• 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.
Your "AI" is still much better than vanilla AI. If your British AI took 2 years to do something than vanilla AI it takes 9 years ans still nothing happens ;) If your British AI defeated Italians and Germans in north Africa then vanilla AI manages to lose Egypt and even Iraq to Italians. If your American and British AI managed to land in North Africa, Italy and France then vanilla AI manages to land somewhere in Island or Norway and if manages to land in France then is easily crushed.

Your Soviet AI managed to hold Leningrad and Moscow then vanilla AI simply hands over Leningrad even to Finish if they are at war. If your German AI sent armies to Caucasus to hunt oil and Stalingrad then vanilla AI sends them to Tundra to hunt white bears.

No my German AI failed to take Lenningrad or Moscow because it was super slow to attack when it had the advantage. It wasn't the UK AI doing great in NA it was the Italian and German AI not sending troops till after they already lost there. And my US AI would do an invasion in Italy only to sit there and not move till after the enemy showed up.

As I said it wasn't that my AI did a good job on the defense it was how bad the AI did on the attack even when they had the opportunity to do so.

And the German AI refused to research any further in encryption because it felt that the other side would never reseach decryption.
 
No my German AI failed to take Lenningrad or Moscow because it was super slow to attack when it had the advantage. It wasn't the UK AI doing great in NA it was the Italian and German AI not sending troops till after they already lost there. And my US AI would do an invasion in Italy only to sit there and not move till after the enemy showed up.

As I said it wasn't that my AI did a good job on the defense it was how bad the AI did on the attack even when they had the opportunity to do so.

And the German AI refused to research any further in encryption because it felt that the other side would never reseach decryption.

While your American AI was sitting in Italy, vanilla AI sits in Washington for 9 years ;)

In NA vanilla British AI surrender to Italians without much fuss even without Italian reinforcements, and Italians never receive German reinforcements anyway.

If your AI is not inventive and aggressive enough then vanilla AI is even twice of three times less aggressive or inventive this is the main vanilla AI problem.

One other thing where vanilla AI completely fails is evacuation it would never make Dunkerque or any other similar operation.
 
-fixed a bug where the GER AI would keep spamming new divisions even when it was low on MP and most existing divisions were low on strength.

Also: No more talk about Canadians or who liberated what. Funny thread is for fun stuff.

I was just saying that the Baltic states weren't liberated... D'OH! I wasn't discussing about it AFAIK. Anyways, mnplastic, are you aware that the OP is supposed to resemble what actually happend during WWII or you're just bashing the epic HoI 3 1.0 AI?
 
I was just saying that the Baltic states weren't liberated... D'OH! I wasn't discussing about it AFAIK. Anyways, mnplastic, are you aware that the OP is supposed to resemble what actually happend during WWII or you're just bashing the epic HoI 3 1.0 AI?

:) Yes I understand it :) I am just saying that even with all their failures RL generals performed beter than game AI.
 
While your American AI was sitting in Italy, vanilla AI sits in Washington for 9 years ;)

In NA vanilla British AI surrender to Italians without much fuss even without Italian reinforcements, and Italians never receive German reinforcements anyway.

If your AI is not inventive and aggressive enough then vanilla AI is even twice of three times less aggressive or inventive this is the main vanilla AI problem.

One other thing where vanilla AI completely fails is evacuation it would never make Dunkerque or any other similar operation.

Are you claiming you never once saw or read about vanilla USA AI doing invasions? Vanilla German AI sending troops to NA? Vanilla UK AI winning in NA? Vanilla SU AI never won in the east? Etc etc? Maybe you have your own AI that is different from what I had in my tests.

Since it has lost its 'fun' let me get serious. We as humans in this game have two major advantages. We have knowledge of past events and we have knowledge of future events. The AI in the game and the real life leaders did not have these advantages. The gripes are based on how a human playing the game would play as the AI. In 1936 very few people would know that CV's would be so important and powerful. Or that massing tanks would be better than parceling them out to individual units. One side tried it and it worked. The other side guessed wrong. Most of the war was based on one side trying to figure out what would be good enough to win. If they won they were called a genius. If not they are a loser.

Next time you play the game ask yourself why you just made the decision you did. You will be shocked to realize its because of what you know will happen in the future. And if the game is ever changed to where we 'hard wire' the AI to do X to counter what you did, you will then do Y in response. The real life leaders did not have the luxury of mulligans and do overs. How many of you won their very first game of HOI ever played? I know I got creamed in my first HOI1 game many years ago. But as a human I adapated. Unfortunately the real life leaders and the AI can't do that.

