Helping with analysis of one aspect of AI issues

[unmerged(12007)]

Okay, so we've all seen the threads about how the Barbarossa campaign goes when its AI vs AI, wherein a few units can overrun the Soviet Union. From what I have been able to read, this occurs mainly for two reasons :

1. The SU ai does not build enough land units, or, does not deploy them.

2. The SU ai, once the continuous front is broken, as it easily is due to problem 1, tend to stack everything on Moscow or one or two other towns.

Now, another problem that occurs repeatedly is the overrunning of British East Africa by a few Italian units. I think this occurs for the following reasons :

1. The British AI either does not build or does not deploy ground units.

2. The British AI does not transport ground units to this region.

3. The Italian AI seems to have no problem doing 1 and 2 for their own units.

Finally, from my own play, I've seen the following :

1. I've Allied minor powers such as Belgium, the Netherlands, and France move forces by sea to threatened regions and also launch at least one invasion.

2. These Allied minor power's naval forces seem to run out of oil, loiter, and lose strength without replacing it. Not sure if this is an AI issue or simply a lack of IC/resources.

Here's another observation from my own game. The Germans declared war on the Soviet Union at a point where it was so soon after a difficult French campaign that the Germans had almost nothing on the Eastern / Polish front. Moreover, the SU had conducted a successful "winter war" campaign as AI. Also, Romania never allied with the Axis. As a result of all this, within a few months, the German-Soviet conflict boiled down into set-piece battles in Eastern Poland, which tended to roll back and forth through that area; sometimes the Russians appearing to have the other hand, other times the Axis German forces appearing to. Russia actually forced the surrender of Latvia and Lithuania, and have severly mauled Slovakia and Hungary. This remains the current situation, a continuous front with numerous battles in Eastern Poland, and its been going on for a year and a half gametime.

Another observation from other players is the total lack of major power amphibious invasions, including the Allies and Japan.

What can we glean from all of this? Here's what I think :

1. The AI does in fact have the capability to be aggressive when it comes to amphibious invasions. Some of the AI modules are doing this, but some aren't. Multi-nation coordination of this among the AI seems particularly problematic. Overall, I'd say that the AI needs to be much more serious about building up an invasion force, coordinating that buildup with other powers, and then launching the invasion when sufficient forces have been accumulated.

2. The AI does have the capability of building units and shipping them to threatened areas or objectives, as we've seen from Italian AI behavior in Africa. Why British AI is NOT doing this in the same area is a matter for investigation.

3. SOME ai modules are not building or deploying land units in any kind of sufficient quantity. Not sure why some AI would have the problem (British) while others wouldn't (Italian). Meantime, the overall quantity of units on all sides seems to be quite low, allowing for easy breaks to continuous fronts even if the subject AI is not plagued with the no building/deploying problem.

4. The AI is capable of maintaining continuous fronts and building and deploying land units to maintain those fronts. However, if there is a serious break, the AI tends to breakdown and forget the principle altogether and go into panic mode. This behavior leads quickly to defeat of the Soviet Union AI by Axis AI.

5. I don't think the AI is making quite enough land units for largely land based forces. I don't think this is a matter of simply tweaking the AI file percentages either. They are either not building or not deploying units, which suggests a somewhat deeper AI problem that, however, could be fixed with some code.

6. From my own play, I've also seen very poor prioritization of builds. I saw the Nationalist Chinese attempting to rebuild their navy after they had no ports and no industry, and also only a few militia units left.

I hope this is some small contribution to the solution.

 

[unmerged(12007)]

If we get a critical mass of descriptions of AI behavior as observed by players on a consistent basis, this may enable the designers/coders to achieve better AI.

 

[unmerged(11746)]

Brits ahve a problem or two (hundred)

1) a ruddy massive build queue of cruisers, destroyers and a carrier (which they dont need but you cant get rid of).

2) Huge resources but no way to utilise them (IC Inhibited)

3) No availability (as the player) to use Commonwealth resources and IC.

4) The AI obviously cant use the above (3).

5) The AI, for player or itself, doesn't use the AnZAC's, SAffers or Canucks in concert with British forces.


Given the above (and you can play it out with the commonwealth mod in the mod forum) then Britain can play the "historical" role it did play: Crap kit and "Thank god for the ANZACs"... i.e you can ship over half a dozen or three armies of ANZACs and then you get to fight both in Africa and do a BEF... AND have enough to do some bloody research (So that by 1940 you dont have Vickers light tanks v german Tigers and Nukes)



- -

When you think about it - and even ignoring the "tactical AI issues" the problems with most of the game imbalances come from the ease with which germany can race ahead of everybody in both tech and units.

THe Yanks are handicapped by low tech (but can make it up over time)

The Brits are (currently) handicapped too heavily by limited ICs. This should be changed to utilise Comonwealth ICs (Taking it up to around 450 IC instead of 300)

BUT


make the cruiser-forces a compulsory build. (NOTE: meaning you cant prioritse over them). What this does is allow for research and the potential to build half-decent units from 39 onwards (if you do it well) ..but still leaving you with enough ground-hogs from ANZAC (et al) forces to cover the HUGE area of ground you need to with some half decent troops (NOT colonial Militia).

