Matrix Games Forums

There have been some threads grumbling about the AI in COGEE here recently so I thought that starting a thread like this might be apropos.

I, like many players, find the AI in War in the Pacific to be pretty pathetic (hopefully the Admiral's Edition AI will not be so bad, but that is a tangent . . .), and I started up this thread Helping the AI to Challenge Me in the WiTP forums to get some feedback.

Here is one one guy, "R8J" suggested

Three day turns.
Do not bomb any airfields. It's to easy to destroy the AI airforce on the ground.
All units involved in an invasion (air/sea) must be preped to 100.
Max aircraft at an airfield is size x 50.
All ships with sys damage 20 or higher must return to base and can not sail until in the green.
No air missions unless morale 60 or higher.
LCUs can not be committed to new combat until fatigue/disruption is down to around 0/0.
Units must be assigned to HQ before being comitted to combat.
All convoys must have an escort ratio of 5:1 (5 transports: 1 escort).
Do not upgrade aircraft until pool is exhaused.
Stick to above rules!

Now if you don't know WiTP, some of that might be 'Greek' to you, and obviously very little of those suggestions are going to be directly applicable to a very different game like COGEE.

But the general point I want to make here: we all need to be realistic about what we get for $50 or $60 in a computer game. Expecting the AI to be as 'smart' as your 10 year old brother is simply not realistic. AI in _most_ games is 'stupid' and only manages to challenge accomplished players with advantages, and/or having information that the human is not privy to.

We are privileged to live in an age in which we have incredible access to information, incredible game graphics, and even simple decision-action algorithms in the form of computer-game "AIs," but calling these things "artificial intelligence" is really a total misnomer.

There is not a single game out there today (with the possible exception of some of the most advanced chess programs) that arguably really reaches a sufficient level of agency and adaptability that is appropriately referred to as 'intelligent.' In the case of the chess programs, you have to keep in mind that, they are dealing with a mathematically VERY simple decision-space. The board is always the same, as are all the pieces, there are a large number of permutations of pieces on the board, and with diminishing numbers of pieces even more, but even so: the complexity of the decision space in which a chess app has to operate is nothing compared to a game like COGEE, or most contemporary strategy games (e.g., Civ, EU, etc.).

Until there are major advances in the AI coding world, and/or in the hardware we use to run these games, we need to be realistic and play appropriately. Criticizing gets no one anywhere; but if you are open-minded, perhaps a bit creative, and mature, you CAN have fun in a computer game that you overall love, but which is nonetheless limited by having a less than brilliant AI. For me WiTP fits that description; I love the game and the chance to play the role of supreme commander of the Pacific theatre is just too engrossing, no matter how pathetic is the AI (and I am not a big PBEM fan). But to really enjoy myself now that I have more-or-less mastered the game, I MUST impose self-rules, or constraints on my own actions within the game in order to not simply mopup the AI.

Though I think the AI in COGEE is considerably better than the one in WiTP, similar principles apply here: Once you have 'figured out' the AI in one of these games, if you really want to maximize your fun, you might want to consider making an active and devout point of NOT taking advantage of your insights, your ability to be creative, your capacity to exploit the poor weakling computer. This doesn't mean you have to 'play stupid,' but simply avoiding doing things that border on unrealistic can often be enough.

Some ideas for what you can do to help the AI in COGEE challenge you? I started this thread for us to discuss/list/enumerate the myriad of answers to that question.

To start, I will quote Russian Guard from the Do You Often Lose Battles thread.

ORIGINAL: Russian Guard

Another thing relating to how well the AI plays is whether you try to recreate "historical" type battles or not, in terms of how you move and deploy your units.

In my experience if you deploy your troops like a traditional Napoleonic battle (center, flanks, etc), then the AI can give you a real good whipping sometimes, especially if you do not play France.

Alternatively, if you "take advantage" of the AI by knowing things it will fall for and moving units around in a fashion that isn't necessarily "Napoleonic", then its easier to beat the AI.

No criticism intended or inferred either way; players play as they will, which is as it should be.

While RG refers to tactical matters of how you deploy, move and use your troops in the detailed battles, the same principles could be applied to strategic level actions, building, diplomacy and the balance of naval and land strategy too.

In short, if you play a game 'in style' for the period, leader, character, nation you are portraying, you are almost always more likely to achieve less ahistorical results, or results with which the AI cannot cope. If you play primarily to whup up on the AI, then you shouldn't be surprised if that is exactly what you accomplish!

So what other methods can you guys propose or suggest for helping the AI in CoGEE to challenge you?

Great post!
Hmm the thing that would make an AI game more challenging for me...
-not building some of the buildings, espesially barracks and bank above lvl 3 or 4 (to keep money and army research down)
-not allowing myself more than 1 arty per army
-only play QC/IC when the AI has inferior numbers
-not leave my borders empty just because I am at war with only one nation
-keep a mixture of around 60:40 in INF/Cav to avoid my INF chewing up the AI
-limit my reinforcement troops to 3.5 morale instead of 4.5

Terje

For naval detailed combat:
- Only attempt close melee fleet actions as England
- Don't grapple and board unless the target ship is close to striking
- Don't sail more that a couple of frigates in fleet containers
- Maintaining the integrity of the line-of-battle even if it means missing opportunities for decisive victory (unless playing as England)
 
For land detailed combat:
- Impose formation integrity restrictions into all movement and combat
- Give formations tasks based on their position in the deployment rather than based on mere unit strength or special abilities (IE, if an understrength Corps is on the flank and a flank attack is required, that Corps in position gets the job although other Corps might provide support
- Artillery units always get infantry support even when the infantry might be better used elsewhere
 
For the stategic game I try to run the economy based on my (admittedly limited) understanding of the nation I'm playing.  Therefore as Russia or Turkey for example, I don't change the feudal levy (taken as implementing some sort social reforms) and the like...
 
