Axis AI does some funny things

[Happycat]

I am sure I'm not the first to notice these quirks. Let me start however by saying that with limited help turned on for the Axis AI, overall it does a pretty good job.

As to the quirks---the AI started by attacking Poland (quelle surpris!). But, in turn 2, when it had an opportunity to close in tightly on Warsaw, it diverted units to attack weakened and isolated Polish units back near the border. In turn 3, it did take Warsaw. After capturing the capital, it continued attacking and killing isolated units, and also moving just to capture territory. Although I suppose the extra experience points for fighting might be of some small benefit, I think that the negatives are far more significant. By continuing to fight, the AI took losses which it had to replenish (and pay for). The additional and unnecessary movement just to capture hexes was a waste of limited German oil supplies.

In France, it was even worse. In June of '40 (after invading the Lows in Nov '39), the French-Anglo forces in front of Paris were in some difficulty. A concerted effort would have cracked the line, and Paris was only held by a weak unit. But, the turn before, I had finally vacated the Maginot Line. The AI decided it was worth diverting THREE panzer corps to obliterate the empty fortresses. So, Paris still holds out, when it shouldn't be.

I can easily understand why the AI does this---I'm sure part of the coding tells it to go after empty fortresses and to gobble up territory whenever possible. But surely the coding could be weighted, so that an empty or weakly held capital would be more tempting to the AI.

I would be interested in the thoughts of others on this, especially CEAW developers and testers.

[firepowerjohan]

It is very complex indeed and often AI has to have some security margins not being so much on the edge as a human player. If you instead of focusing on grabbing land and resources start to attack "the center" and fail, then it can very well be that you are stalled for several turns doing the same attack and failing. Grabbing the enemy resources and beating the enemy units is a safer way, but all ways will provide cases where a human by logic understands which alternative is better.

What if we suddenly change the rules so that some countries do not surrender until some turn later, that would mean a single advance into enemy Capital could be met with a large counter if you have not properly tried to defeat the enemy first.

Furthermore, adding several capitals like USSR for instance having Moscow and Perm then defeating the USSR units is more important since when you capture Moscow what will determine your continued march towards Perm is also that you have caused USSR anough losses so that they cannot slow you down.

[Happycat]

What if we suddenly change the rules so that some countries do not surrender until some turn later, that would mean a single advance into enemy Capital could be met with a large counter if you have not properly tried to defeat the enemy first.

Yes, I remember at least one computer game that had this rule in place. IIRC, it was applied to ALL capitals, which was a poor idea---if Warsaw falls, it falls. The conditions required to permit an encirclement of Warsaw pretty much preclude there being any significant Polish assets left for a counter-attack anyway.

France, otoh, perhaps should have the ability to do one counter-attack. However, could it be coded so that the conditions for a counter-attack include a certain minimum level of French assets still be on hand? I know that I would be extremely p.o.'d if I had defeated the French in detail, took the capital, and then had to wait another turn just because of a game mechanic.

It's a tough decision isn't it? I honestly don't know which way would be better. Perhaps it can be an option that can be clicked on or off prior to starting a game? (e.g. Major Power capital counter-attack)

[IainMcNeil]

I think that Johans point is that he has tried to make the AI general enough so that it tries to defeat your troops and grab resources as well as capture captials so that it behaves realistically and as a result it remains flexible enough for us to tweak the rules. The player however is trying to maximize results and exploits all the rules/mechanisms to their full potential. If we made the AI very specific any changes to the rules would completely throw it and could completely break it.

E.g. If we decide to change surrender rules so they requires the captials plus 50% of all cities belonging to a side, the AI will cope well with that without any changes really. If we decide to make the surrender based on a % of land under your control, the AI will also cope wth this. If we decide to make it so that surrender occurs when the military might of the nation is below a certain level, the AI will also cope with this. That's pretty cool

[Happycat]

I'm sorry, I'm not fully understanding either of you then. At first you appear to be saying that the AI, in its general approach, does the job as well as it can. But then I get a sense that you might be open to tweaking the AI some more.

I hope I am being clear: I think the game is very good, and enjoyable, and somewhat challenging against the AI. However, the moves that the AI made in France were absolutely counter-productive in the last turn that I refer to in my earlier post. When the Axis has major units, with little or no damage, within two hexes of Paris, and the defensive line is tattered, it makes no sense for those units to divert and start killing empty fortifications on the Maginot Line. This move took the units a good distance away from the main point of attack, and allowed me time to adjust my lines and strengthen the weak units. It has probably cost the AI two turns minimum.

As you know, in a wargame of this scale, losing two turns in 1940 pretty much screws up the timetable for 1941.

Don't misunderstand me---I realize that I am playing against AI and that it is non-sentient. I don't expect miracles, but the movement I described seems to me to be something that the AI could be deterred from doing.

Of course, I haven't yet tried this against an Axis AI receiving maximum help. Perhaps I need to do that...

Thanks for listening!

[syagrius]

The game ok and is balanced as it is. Maybe adding Stalingrad and Leningrad with Moscow and Perm would be good.