It actually works VERY well.

- - -

If you then take that as a "principle" you can sort out a lot of the game issues by simply chaning the world market to not allow trades with enemies... thus making resources a strategic goal once more.


Without doing so you ALWAYS handicap the (Brit) AI at the strategic level - even if the AI was Erwin reborn, it wouldnt be able to do anything with the brits 999 times out of a thousand... at the "tactical" level.

Unfortuantely it would probably also mean that they'd have to introduce a better convoying system also - as the conveyance of resources is also fundamental to the strategi game, rather, it SHOULD be.

As it stands right now there is ZERO importance to the Atlantic Bridge.. its totally pointless even thinking about protecting it... and only the italians can harass British supply lines. This, of course, can be worked around in any case using the convoy system so that you dont need to use ANY escort and NEVER loose a ship (except the odd one close to Britain once in a very rare blue moon).


Oddly enough what the Brits should be handicapped by is Supplies - which is very strangely NOT a tradeable resource. This is theone thing that really could have been mdoelled to make an Atlantic bridge, trade or military, between the Brits and The Colony (Caps) pretty much essential.

Most of the ellments for utilising supplies are allready in-game. I played once with deliberately underscroing the level of supplies to the generla population and wound up with 40% dissent and terrible ICs etc as a result - and that was JUST by putting supplies at the required level minus 1 (one) point for two years.

On her own there's no way Britain (in the game) can get all the supplies she needs UNTILL war breaks out (halving the supply requirements)... or, at least, there shouldn't be.


Now, if that isn't, ironic, I dont know what is. ;-)

- - -

Anyway - if you sit back and really study the game you'll see there are two core issues with the AI overall:

1) how it handles strategic-tactical battle actions/engagements. - Thats something that can be set aside as a "must have".

2) How 1), above, is actually constrained by, or not, the resource element of the game. Where we see rampant and blatant excesses of the AI it can almost always be traced back to either a preponderence or abscence of resources, be that IC's or Supplies, but , RARELY the actual physical resource pools such as coal, iron, rubber and oil, that are situated ON THE MAP..... it all exists in the Market, theres no need for the "real world".

As I say, there's not much that needs changing - but the market HAS to be tied onto the map at some levels...and the moment that they do that you can then balance individual nations much more succinctly. NOT to make it historical, but to make it more challenging and to actually make the STRATEGIC physical element of the world count for something - other than as "points" on a final tally.

There's a good blueprint for a rocking game here... but atm it's falling well short of its potential... and that even after all the niggles such as GUI and battle-level AI issues.

Put the "strategy" back into this "strategy" game please.

 

[unmerged(12007)]

Darkmoor : I find your observations to be far more precise than mine. Lets hope others can contribute at the same level. We all know we see the game potential, and we all hope that with the patching process it can attain some baseline expectations that we had for it.

 

[unmerged(11982)]

A couple of my own observations here -
Mostly playing as Brazil and watching what the majors are doing. Seeing pretty much the same problems mentioned here and elsewhere.

UK - that massive build que that it starts with cripples the AI. UK pretty much produces nothing.

USA - produces lots of units and does almost nothing with them.

As an experiment I let the game play for a couple of years then did the load save game trick to play as all the major nations. The UK was still stuck trying to get through its starting build que. The US had a lot of units but not one single transport ship. Japans build que had 2 fighters in it - thats it, and needed a lot more infantry. Russia was short on everything - didnt look like it had built much at all. Germany looked ok though seemed to have far less tanks than I would have expected (5 divisions armor compared to 50+ infantry) - and had wasted a lot of resources building submarines.

So for each case I "fixed" the problems - ie for the US I put 10 transports in for building, for the UK I added a lot of infantry/mech and a few transports and set their priority above all the ships. Same kind of things done for the others. I then resaved and went back to my original country - after 6 months to a year it was a WHOLE new game, the AI was actualy out there trying to win the war. Far more enjoyable to watch

Myros

 

[unmerged(12007)]

Very interesting Myros.

 

[unmerged(12214)]

A couple of additions and some support, here.

I'll echo what's already been said about Allies, and particularly the Commonwealth. The AI goes brain-numb when faced with making use of all available force. I.e., the British aren't exactly the Indians or ANZACs, so the AI doesn't consider their needs and goals as a cohesive whole. A similar problem cropped up when I, as the U.S., invaded North Africa. After dealing with the Italian army there, I set out for the peninsula of Italy but the British stayed put. I had no way of ordering them to join me.

In the Pacific, Australia has taken on Japan almost single-handedly and nearly won. Meanwhile, Italy magically sent fleet after fleet of transports to the U.S. West Coast, where they were slaughtered and still the Italians are sending unescorted transports.

