A couple of AI issues

[JudgeDredd]

A couple of things I noted prior to any landings by the AI Allies...

• The Allies seemed very keen to bombard the ports. Not a problem, except they would "stack" naval units on a port without any possibility of some units being able to fir...for example...a port with 3 sea hexes would have 3 allied naval units there to bombard...but the AI would move MORE naval units and put them in the hexes "behind" the attacking naval units because there was nowhere for them to go! A waste of resources there...seemed like they had an "instruction" to move to the port because there was an enemy unit there, but did not know that there were no sea hexes available for them to move next to the unit to bombard
• Still no sign of any troops in North Africa or Italy

March 1944
The allies have landed....only Brits at the moment...but it looks like the yanks are going land directly from the US...I never did see US troops on British soil! I did however spot a British unit (unless it was Canadian - I haven't played as Allies yet) in the US!!

Also, a UK sub unit did appear in the MED in early 44....still the landings seemed piecemeal at the moment...considering they have the whole French coast to attack.

[Dave Ferguson]

allied naval stacking ouside ports -
 
common problem with AI's, they make a plan BEFORE the moves but do not take into account the change in situation AFTER the first unit has moved. To do that they would need to make a new plan after every unit has moved. It could take forever?

[JudgeDredd]

Not really. I mean, it would take longer, but only because of the change in situation meaning there may be "more options" to work out.

The point is, I think, that the AI has to make a calculation for each unit...if it does it at the beginning of the turn, or on a unit by unit basis doesn't necessarily mean it's going to take longer per se....simply that because of the changing situation there may be more parameters to take into account.

What it does show is one of two things...

• Either the AI is making it's unit by unit decisions prior to a global movement phase
• or
• The AI is making it's unit by unit decisions prior to a unit only movement phase

If it's doing the former, the the AI is actually poorer than expected simply because the fluidity of the battlefield has not been accounted for when it was designed. If it's doing the latter, then it's doing it wrong.

I don't want to or mean to be condescending with my comments. You (developers) have managed to program a better AI than I could and working in the programming world I know that "users" come to me saying "Can you do this?" thinking it's pretty simple...and on paper it is...but taking in all other factors ignored by users makes something so simple into a very daunting task.

I'm just pointing out the fault I see in it.

[Dave Ferguson]

Pointing out faults is always good feedback but of course I have nothing to do with this game other than as a player. Personally I don't particularily like playing AI's but have had some experience of them during playtesting of other games. IMO you could put 50% of your development effort into the AI and it would still not be enough.
 
Dave

[JudgeDredd]

Absolutely. I do agree that AI is very, very hard to get to a decent level. Then you have to tweak it in various modes for varying levels for the customer.

I understand that AI programming is a massive chunk of development time, however, I'd hate to see it in the decline.

I'm not an online player...too many cheating bastards who want to win at all costs, ergo making the experience pretty usless, being as you never really know whether you are playing on a level field or not. This all came to pass, for me, in that massive multiplayer game that was Counter Strike. Loved it then realised that very few people actually played within the parameters and limitations of the game. The same with Battlefield 1942 and Battlefield 2. I loved both of those, but cheats were out before the games were released!

Whilst I expect cheating is far less likely in the wargaming industry, I think it still goes on, and for that reason (not knowing who's doing what) I never play online.

I do not mind losing, but I don't like losing when the odds are unknowingly stacked against me.

[Forwarn45]

With a game as Allies versus the Axis AI, I have the following observations:

(1) On the East Front, the land AI was playable.

(2) But on the West Front (after the Fall of France), the AI needs work. It ignores Africa and - more importantly - did not respond to my invasion of France. It left La Rochelle open so I moved to invade in that direction. Then it moved its panzer from Brest down toward LaRochelle before I invaded. So invade Brest. The AI was doing fine on the Eastern Front by the time I decided to invade and had plenty of troops to draw from. But it did not reinforce France to contain my beachhead. It just left its troops in the cities (Paris, Cherbourg, etc.).

[Barthheart]

Wow. You are at least the second person that said this. I never seen the AI leave a coastal city/port open to invasion. I've always had to fight for them. I always send a fleet to check out what's open and they are all always covered.... on normal difficulty... invading in mid 42 to late 43. What level are you playing? What year did you invade?

Now, after I have a toe hold the AI just siits on it's cities and makes me pry him out. The odd armour/corp shows up to fight but is usually quickly cut off. Not until I apporach the German border does the resistance get much stiffer...

[Forwarn45]

I was playing with moderate advantage to the Axis AI. I invaded in late '42 - I think it was October. My navy had scouted out LaRochelle and saw it was open. It is funny that the Axis AI would think to cover the opening - but in so doing, would open another port! I did not have to actually invade at LaRochelle to get this response - once Brest was open - my troops were diverted to unload up there instead.

[IrishGuards]

In this scale the AI should never ... ever leave a city empty .... unreal
IDG