[Stryder]

the truth is the real Axis did some funny things:

allowing Dunkirk
ignoring Malta
not adding more forces in Libya when it could have made a difference
not going into winter quarters before blizzards hit and not being prepared for winter 41
trying to go after Stalingrad and the Caucusus at the same time and getting neither
not trying to break out of Stalingrad (if possible)
keeping the toehold east of Crimea 42-43
Kursk.. kept delaying the attack until there were massive defensive lines and then still going forward... what happened to avoiding strong points?
not abandoning the Crimea 43, losing troops there
rushing reinforcements into Tunisia when it was too late
Leaving alot of troops to defend Courland '44 when they could have been defending Germany
Pulling troops out of Poland to defend Hungary and leaving the path to Berlin open
Offensive in the Ardennes using most of the reserves and armor left on a desperate gamble instead of protracted defense
.. and I am sure there are others


of course, it's easy to have opinions now with the benefit of time and alot of good analysis available to read and many of these points are open to debate, but very often it's the side that makes the least mistakes that wins the war

[gmothes]

Nice observations Guderain. The AI needs help in the Med on both sides, but does well in Russia and well enough against the west...

[sonnenwalde]

While I've only played a couple of games (as Axis) the AI has performed reasonably well in the USSR and the West. The main problem, as Guderian states, is in the desert. As Axis I found it far too easy to defeat the Allies and therefore capture much needed oil supplies. This theatre of conflict definitely needs looking at.

[vypuero]

The one thing I might say about the Ardennes offensive is that Hitler had to know he was a dead man if he lost, so a normal leader would do the best for Germany, but for Hitler it was all or nothing, and in the end he cared nothing for his country or its people.

[sonnenwalde]

By December 1944 the Germans had lost the war and the Third Reich only had a few months to live. The Ardennes offensive was the last throw of the dice for a desperate leader. Attacking in the East would make little or no difference with the massive Soviet manpower & materiel advantage, so he had to try something on the Western front.

[Peter Stauffenberg]

My main issues with the Axis AI is that it almost never attacks Denmark and Norway and that it doesn't attack Holland before it attacks Belgium.

It's very easy for the AI to attack Denmark on turn 2.

Turn 1: Rail the units in Prague and Munich to the hexes near Kiel (1 at the Danish border and the other adjacent to the port).
Turn 2: Use the northermost fighter to bombard Copenhagen. Use the battleship to shore bombard Copenhagen too. Move the unit at the border to the island west of Copenhagen and attack. Load the unit adjacent to the port and unload it immediately to the hex south of Copenhagen and attack.
Turn 3: Repeat the process if Copenhagen didn't fall during turn 2.

The turn after Copenhagen fell you move the air units to northern Denmark and load your units in Denmark aboard transports and move them south of Oslo. Use the destroyer and battleship to protect these transports. If the air units and/or land units took lots of hits in the fight for Warsaw and Copenhagen you postpone the move towards Norway. You simply rebuild the strength of these units to maximum.

The turn after you moved the transports you declare war upon Norway and bombard Oslo with these 3 air units. You land 1 unit east of Oslo and attack and the other west of Oslo. Use the battleship to bombard Oslo too. This should put Oslo to its knees. You may alternately put 1 fighter in Western Germany to protect against strategic bombardment.

The units who attacked Poland should be railed and moved to the border towards Holland. When enough units are placed there then you declare war upon Holland, but not Belgium. Attack Holland and quickly capture Hague. Hopefully the units who went for Oslo managed to capture Oslo so the air units can be used in Holland and Belgium. You probably built a tactical bomber early with the Germans so it's ready for use as well.

From now on the Germans can attack Belgium and France the usual way. Belgium will fall quickly when they can invade from several Dutch hexes as well.

By doing this victory in the west should be achieved by the Summer of 1940 at the latest, thus giving Germany a chance to go after Yugoslavia and Greece before attacking Russia.

Strengthening Libya should be done too.

I think the German advance into Russia is not so bad. He's quite aggressive and the Russian human player must be careful and retreat unless he wants he front to collapse completely.

I think it should be possible for the Axis AI to be instructed to go after Denmark, Norway and Holland before they attack Belgium. These fights are easy to make and will give Germany extra income earlier. OK, the Germans need some units to protect Oslo, Bergen and Copenhagen. But the 2 corps going after Norway can remain there so you don't have to pay PPs to transport them back to the continent and even more PPs to transport garrisons to Norway. Copenhagen can be protected with a garrison rail moved from e. g. Vienna.

[firepowerjohan]

Norway invasions is very suspect to counter by humans doing "win everytime" strategies and then complaining AI is weak. Similiarily, IrishDragoonGuards showed how Allies can really punish Axis if Axis DoW Denmark early by bringing counter in the west and in the Baltic. So, we try to make an AI that has a security margin hence is abit more cauious than humans, so that is does not make too many mistakes

[syagrius]

In my first Allied game, I noticed that the Axis AI doesnt care enough about his flanks: in Russia I destroyed many Axis armor and corps out of supply because they went to far into my lines.

[syagrius]

Yesterday I have finished my first game as the Allies, playing with slight advantage Axis. I won easily in 1944 taking Berlin and Rome.

The AI's flanking problem I mentioned in my previous post is the main cause of the Axis defeat. The Germans lost too much armoured corps against the Russian. It seems they try too much to reach Moscow and neglect the north and south of the Russian front, which results in flanking and cutting of supply. I will try next game with moderate advantage Axis, unless the upcoming patch solve this.

[ancient]

The Axis AI still refuses to defend the French coastal cities. What's the logic behind that?

[syagrius]

In my Allies games the German always had infantry Corps in those cities, however their army in France is too small to stop my invasions.