All I was trying to do is to show that while the AI makes poor decisions so did humans when faced with similiar situations. Can the AI be imporved? I sure hope so. Should it be improved? I'm sure PI and everyone would want that. But let's be a little honest with ourselves. We as humans have a super big advantage that might never be possible to overcome.

I hope most enjoyed my little sidetrack and thank you all for the nice comments.
 
jju_57
Respect for a brilliant, intelligent and funny thread. Classic stuff! :)
 
Are you claiming you never once saw or read about vanilla USA AI doing invasions? Vanilla German AI sending troops to NA? Vanilla UK AI winning in NA? Vanilla SU AI never won in the east? Etc etc? Maybe you have your own AI that is different from what I had in my tests.

I went through 3 pages of AARs. Only once Soviet AI beat Germans because Japan player invaded Europe fighting against Axis.

Only once American AI successfully invaded Europe and even managed to liberate France, but then German and Italian AI beat them back and destroyed invading forces.

Only once British AI won in NA but because Polish player conquered Italy.

I am afraid it is not just my AI ;)


This example plainly shows that non Axis AI perform sometimes better just because player directly or indirectly help it.

In my current Lithuanian game Americans landed in Europe several times too, but sent only hand full of divisions which were destroyed by German AI without much problem.

The problem with AI is that it can not perform mass landing in wide area and then support it.
 
How many of you won their very first game of HOI ever played? I know I got creamed in my first HOI1 game many years ago. But as a human I adapated. Unfortunately the real life leaders and the AI can't do that.

I think I won my very first HOI3 game as Germany on hard or very hard level. However, I lost my first game as Lithuania ;)

I am not claiming I am some sort of genius. I just had a lot of experience playing hundred or so different strategic games. Irrespectively what game was or what era everywhere worked the same strategy - mass your forces and overrun your enemy's scattered forces one by one even if enemy has much bigger overall army.

Current HOI3 AI do not understand that simple strategy. It tends evenly spread forces or put approximately same force as enemy in front.

Your AI or RL generals knew that strategy quite well. The problem that other general learned how to counter that strategy. A very good example of that was Kursk.
 
Are you claiming you never once saw or read about vanilla USA AI doing invasions? Vanilla German AI sending troops to NA? Vanilla UK AI winning in NA? Vanilla SU AI never won in the east? Etc etc? Maybe you have your own AI that is different from what I had in my tests.

Since it has lost its 'fun' let me get serious. We as humans in this game have two major advantages. We have knowledge of past events and we have knowledge of future events. The AI in the game and the real life leaders did not have these advantages. The gripes are based on how a human playing the game would play as the AI. In 1936 very few people would know that CV's would be so important and powerful. Or that massing tanks would be better than parceling them out to individual units. One side tried it and it worked. The other side guessed wrong. Most of the war was based on one side trying to figure out what would be good enough to win. If they won they were called a genius. If not they are a loser.

Next time you play the game ask yourself why you just made the decision you did. You will be shocked to realize its because of what you know will happen in the future. And if the game is ever changed to where we 'hard wire' the AI to do X to counter what you did, you will then do Y in response. The real life leaders did not have the luxury of mulligans and do overs. How many of you won their very first game of HOI ever played? I know I got creamed in my first HOI1 game many years ago. But as a human I adapated. Unfortunately the real life leaders and the AI can't do that.

All I was trying to do is to show that while the AI makes poor decisions so did humans when faced with similiar situations. Can the AI be imporved? I sure hope so. Should it be improved? I'm sure PI and everyone would want that. But let's be a little honest with ourselves. We as humans have a super big advantage that might never be possible to overcome.

I hope most enjoyed my little sidetrack and thank you all for the nice comments.


+1
 

So I wait for next year, and this time I conquer all the way to Stalingrad and I'm about to take the city, the Russians only have like 300 defending troops left, and then what? My idiotic AI Romanian and Italian allies get crushed by a Soviet attack, because they haven't bothered to research any doctrines or upgrade their troops or anything. And next thing I know the SU has pocketed half my Panzer divisions because the AI couldn't defend my flanks, and when I try to rescue them guess what? Out of supply again.


To my knowlege thats kinda what happend in realife. The Romanians couldn't hold the line behind Stalingrad and broke! That left the entire German 6th Army cut off.
 
