AI sends troops everywhere!

[bob.]

Noticed that the AI apparently has no concerns for nationality of stacks when sending them around which results in some really weird stuff...
mostly small things like Germans holding the line in Galicia or Russian siege artillery in Königsberg.
But here's the most absurd one I encountered so far:



Well, Lemberg won't fall too fast. Problem is: Armenia is now completely undefended.

[Owl]

Could probably be easily fixed by giving the troops from a certain area a regional restriction like in RUS, but that would on the other hand prevent some interesting strategies for a human player...

[Reiryc]

try turning off AI detection bonus in options. Go to options, AI, and set ai detection bonus slider all the way left.

i'm not sure it will affect your current game however.

[Owl]

Wouldn't that negatively affect AI performance in other areas, though?

[Reiryc]

Wouldn't that negatively affect AI performance in other areas, though?

Not really... depends on how you play against the AI. If the AI makes deep raids then you can cut them off--which is what I do. Some try to hit them head on. If however they stick together for the most part, then the front pushes a bit more together because the AI doesn't see such big gaping holes and sends everything through them.

[bob.]

I have AI detection bonus on low already... I guess I could turn it off but I doubt that will change a lot, tried the same in CW2.

Just loaded the other two sides from my Eastern Entente game to see how the AI does. Wow. I am amazingly disappointed. What the hell?!?
So, I have it confirmed for myself now. AI is still absolutely shit. Shame. I guess I will have to find some PBEM players soon.

Want some best-of?

"We need some ships to patrol French Africa! About 1000 ships should suffice!... obviously not the whole fleet, since we need 30 submarines at harbour in Ottawa!"


"What are those front and supply lines you speak of? I'm gonna send a whole army starving somewhere in a German forest while the front line is at the border!"
(Notice also how there is no front line to speak of, just some stacks dispersed... somewhere!)


"I know how we will win! We just make a big fucking stack! And we will call it "Armee Francaise". Vive la France! 227 CP needed, 52 available? Meh, who cares! And where will we send the big stack? Oh, obviously INSIDE A FORTRESS TWO PROVINCES BEHIND THE ACTUAL FRONT!"


"Hm, where should we retreat to? Into the province that is our front line and where we could nicely defend over a river or... into a province surrounded by Russian armies?" You have one try to guess correctly what the AI chose..


So in summary, I am sorry but the AI is a disgrace. It doesn't even do the most basic things correctly, like not making 200 CP stacks or looking out for at least something resembling a front line. Very, very disappointing!

[TrenchFoot]

Maybe a future patch will give the AI a brain. At this point I would even settle for half a brain....

[Bismarck1940]

The AI needs fixing to make this game playable, except PBEM. I will stop complaining now and wait for the fix. The AI does some things fine; but the problems are overwhelming at this point. CW2 and Espana '36 have much better AIs (either side)

[Respenus]

Hmm, this is very odd behaviour. Right now, my CP campaign is up to August 1916 and while I've seen the AI pull one or two less then smart moves, I've otherwise been confronted with a responsive AI that can cause problems if you don't plan your moves smartly.

[Ace]

It is HIGHLY recommended to give AI no bonus in detection. Whith higer detection, it sees that objective cities behind the front line are garrisoned. And, then it tries to capture them by forming a large breakthrough stack (like in your example). When no bonus is given, it behaves much more logically.

[Owl]

It is HIGHLY recommended to give AI no bonus in detection. Whith higer detection, it sees that objective cities behind the front line are garrisoned. And, then it tries to capture them by forming a large breakthrough stack (like in your example). When no bonus is given, it behaves much more logically.

Would you argue for that in general in regards to AGE games or just specifically for EAW?

[bob.]

I guess I will set it from low to none then if I start my Western Entente game (my Eastern Entente game I can consider "won" now at turn 15 because of the CP stunning incompetence...)

Still, this does not explain nor exuse things like a 16000 power ship stack off French North Africa nor does it excuse forming these huge "breakthrough armies" with severe penalties in the first place. I am disappointed that apparently there have been no improvements to the AI whatsoever since years for AGEOD games. I notice the same retarded behavious in AJE, PON, CW, CW2, TEAW, RUS, ROP, WIA... and I am sure there WERE improvements to the AI but I just don't see much of a change.

I am not asking for a "great" or even "good" AI. I realise that this is basically impossible with a game of this complexity. And I can make some house rules for myself so the AI has a chance. However, playing against an AI this bad is just dreadful. I don't know why with every new release I expect the AI to become better...

[Ace]

It is so on most ageod games. For example in cw2, AI tends to do silly deep raids towards empty enemy objectives. But, here it is more obvious because of the nature of the war. So, my personal tip - no detection bonus to the ai.

