till 1945 by itself . Thanks for the pack guys.
 
You must log in or register to reply here.
Top Bottom
Back