Great reading jju57¡. In discharge of real WW II leaders, it has to be said that RL is "slightly" more complicated than this game, where we just follow a few rules, and it can be easily discerned what´s the best choice. And, above all, we have the huge adventage of retroespective: we know what happened, why and which were the mistakes.
 
Stroke of genius, pure and simple.


On a more serious side.
@DerKomtur:

We as humans have a super big advantage that might never be possible to overcome.

Actually, taking away that hindsight advantage could be done in two ways:
i) Removing every single bit of historical scripting, so you won't know what to expect. Would that still be WWII? Of course not.
ii) Making the AI make all the minimax optimal decisions. Would that still be WWII? Of course not. Germany would probably be attacked by the Soviets at the moment they start going into France, which would have twice the strength as historical.

My point is that you can not have a game both historical and rational, because history is not rational. Some of the causes and effects can be modelled, but you can never get the same chain of events to always happen with a simulation. Life is stochastic.
 
Hindsight can never be fully eliminated in a HOI-like game and some things will have to abstracted in one way or another because of the fact that it's not a simulation. Still, what the developers should strive for is to "nudge" the player and the AIs in the right direction without forcing things. Of course that life is not deterministic, but I wouldn't confuse the words "rational" and "gamey". I also think that the devs should rethink their approach to information availability to the player. IMO it should be more tied to espionage. It plays little role in the game ATM, while it should be important and IRL it was all about information-gathering. You know that ~95% of the time the war will start in 1939 if Germany is played by the AI and that there will be no Japanese-Soviet war and that France will be defeated in 1940 and that battleships will be obsolete by 1940s and... The historical leaders didn't know that. Of course, 70 years after those events everyone is a great general on paper.
 
My point is that you can not have a game both historical and rational, because history is not rational. Some of the causes and effects can be modelled, but you can never get the same chain of events to always happen with a simulation. Life is stochastic.

+1

thank you for saying this Slan. You are exactly right. And, what further complicates our views of the past is that we often skew our views and even, without knowing it often, perceive the people of the past as of the same mind as ourselves (meaning the exact same rights and wrongs or views on life and death). But, we take for granted so many of the advances we have. Hell, in the field of medical science back even in the first half of the 19th centuries, you were more likely to die from infection after surgery than from your wound itself (the top development in medical science which prevented this was for doctors to wash their hands and sterilize equipment before use... and it took them until the mid 19th century to figure this out!). Life expectancy was shorter. That would have a huge impact on the way society would have thought and acted. Then factor in socio-cultural ideologies circulating at the time and you realize you have a very different world on your hands.

Now, the second world war is not as distant as the example I used, but the same principle applies. Simply put, times were different then, and even though we have HUGE records on the war, we don't know everything. A fellow student of mine put it this way in class "I'm writing out these lecture notes as if I firmly believe what's being taught to me, but in my mind I could be thinking the exact opposite, or about something else entirely. But people studying this class in the future will never know, because they'll only have these notes to work with."

What makes history (particularly military history) is often looking at outcomes, and then trying to derive motives from those outcomes. But we also have to remember the ends don't necessarily justify the means. I don't even mean morality in this instance, I mean the outcome of something can often be completely different from what was intended.

Let's take an example for a second. the German Reich in the war. Looking back, we don't understand some of the decisions the Nazis made, especially in regards to the Holocaust and the final year of the war. "Why did they devote MORE military personnel and equipment to death camps when they were losing a war? Why even kill off the population of Jews you were using as slave labour? it doesn't make sense." It's true, it doesn't make sense. From OUR perspective. But from THEIR perspective, it made total sense. The core foundation of Nazism is RACIAL PURITY and anti semitism. This by that point (44-45) in the war was more their goal than winning the war.

it seems like I've jumped around alot in this comment (I have), but there is a point to this. What we percieve as rational in our hindsight must not be mistaken as what those of the past must have also percieved, particularly in the case of nazi Germany, where much of their leadership was NOT thinking rationally as we would define it. Likewise, much of the decisions other leaders made at the time don't seem rational to us because we're largely basing our opinions on the OUTCOMES rather than their perspective of the situation.

There's a few old assumptions about history and the past, one of which is "Hindsight is 20/20" but as we know here, nothing could be further from the truth.
 
In any case, no matter how broken it is, if the AI can't handle it, it wont get fixed. That is very frustrating and sometimes makes me wish there was no AI. Its not like the AI is such uberplayer after all even if you don't fix broken features.