[Owl]

It is so on most ageod games. For example in cw2, AI tends to do silly deep raids towards empty enemy objectives. But, here it is more obvious because of the nature of the war. So, my personal tip - no detection bonus to the ai.

Thanks - I find that interesting though, since I remember at some point reading in the RUS forum that somebody suggested detection bonuses and I've generally stuck to that - guess I will try no detection but keeping the activation bonus the next time I play the AI to see how it fairs. Also the game tool-tips suggest the detection bonus, maybe they ought to be changed?

[HerrDan]

Basically I've seen the AI performing poorly, but these screenshots you posted look very strange indeed. I agree with Ace that the less detection bonus you give to the AI, the better the results.

[Kensai]

The bizarre behavior described by bob must be rather uncommon to extract such generic remarks about the quality of the AI. I am not entirely convinced AI detection plays such a major role, cause the AI is hardcoded NOT to make such big stacks. I always play with Medium AI detection and never seen Turkish troops in Austria-Hungary or stacks-of-doom.

Just keep something in mind, your mileage may vary. It is generally recommended NOT to try any current generation AI in a gamey way. If you do, it will kick you back with some strange behavior. I don't know what went wrong here, honestly, but the most "silly" behavior I've seen by the AI is sending (or keeping) some stranded units behind the enemy lines. Definitely not... these screenshots!

[bob.]

The bizarre behavior described by bob must be rather uncommon to extract such generic remarks about the quality of the AI. I am not entirely convinced AI detection plays such a major role, cause the AI is hardcoded NOT to make such big stacks. I always play with Medium AI detection and never seen Turkish troops in Austria-Hungary or stacks-of-doom.

Just keep something in mind, your mileage may vary. It is generally recommended NOT to try any current generation AI in a gamey way. If you do, it will kick you back with some strange behavior. I don't know what went wrong here, honestly, but the most "silly" behavior I've seen by the AI is sending (or keeping) some stranded units behind the enemy lines. Definitely not... these screenshots!

I have to disagree with you here. As I wrote "I notice the same retarded behavious in AJE, PON, CW, CW2, TEAW, RUS, ROP, WIA... and I am sure there WERE improvements to the AI but I just don't see much of a change."

This is NOT a one-time problem. I was hoping maybe TEAW would improve it because I read in some AARs about the AI actually building fronts or something. I noticed that a little bit but with the otherwise terribleness it did not matter.

About big stacks... I have noticed this in all other AGEOD games as well. The AI does not seem to care about supply or CP penalty at all. Or at least at a very low priority.
When I first saw that 16000 power fleet, I thought the game data was corrupted or something

And I am in no way "gaming the game" or anything. In fact, I love making slow, deliberate advances in this game. They always get frustrated by the AI sending small stacks to cities in my back which is annoying as hell and serves them zero purpose. Really saps the fun out of it.

Frankly, I am having problems imagining any beta tester actually playing the game to 1918/1919 against the AI and having something even resembling a cohesive front throghout the game. I would love to find out if I am doing something terrily wrong, though!

Edit: BTW, these were my settings for the game.

[PJJ]

That screenshot showing the HUGE British naval stack in Africa is both fun and sad at the same time!

I've also noticed that supply is definitely a problem to the AI. That's why I always play these games with the easiest supply options for the AI.

[Kensai]

I think I remember Pocus had introduced engine-wide code which makes huge stacks of armies and fleets less probable. He probably needs to up the effect. I repeat: it could happen, but it happens rarer in the latest patches of all games.
Regarding the "retreat bug". It is a probability chance here as well, if you see the settings code you will see it happens quite rarely (there is a very low percentage to retreat in an unfavorable region). It is a natural reaction of a human player to think it happens that often but if you count all the other times the armies retreated in a "good choice" of region (and you didn't notice) you will understand it's a nice mechanism: once in a while, there is an unlucky army movement.

[bob.]

Also, I want to note one more thing about the AI. I had the feeling when playing that the AI has zero understanding of the concept of entrenched defenders. In Serbia I had at least five battles where they attacked my entrenched main force in hilly region, always suffering large casualties with me suffering tiny casualties.
Same at the Russian front where I constantly defended against small-scale attacks which NEVER succeeded. There was not a single big offensive push in these 15 turns apart from the first two turns where they attacked my initially-locked armies.

[Kensai]

Also, I want to note one more thing about the AI. I had the feeling when playing that the AI has zero understanding of the concept of entrenched defenders. In Serbia I had at least five battles where they attacked my entrenched main force in hilly region, always suffering large casualties with me suffering tiny casualties.
Same at the Russian front where I constantly defended against small-scale attacks which NEVER succeeded. There was not a single big offensive push in these 15 turns apart from the first two turns where they attacked my initially-locked armies.