As England I do not build walls or guns in home provinces.
 
Best Regards

If war has not occurred with a nation that historically your nation did war with declare it.
Do not make friendly diplomatic agreements with nations that in real history your nation was enemies with.
Do not offer protectorate status to far away provinces.
 

Well, the obvious one is to change the victory conditions to something more challenging. That's what the glory modifier is for.

ORIGINAL: Joram

Well, the obvious one is to change the victory conditions to something more challenging. That's what the glory modifier is for.

Indeed, that will make ultimate 'winning' of the game more difficult, or at minimum more involved in terms of time and effort. But the slightly different issue that is the focus here is: playing so that the AI, with all its limitations, will be more challenging on a moment to moment basis. Having to get a higher score does not necessarily make the AI more challenging, though you would probably have to beat up on him for a longer period of time.

Obviously, upping the difficult (i.e., 'human handicap') settings is one way to make the AI more challenging. But the idea is that self-imposing "Helping the AI" rules of play style are another way that I think we all too often tend to overlook when we play these games.

Me personally, I've _never_ enjoyed playing the AI at really high levels. It just feels silly when I'm using _all_ my creativity, ingenuity and adaptability (and consequently playing in HIGHLY unrealistic, ahistorical ways) to face an AI that has astronomically better resources as a result of the various bonuses that get passed off as 'difficulty.' I agree with PBEM advocates that human opponents are the best; but it is just not very convenient, it takes way to long for it to be fun (for me), and it is fraught with problems. Not to mention the tactical battles remain undoable for now. So the spirit is to discuss or figure out ways to play so that the AI is more challenging even on the intermediate levels of human handicap.

As a general example: once we Grogs have played a particular game for 15 or 20 hours, I wager we have developed a pretty strong intuition (if not explicit paradigm) for how the AI behaves. This puts us into a position to 'try things' to further explore the limitations, biases, and predispositions in the AIs decision-making. With this knowledge, we have 'learned' (something the AI never does) about strategic and tactical opportunities afforded by the AIs 'psychology.' If we then begin to take slightly different in-game actions to capitalize on this knowledge, we are effectively exploiting the AIs limitations. As RG said, this is 'as it should be,' every player playing his own way. 'Figuring out' and crushing the AI is fun, and an integral part of mastering and eventually getting bored with any given game.

The idea here is to point out that: there is also MORE than that too be explored. You can also explore the ecology simulated in the game in other perhaps deeper ways by (perhaps in matches subsequent to crushing the AI) adopting a somewhat blindered 'Helping the AI' play style.

Careful consideration of these topics could even contribute to future insights into how to make slight changes to AI decision-making algorithms so that they in part simulate these human-helping AI self-restrictions . . .

I agree, I'm not a big fan of upping AI bonuses either though I understand why it's done. But that's why I suggest changing the victory conditions a bit. Yes, you are still playing the same AI but the challenge is still to beat it before it beats you. If you bias the victory conditions, at least you won't feel cheated which you might if you gave the AI a million bucks, or doubled it's strength or something like that. In a way it's the same thing but it is an option to make it more challenging in a manner that might suit some people better.

Goodaye,

Hey, well said.

Cheers,
Lancer

Try a Path of Napoleon game.  You actually have to exceed Napoleons historical accomplishments and its quite difficult to do it all in the time you are given.  Some of the achievements require several wars (with the same power, so 18 mo enforced peaces between wars) and you dont have much idle time at all, you have to keep pushing towards your ends or you will find yourself out of time.  Also the multiple front nature of the goals will require careful strategic focus and good fundamentals re logistics, economy of force, etc.

I am almost done with one, been playing a turn here or there in my spare time, at this point I have huge detailed battles every other turn, had to replay all of 1810 due to a weird little cede province bug that popped up (I believe being addressed in patch?) that made it impossible for Austria to rightly cede me a target province, but other than that has been a hell of a game.

(Moved thread to the War Room so that it will be easier for future players to find.)

This suggestion deals with game/AI mechanics.

When defending in detailed battle you should take control of any garrison unit that is attracting the AI's attention and unduly distracting it from its offensive. For instance, it dithers endlessly around the 'watchtower' hexes that are garrisoned, often taking several days to ring it on all sides before attacking it. By taking control you can, either, attack with it, or run with it, and let the AI get back to its planning.

Likewise, remove any garrisons in the key 'Star' hexes in villages and use your own troops for any such garrisons you intend to maintain. This keeps the AI from diverting extra troops to capture garrisoned hexes you don't intend to personally defend and let's it get on with its advance.

Don't use the 'Farther Start' advantage. The AI typically takes most of two days just to come into contact under this handicap.

I dont know how to take control of garrison units.

When they are fortified you hold your cursor over them and hit a key. Think it is e but it might be w. Can't recall for sure; it was different in FoF. Just try (slowly) a couple different keys.

Just hold the cursor above the unit and hit the 'U' key.