Winning battles a little too easy

[pipsqueak]

I'm not a great general or tactician but I am winning all battles. Even when outnumbered by 20K. And with losses that should be unsustainable for the enemy. My losses tend to be 2.5K and the enemy about 20K. These apparently are "minor" victories.

The AI needs to do much better. It is too easy now to beat it every time. The AI is easy to flank and dislodge from fortifications. The AI attacks piece by piece even when it outnumbers me greatly.

I am loving the game but the game will have limited longer term repeat play when the AI cannot win one single battle.

Here's hoping the AI is given some kind of help to make things more interesting in the future.

My 2p worth...

[Fred Sanford]

In the end, that will be THE criteria by which the game is judged.

[Ormagni]

yeah, the AI can be a little screwy. but I will say it has improved TREMENDOUSLY since the game first released. so as long as the trend of improvement continues, I'm still optimistic.

To increase the challenge, for myself anyway, I'll create my own unique challenges that I have to adhere by. For example, I'm currently running a union game where my goal is to always maintain a surplus economy and never take out a loan. it's definitely hard. I'm trying to maintain my boarded with less than 40K men across the entire front. because equipping troops just gets too expensive.

[Duck]

Originally posted by pipsqueak:

I'm not a great general or tactician but I am winning all battles. Even when outnumbered by 20K. And with losses that should be unsustainable for the enemy. My losses tend to be 2.5K and the enemy about 20K. These apparently are "minor" victories.

The AI needs to do much better. It is too easy now to beat it every time. The AI is easy to flank and dislodge from fortifications. The AI attacks piece by piece even when it outnumbers me greatly.

I am loving the game but the game will have limited longer term repeat play when the AI cannot win one single battle.

Here's hoping the AI is given some kind of help to make things more interesting in the future.

My 2p worth...

I have noticed casualties get skewed like what you mentioned when you are playing defensively as the CSA and/or are abusing skirmishers. Try switching to Union?

[bylandt11]

I think the AI is not bad at all. To program it for such complexity in battle (it's not chess) must be a nightmare. But yes, it is inadequate against humans. It should learn to attack more concentrated. And perhaps give it a few cheats like less fatigue or better vision of the enemy positions. My own house rule is that I ignore VP locations.

The problem is that the campaign game is decided in these land battles that the player plays out himself. The AI can survive the casualties or the loss of territory, but not the hits to national morale. I just played the CSA on very hard. Union surrendered in less than 5 months, on 2nd of november 1861, from sheer morale loss from battles.

[Scotters]

I would like an option where we can turn off fog of war, feuds, and delayed orders for the AI, while all of that would still apply to us.

[rogersdrums]

Yes the ai will need some kind of "perks" in the tactical battle set ups.....maybe give it the ability to flank easier...or make it much more difficult for the human to flank the ai.....even though some have mentioned giving advantages to the ai is too gamey,remember giving bonuses in the initial game set up to the ai opponent is gamey also....the ai does need some tactical battle advantages....Ive noticed that the ai many times when attacking just cant seem to attack en force....so many times the ai sends some of its brigades on suicide missions separated from the main force...and sometimes attacks in waves.....and in second day/etc. of battle set ups the ai sometimes does not set up intelligently....i recently had a large battle whereas the second day of battle the ai forces set up in a nondescript blob right in front of my main defensive line.....I just want to see this game get to a point where the human participant isnt sure of victory in almost every remotely close odds battle....right now the tactical battle aspect doesnt play that way most of the time in my experiences...the ai has come along way for sure....all credit to the dedicated developers...but most everyone here knows its way too easy to defeat the ai opponent in tactical battles.

[Basilhare]

Originally posted by rogersdrums:

...the ai does need some tactical battle advantages....Ive noticed that the ai many times when attacking just cant seem to attack en force....so many times the ai sends some of its brigades on suicide missions separated from the main force...and sometimes attacks in waves....

This.

The AI simply needs to use its entire force vs. piecemeal attacks...it will throw 25% of its force at you so that you can always defeat it...

[pipsqueak]

I have noticed casualties get skewed like what you mentioned when you are playing defensively as the CSA and/or are abusing skirmishers. Try switching to Union?

This is playing the Union and not with skirmishers. This is using lines of infantry and cannons. And flanking infantry.

[pipsqueak]

Originally posted by Ormagni:

To increase the challenge, for myself anyway, I'll create my own unique challenges that I have to adhere by.

I have started doing something like this too. Role-playing limits... I will try not to outflank the enemy when they have built fortifications. I try not to out flank the enemy from behind. I try to not constantly kill all the AI artillery the AI always insisting on driving out front and alone like scouts. All this really just delays the inevitable.

AI tactical battles are beyond the developers, and most games. But maybe, as already been suggested, perks and cheats could be given to the AI? Make human units slightly less morale and quicker to break. Quicker to tire. Less accurate. Artillery less morale impacting for the AI. And so on. Small buffs here and there. Lots of them. But small ones.