Ehm, silly attacks against fortified lines happened in the real WW1 as well. Remember Verdun?

This is WAD.

[bob.]

Ehm, silly attacks against fortified lines happened in the real WWI1 as well. Remember Verdun?

This is WAD.

They happened, but the commanders were not THAT stupid. If they attacked fortified lines, then not without artillery and a considerable local troop superiority. With exceptions maybe, I am not a WW1 expert. Still, even if it did happen in reality this is not how a good AI should behave! Especially not five times in a row with the same forces.

Well, I am going to try playing the Western Entente now with same settings but no activation bonus. Maybe it will not be so bad.

[Kensai]

The recommended settings say for low-medium AI activation. Actually, there is UI bug here, each switch gives a different suggestion. In all AGEOD games the recommended setting is medium. I always set it to Medium. Try neither higher nor lower. With no AI detection there will be no strategic depth and imagination for the AI armies.

[PJJ]

Also, I want to note one more thing about the AI. I had the feeling when playing that the AI has zero understanding of the concept of entrenched defenders. In Serbia I had at least five battles where they attacked my entrenched main force in hilly region, always suffering large casualties with me suffering tiny casualties.
Same at the Russian front where I constantly defended against small-scale attacks which NEVER succeeded. There was not a single big offensive push in these 15 turns apart from the first two turns where they attacked my initially-locked armies.

Regarding Serbia, that's exactly what the Austrians did in 1914. Under the incompetent leadership of General Potiorek, they launched several offensives against well-entrenched Serbs, with disastrous results.

[bob.]

OK, fair points.
Then let me say this: if, with the AI being as bad as it is, making it attack entrenched defenders at bad odds was a conscious decision by the developers, then I do not think it is a good one because the player is not constrained by it.

[Kensai]

Do not underestimate Athena, especially in its later iterations. Unless you have a save game that proves otherwise, I bet the AI throws to you minor attacks (conservative or probing) that in the end take little damage themselves. Yes, it brings up the battle interface, but damage is limited for the attacker. This is what I have been seeing myself in the game, at least. The AI makes calculations before launching an attack and its stance seems to depend a lot on the power ratio. The AI will never do suicide attacks unless there is a compelling scripted regional bias.

[vonRocko]

With every new Ageod game I pray they improve the AI, but every time I see the same old Athena nonsense. I thought a WW1 game would force them to alter it significantly, but I guess I'm wrong.
IMHO this engine just isn't up to too this.

[bob.]

Kensai, I do not really understand why you argue so heavily that the AI is so good. I am not trying to bash the game here, I love all AGEOD games! They are in fact more or less my favourite PC games in general. That does not change the fact that the AI is horrible though.

Sure I have a save game that proves otherwise, the one I am posting about here! If you really need more "proof", here's some more screens:

Central Powers lost 1 MILLION men, a lot less than the Entente powers combined. And most of those casualties were, obviously, inflicted by me. At least 600.000 anyway.


The East Prussian "front":

notice that while I have a 100 % continous front line from Königsberg to the Romanian border, the AI has nothing but some scattered troops. They do not have the strength to resist me anywhere.

"But where are all the German forces" you may ask. Good question! They are, obviously, at Breslau, away from any useful position, in one huge army that attacks exactly nowhere.

A striking similarity to the "Armee Francaise" above, showing that this is NOT A SINGLE OCCURENCE. The AI does this all the time!

On and one last screen: German artillery is going for a "behind the lines recon mission". Ah yes, the good old sneaky 122mm gun (it's propably not a 122mm gun), easily concealed for those surprise attacks behind the lines...


You know what, especially this last thing has convinced me that in fact there has to be something wrong with the AI in this specific game. In other AGEOD games, it was never THAT retarded.

[vaalen]

Well, Bob, you have convinced me. Your screenshots show me the worst AI deployments I have ever seen from Athena, and I met her on her very first outing, and have every game she ever appeared in.

Something is seriously wrong here.

[caranorn]

Do not underestimate Athena, especially in its later iterations. Unless you have a save game that proves otherwise, I bet the AI throws to you minor attacks (conservative or probing) that in the end take little damage themselves. Yes, it brings up the battle interface, but damage is limited for the attacker. This is what I have been seeing myself in the game, at least. The AI makes calculations before launching an attack and its stance seems to depend a lot on the power ratio. The AI will never do suicide attacks unless there is a compelling scripted regional bias.

If that's the case you might want to look into Metz, I've seen repeated French attacks against that position (stopped with the winter turns, hope it won't start again in spring) which led to the complete destruction of several of their divisions. Of course Metz would be an important target, but not to such a point...