If the human player is operating one of the Allies or Axis powers, I think he should be able to intervene and give orders to his allied nations' forces. And if the AI is running it all, it should basically consider all the allied nations as one force (within reason -- don't need Indian troops garrisoning New York!).

I think this must somehow be related to the neutral AI problem. As has been noted here before, a player-run neutral has a blank check to invade and conquer all he wants, as the AI is not programmed to declare war on a player-neutral as it is vs. Allies or Axis powers. Granted, Mexico wouldn't get very far invading the U.S. but by avoiding direct action against one of the great powers you can do amazing damage to the world map. That's a major AI omission.

It would be productive, I think, if the AI recognized "spheres of influence" and "operational packages".

A sphere of influence would limit or dictate an AI country's actions. The Suez is within Australia's sphere of influence, for example. But California is not within the Italians'. South and Central America are within the U.S.'s, which would obviate a Brazilian conquest.

"Operational packages" would be self-contained procedures that tell the AI "build troops here, along with transports, escorts and air cover. Then, 6 months later after x divisions are built, load them in transports for the invasion at <target>"

Right now the AI knows to build stuff but it doesn't seem to have the smarts to then know what to do with it. By building packages, all the planning would be incorporated in one process, making it look like the AI is actually planning ahead.

It's one of the basic tenets of warfare that you must have a cohesive plan and not spread your resources too thinly. And that is the first and biggest mistake made by this AI.

 

[Jon Shafer]

Good post Bolt.

The two most serious problems I have with all of HOI is that the AI doesn't seem to build enough units and that they don't like large amphibious invasions. If I could have two and ONLY two things changed at all, those would be them.
As Germany I focused ALL of my economy on research (Panthers in 1938), and completely neglected my army until 1938. Despite that, I had over 5 times as many divisions as France in 1940, and over 4 times as many as the USSR had on its western front... I don't care how good you or the AI is, you simply can't win being outnumbered like that without having vastly superior technology (which was unfortunately a category in which I completely outclassed everyone else in the game). The lack of serious AI amphibious invasions is a huge factor in the war in the USSR, because a German player can assure himself that ~5 divisions can hold all of western Europe, while that would be impossible with a human Britain/USA.

 

[unmerged(12007)]

Since this thread is intended to coagulate various player experiences into one place where AI analysis is helpful to the coders, here's another thing I like to make sure gets mentioned :

South America. I myself have seen in my game, and I've seen at least ONE OTHER player comment on the exact same thing.

The Argies (Argentina) sometimes allys with the Axis. That's all well and good.

However poor the AI might be at launching amphib invasions, they don't seem to have any trouble targetting and landing on the southernmost province of Argentina. This happened in my game and also in a game I saw another player talking about.

What is the obsession with this province? If the AI can do this, why can't it invade France?

Why do the AI invaders typically take this province by invasion and then do NOTHING thereafter ("the troops sit there growing old").

 

[Jon Shafer]

How many divisions were involved in the landing?

 

[Arnir]

I will agree that the Commonwealth faces some serious difficulities.

I do applaud that the CW is not a single entity in the game as it reflects many of the political tug-of-war matches that took place throughout the war. Contrary to popular belief, the CW nations didn't just give Britain everything it asked for.

However, all of the above being said, I don't think that the advantages of the CW are reflected. Playing as Canada, I have to research all of the tech that Britain just handed to much of the CW. For example, (unless I'm missing something), Canada did not separately develop Spitfires, Hurricanes, Halifaxes, Lancasters etc.. I'm not sure how I get British Tech in the game. I always give all of my tech to the UK, NZ and AUS as I feel it is the thing to do. (sometimes SA).

 

[dsteve3]

Originally posted by Myros
*snip!*
So for each case I "fixed" the problems - ie for the US I put 10 transports in for building, for the UK I added a lot of infantry/mech and a few transports and set their priority above all the ships. Same kind of things done for the others. I then resaved and went back to my original country - after 6 months to a year it was a WHOLE new game, the AI was actualy out there trying to win the war. Far more enjoyable to watch

Myros

How? I can't run the editor that came with the program (exception errors galor...)

Can you tell me how to do these things?

 

[Alwin von Arlt]

Interesting

Pretty much seen the same thing although my argentina game, did conquer all of south america =) joined the axis, and spent the last 3 game years fighting a war with the English in the Brazillian rain forest, no clue where from cause of fog of war, I would of lost eventually as their tech massacred mine for the most part, but my alliance won in the end =) so I'm happy.


in '47 the English did send 20 units to argentina proper with an amphibious landing, and won the initial battles, so i don't really know how bad the amphibious ai is.

 

[unmerged(12007)]

Originally posted by dsteve3
How? I can't run the editor that came with the program (exception errors galor...)

Can you tell me how to do these things?

There is a save load method that enables you to switch playing to any nation you want. Its in the FAQ now.

 

[unmerged(12007)]

Lets keep as much serious AI analysis goodness flowing to the coders-that-be.