Explorations in the AI Editor

Advanced Tactics is a versatile turn-based strategy system that gives gamers the chance to wage almost any battle in any time period. The initial release focuses on World War II and includes a number of historical scenarios as well as a full editor! This forum supports both the original Advanced Tactics and the new and improved Advanced Tactics: Gold Edition.

Moderator: Vic

Post Reply
miral
Posts: 170
Joined: Thu Dec 20, 2007 7:57 pm

Explorations in the AI Editor

Post by miral »

First let me warn you that this will probably be a long thread, due to have complex I have found the AI editor to be. So I wil probably break it into more than one post. We'll see.

When I say the AI editior is complex I don't mean to use; like the rest of the game it is elegantly simple to change settings. But, also like the game, it is deeper the more you look at it. And this thread is by a beginner not versed in computers mostly for same, though I hope all can find something useful in it. When I say I have 'discovered' something I mean discovered for me; I am sure that more experienced and astute players than me have already discovered the same thing.

For example, I 'discovered' in the AI editor that the AI personalities are randomized for each nation (faction/side). I either missed this in the manual or it is not there. Ohh, you get to the AI editor by creating a random map, clicking the Edit button in lower left. There are a whole bunch of buttons at top left and center. Click Settings. Then, on left side is a box with a lot of lines rather than buttons. Click AI, near the bottom of the list and you are there. Also, take a look at all the other things in this list for a host of other settings you can change.

I had a post called Use the AI Editor for a Much Better Game in which I extoled the Editor for its simplicity and the many things you can change. I still do, but it is not as simple as I thought. The reason for this, which should have occurred to me, is that many the parts of the game are interconnected so that if you change one setting, it affects others that you may not have meant to change.

Now, so this doesn't run to a book (and, indeed, one could write a small book about the AI editor) let me start by giving an example, which may or may not apply to what others want to do with the Editor, but should still give useful information about general concepts.

First, you need to decide what you want to do with the Editor. That is, you can just change some settings to see what happens or have a more specific idea of what you want. I started with the thesis that most people who criticise an AI for being weak do so because of its Strategic weakness but that it can be strengthened by changing it tactically; in AT by changing the size and composition of the units and the objectives set them.

Since the dawn of Civ I it seems to me that most Ai's have wanted to blast out a huge number of weak units. I find the default settings of AT tends to do this, so I started with increasing the weight in stacking points of each individual unit (No. 182 in the Ai Editor list). The default is 10, which seems to me far to low, because although this is the minimun stack points when I have used it it seems to function as the norm). So I set it to 250. Bad move, as will be seen. For one thing I had forgotten the penalty for overstackin in a hex.

No. 156 will add more support weapons (bazookas, MG, mortars ect) to units and I lowered (you increase some settings by lowering the default and others by raising it so be careful) as, in the games I play the AI seems to produce a very narrow range of weapons.

Next it occurred to me that, if I changed the % of front the AI tried to cover (No. 155) this would cause it to produce fewer units, so I changed AI Front coverage to 50%. These two settings, 182 and 155, are, by the by, two of the most important. Changed other settings too, but let me keep it simple as I have probably already bored many away.

So I played some games and found that I got - a huge number of very large units composed mostly of infantry! What had happened? I should mention that I play usually, and tested this, on maps of small to medium size, both because that is my game preference and smaller maps enabled me to test different settings faster. Now, I turned fog of war off so I could see the composition of enemy units. I wish there was a way to switch back and forth from AI to human to AI with the same side in a game but it seems there is not. This would enable me to see what the AI is doing with production and PP. Since I cannot I must infer what it is doing from the make up of the units and what they do on the map.

So why had my wonderful plan not worked? Because I was giving the AI contradictory orders. I told it to produce large units of multiple weapon systems AND cover 50% of the front. Now, the front coverage setting is one of the most important of all. Even on a medium map 50% of the front can be a lot of hexes. In order to cover that the AI still had to make a lot units and, since I had told it make very large units, it could not do these two things and stil have the PP to produce lots of support weapons. (I also got gigantic HQs to support this vast hord of infantry). This is what I meant by being aware when you change a setting that it can interact with other settings. So, how to get what I wanted?