[rogersdrums]

Originally posted by Scotters:

I would like an option where we can turn off fog of war, feuds, and delayed orders for the AI, while all of that would still apply to us.

All stellar ideas to aid the ai.

[Michaelsmithe]

While i hope to see the Battle Ai improve, which i believe it can. I also use this point as reasoning to why Multiplayer is essential to this game. Even a 1v1 style multiplayer scenario is needed. It adds another human that is trying to win and can out maneuver you in many ways.

I hope the Development team, will complete 1.0, successfully get singleplayer released and what not, then maybe start looking into the prospects of multiplayer. The game is setup perfetly for it, it would work well and hell if they need beta testers for it, i'm sure there are a thousand people ready to sign up and test it out with their friends.

[Duck]

Originally posted by pipsqueak:

I am loving the game but the game will have limited longer term repeat play when the AI cannot win one single battle.

but i mean in reality, isnt that the case with most games featuring AI? For example, I played UG:CW for like almost 1500 hours and the AI won like one battle, total. In this game, I have occasionally found myself outnumbered 2:1 and the AI makes a concerted attack and I lose. Not often, but more so than other games...

[meghostryder]

Originally posted by Duck:

Originally posted by pipsqueak:

I am loving the game but the game will have limited longer term repeat play when the AI cannot win one single battle.

but i mean in reality, isnt that the case with most games featuring AI? For example, I played UG:CW for like almost 1500 hours and the AI won like one battle, total. In this game, I have occasionally found myself outnumbered 2:1 and the AI makes a concerted attack and I lose. Not often, but more so than other games...

this thread is amazing and is a full example of how the general gaming audience has been hoodwinked into accepting half baked attempts of A.I. in games. It even suggests ways to let the A.I. cheat to get better!

Good A.I. is possible. It's not even hard if the programmer is competent. But it does take time. Significant time.

since we as players don't demand it developers put little effort towards it. But don't be fooled. This game could have good A.I. that plays by the same rules you do and repeatedly beat you. It's not gonna happen in 14 days (the average I've always been allotted when I worked in this industry) but if we started demanding it before money left our wallets we'd see it.

[Commander Cromwell]

I agree that it would be better to have a fully fleshed out and competent AI than an AI that has "perks". I don't recall any "tactical/operational game" that has ever really provided a challenge unless heavily out-numbered or because of historical circumstances.

[meghostryder]

You right. I can't recall one either. but that is because, again- no one had to. Worse those type of games are niche games with small teams with small budgets. they're also likely limited to a cheap engine. I've worked on such teams. Outdated obscure cheap engines that use some restricting scripting language and even worse an IDE that restricts what GUI you can have and so on.

It is no accident these games are known for bad GUI's and bad A.I. Worse the majority of the developers are either older or self taught. The older ones went to college (like myself) long before compiled languages like C# or C++ were a thing. We were taught in assembly.

The self taught generally also skips the compiled languages and gravitates to simpler scripting languages.

Even in today's world if you go for a software engineering job it's highly unlikely any of your courses deal with A.I. programming in any meaningful way. Their goal (and most of the students) is to land a job with starting wages in the 6 figures. This would be companies Like MISYS who do medical software and have entry level wages near 100k.

No one pays near that in the gaming industry. If you had an idea and a good tech demo you could maybe sell the idea to a publisher who'd fund you or start a kick starter campaign-- but you better have the skill and knowledge to pull your idea off. Your still going to have a hard sell on the A.I. that will take not days (The norm) or even weeks but rather months.

Now I've seen specialized courses offered aimed at game development. Nearly all these focus on getting you started in different genres (FPS, RPG, Strategy etc.) and GUI design but none really dive in deep in any one field or A.I. in general.

The programmer has to devote his own time to this and learn from available sources. There are many. Afyerwards then you have to find an engine that supports a full language with libraries that can do the A.I. as well as the game that you can afford.

If you do that you need do it but once as your developed A.I. can be used in all future titles you do using your design.

Or in short they dev one for this game and can easily use it in other titles using this basic design be it WWI or WWII or whatever. The point being the A.I. and team will get noticed and garner a lead over all other competition. They will be the 'go to' for strategy war games.

And although you can't name a strategy game I can name a FPS that did this. F.E.A.R.

That games as far as FPS's wasn't special. What WAS special was the A.I. Up to then FPS's generally used FSM's as an A.I. (Finite State Machine). Basically a simple sit til they see you and then change their 'state' to attack. Low overhead and can be done with a low overhead scripting language like LUA. this was necessary as consoles only had at best 256k of memory. Your not running c# on a console.

But F.E.A.R. was a computer game and the programmer (Jeff Orkin) used a full language to produce an A.I. that communicated with fellow guards, took cover, formed plans, surrounded and ambushed you and so on. He called the system GOAP (Goal-Oriented Action Planning).

The title was later ported to consoles by a hired third party minus the A.I. and did rather poorly.

However if such a system was used here the A.I. would not need to cheat to win. I could go into more if you want but it would require a short course of different types of ways A.I. can be done and why certain types are superior to others.