Let me stop here, both for a rest and so I don't accidently click close and lose it all. I did find a solution and will describe it in next post. Thanks.
miral
Posts: 170
Joined: Thu Dec 20, 2007 7:57 pm

RE: Explorations in the AI Editor

Post by miral »

I in common with Shakespeare only two things; first name and the fact than I am never brief. Sorry for the length but I don't know how to get into this deeply in an abbreviated manner.

Part II - Before I tell you how I solved the problem talked about in the first part of this post let me comment that we don't know the relative power of the various settings of the AI editor. That is, when two or more settings we make are contradictory or in some way not completely harmonious, which one takes precedence? The only way to find out is to do a lot testing. I have done, if not a lot, a fair amount, and it seems to me that Front setting trumps most other conflicting settings, probably because it runs with the AI's general proclivity to mass produce large numbers of units.

So...I kept turning the Front % coverage down and down and still getting hordes of enemy units. Then it occurred to me that the whole concept of Front coverage is both peculiar and unhistorical. For example, for Russian front in WWII, we see in history books maps with nice neat continuous front lines drawn from Finland to the Caucas. This is nonsense, of course. The Germans never had enough troops for a continuous line like this, especially given the funnel shape of Russia, the further east they went. Nor did the Russians, for that matter. There were always large gaps in both their lines. Except possibly for the Western Front in WWI the concept of the continuous front line is more of useful map-myth than a reality. Besides, it has been proven in many modern wars, that defense in depth is more efficient anyhow.

So I wondered what would happen if I almost abolished the frontage %. I turned it down to 5%. Then I came across another setting that is of great importance and most useful both in controlling AI army size AND in getting the AI to do more research for it always seems to me that if you use the default setting the AI does not much research and one of the reasons you can usually beat it is you are fighting it with better weapon systems than it employs.

Setting No. 226, Research Investment, is a real winner. Since I had gone whole hog with Front % Occupation (No. 155), what if I did the same with 226. I set the AI to Spend 90% of its PP on research.

Lastly I set No. 182, Minimun stacking points, to a more sensible 50. And the results were: it worked! I got AI armies composed of good sized but not unwieldy units, in moderate rather than enormous numbers, with a most challenging Research developement (in one game one AI had Medium Tank IV by turn 8!).

I will answer some objections to this approach in a minute but I forgot to mention an important factor. My city size is usually 3. You can help to cut down on enemy army size of course by using smaller cities. And I play and tested all games with AI+.

Now, one criticism of this is that with 5-10% front coverage you, in effect, don't have a front. My answer is, yes you do, you just don't have a rigid, continuous one. I find, that, no matter what I do, the AI will produce sufficient units; you never have to worry about an AI unit drought unless you turn off the AI's ability to create units competely (No. 249). It will produce enough units so that you have a front that is in roughly checkerboard shape, in other words, a defense in depth. This is more efficient than the linear front because, if an opponent breaks the linear front, there had better be reserves to counterattack or chunks of the front may well get cut off from supply. But, with a defense in depth, the reserves are already there for sure.

Another quick way to reduce the number of AI units is to raise the PP cost of the initial production of units and HQ. This is a seperate setting field just above the AI editor in the list you got to the AI editor from. I now regularly raise the cost of creating a unit to 2 or 3 PP and a HQ to 7 or 8.

I will stop here but I'm not through yet. The reason I'm pouring all this out so much and so fast is I have limited access to the internet and I want to get in what I have to say before I run out. In next post I will talk about AI random personality, Air campaigns, and some useful but settings that are not as important as the ones I've discussed so far.

Oh, I have not used and so will not talk about naval campaigns as I cannot figure out how to increase the range of ships, which is so slow on the default I find I just don't have the patience to use it. Besides, I've had hands full with all these other things.

I would appreciate all comments, advice, criticism and corrections of what I have said, especially if others spot weaknesses in the inferences I'm making about what the Ai is doing out of sight of the turned off fog of war. Thanks.

rickier65
Posts: 14241
Joined: Thu Apr 20, 2000 8:00 am

RE: Explorations in the AI Editor

Post by rickier65 »



Miral -

Hey - thanks for the research and results! - and thanks for starting this thread! I haven't done much fiddling with the AI setting yet, but I'll have to make notes of these and combine with Vic and Webbers tutorials for future use!

Rick
miral
Posts: 170
Joined: Thu Dec 20, 2007 7:57 pm

RE: Explorations in the AI Editor

Post by miral »

Part III -

Here I want to say a bit about one of the most interesting and mysterious of the AI settings: No. 248, Switch on or off Ai random personalities. As I mentioned I did not know there were AI random person. until I got into the AI editor. Each AI is rated seperately for offensivemindedness, air campaingmindedness and researchmindedness. But, unfortunately, we are not told how they are rated or what the ranges of each is or, and this is of great importance, how this affects or trumpts any other AI setting changes we make.

For example, Setting No. 224. Raise the default and you get more air for the AIs, lower it and get less. But how does this interact with No. 248. Do all the AI's still get more if 224 is on but some less than others or does no. 248 overrule 224 completely?

In fact, in general, if you are making a number of AI edit changes, is it better to turn No 248 off for them to take effect or does 248 work hand in hand with them? Dont' know. Again, it would be most helpful if it was possible to switch a side back and forth from AI to Human within the same game. My impression, and it is only an impression, I won't go so far as to dignify it even with the name infererence, is that there is some interference of No. 224 with changing other settings. For example, and I have only about 3 games to go on, No. 156, increasing the % of support equipment in infantry units, seems to work better when No. 224 is turned off. In lieu of the developer telling us more about the AI editor's functions it will take a whole lot playing to begin to answer this.

And I called this function because it has tremendous influence on the game even if all the defaults are used but we know so little about it. For example, airmindedness is randomized. Does this mean the amount of air units, the agressiveness with which they attack and defend, the various kinds of air units, or all of these? How does this, or offensivemindedness (ofm) for that matter, interact with researchmindedness (rm), can't keep tyinp these loong words? We don't know. At least I don't.

If anyone has knowledge about this most important and mysterious factor I would appreciate greatly hearing it.

There, this one was much shorter; must be exhausting myself, and certainly anyone attempting to read all this. I have one more post to make on some of the minor but useful and interesting AI Editor changes you can make. Thanks.
miral
Posts: 170
Joined: Thu Dec 20, 2007 7:57 pm

RE: Explorations in the AI Editor

Post by miral »

Part IV (and last for today),

This post will concern various odds and ends of the AI editor and some of the lesser settings.

First, I am completely befuddled by the meaning of Nos. 153, 154 and their concommitants, nos. 164 and 165. I suspect that these are important settings but I have no idea what the word mobilizes and % of units the AI will mobilize in defensive or offensive situations. I put a seperate post up about this and got two interesting answers, but, unfortunately, I'm not sure either is correct. Would appreciate help.

If you use no. 182, minimun stack points in unit, be aware that it pairs with no. 247 and developer says the two values should be close or the same.
I suggest 40-60 as the optimal range for the stack points in unit setting.

If you change the % of front covered setting (no. 155) you should probably have it to cover a greater % the larger your map is. But, as I don't play on huge maps, this is just a guess based on my play on smaller ones.

No. 222 can be raised to have AI produce more Rear Area SFtypes to further diversify your AI units, but this interlocks with other rules as per my example of trying to produce huge units, so adjust all accordingly.

Many have complained about a single unit breaking through and destroying a lot of air units that are unguarded on an airfield. No. 251 can be set to guard air units better.

You can set defense in depth and size of behind the line garrisons with Nos. 252 and 254 but be aware that if you set the depth too deep it will suck up a lot of the AI's units.

No. 256, 'always defend large towns' is default set to 10,000 so the AI wont make much defense of towns and cities with less. This seems far too high. I always lower to 2000, less for scenarios where using low pop cities.

No. 225 - I actually consider this a very important setting but put it here because I think many will disagree. By upping it the AI will make more effort to destroy your units rather than capture your towns. Destroying units completely, especially veteran units high experience ratings is very important. This rule bears on one of the oldest arguments in warfare; should the enemy armies or cities be the prime target. Clauswitz says it depends on the particular situation and I agree, even if I tend more towards the destroying units side. This setting will depend greatly on the geography of your map.

I will experiment more with the settings I have talked about to see if my initial conclusions hold up and report back. Thanks.
Post Reply

Return to “Advanced Tactics Series”