Jump to content
Sign in to follow this  
Javaslinger

Are quick battles vs the AI worthwhile?

Recommended Posts

Not playing a human being is the primal thing for me. I can't stand humans who play wargames. I gave up playing wargames forty years ago because all I saw was old men arguing about what was or was not fair or realistic. Of course to a teenager the old men were probably all of thirty.

So PC wars gaming when I discovered it in the time of Sid Meier's Gettysburg was great. Good enough AI for an hour or two of history games and no bitching because my favourite tank hit a mine. Reload, replay have fun.

LOL yeah I'd say part of what took me so long to try hth play was concern that the folks I would be playing against would just annoy the crap out of me and that was partly driven by the way folks behaved on forums. I have to say though my experience with HTH play has totally spoiled me. My opponents across the board have been really really great, a real credit to our community. They have completely changed my perspective and in effect ruined solo play for me now (bastards). I still do play solo, but the fact that the AI can't react to my moves just drops the experience down a notch from the way I use to view it. Funny thing is it has also made me appreciate good scenario design even more. As an example I really loved the CMFI scenario Bad day at Beach Red as the AI plan almost felt like I was playing against a human the first time I played it.

Share this post


Link to post
Share on other sites
It's all to do with size... as always.

In this case it's that CM1 was almost operational in scale with a Regiment plus on each side on 4Kx8K maps.

I expect that CM2 will get there eventually. However, the complexity of CM2 mitigates it being as much fun and playable as CM1 is at that scale.

That may be true that you could do larger engagements, but larger doesn't neccessarily translate to fun. Cmx2 may never get there, but honestly I wouldn't mind. At the level of detail I am watching CMx2 games I am gradually coming to the conclusion that larger simply equates to making me miss the detail that is going on. CMx2 has made me appreciate small unit tactics more than CMx1 ever did. I have played a couple good size games lately both HTh and vs the AI and my feeling at this point is too large simply detracts. It's like going out for sushi - you can go to the place that serves large pieces of mediocre fish or you can go the place that serves small pieces of really high grade fish. I'll go with the small pieces any day of the week (and no before you go there large pieces of really good fish isn't generally going to happen- good sushi places aren't there with the idea you gorge yourself.)

Share this post


Link to post
Share on other sites

And no Womble, I'm not neglecting those other CM2 advances, and you are (as always on this forum) taking me out of context and yourself neglecting the overall thrust of my argument, which is about the impact for the solo-gamer.

[snip]

So for me, as a solo-gamer playing big WW2 battles, overall, CM2 has so far proved to be a backward step.

The reason I'm not commenting on those points is because they're personal opinion, taste, and no one can disagree. For you, CM2 isn't as good as the classic stuff. Fine.

Forgive me if I pick on something that actually can be discussed, rather than just mused over.

Share this post


Link to post
Share on other sites

Not sure whether playing a game can be equated to eating sushi. Many of us just like larger engagements that require some logistical issues, resupply, reserves, and most importantly, room for maneuver. The large majority of CMBN and CMFI are straight ahead frontal assaults, and that is like eating sushi ALL the time lol. I like red meat too!

My ideal would be to have CM similar to CLOSE COMBAT or ROME - TOTAL WAR where you can play on an operational map, allocating divisional formations etc. and then drill down to play tactical on what we now have in CM.

Share this post


Link to post
Share on other sites

The large majority of CMBN and CMFI are straight ahead frontal assaults, and that is like eating sushi ALL the time lol. I like red meat too!

This is true, but it's more of a map size issue than a force size issue. I don't have any trouble commanding battalion + sized forces in CMBN. Finding quality maps of a sufficient size is more challenging.

Share this post


Link to post
Share on other sites
Not sure whether playing a game can be equated to eating sushi.

Of course it can. Streety thinks, for his own reasons, which none of us can really gainsay (though the whys and wherefores of his coming to those reasons may be debatable), that x2 is a backward step. I do not. The single player experience is different, IMO, but still "better" overall. That's opinion, taste, peccadilloes, and trying to change either viewpoint is largely pointless.

Many of us just like larger engagements that require some logistical issues, resupply, reserves, and most importantly, room for maneuver. The large majority of CMBN and CMFI are straight ahead frontal assaults...

Again, a matter of taste and perception. I find plenty of room for maneuver on most maps. Part of the tactical problem is creating or identifying enemy flanks in an apparently homogenous wall of defenders. For me. In an MP game that's near completion, maneuver has pretty much won me the game, my reserve has fought on both flanks and I'm just resupplying some of the most heavily used squads with rifle ammo, having long ago emptied most of the vehicles of SMG ammo. This (apart from the winning bit) is not atypical IME.

My ideal would be to have CM similar to CLOSE COMBAT or ROME - TOTAL WAR where you can play on an operational map, allocating divisional formations etc. and then drill down to play tactical on what we now have in CM.

[shrug] Trouble with that approach is that if you fight the op layer right, you don't have many battles (ideally, none) you'd be interested to fight at the tactical layer. It would be a very fine balancing act between evenly matched human players that demanded "balanced" tactical battles, and there is a vanishingly slim chance that could be arranged in games against an AI.

Share this post


Link to post
Share on other sites

And to those in this thread who are essentially saying "well, get yourself a human opponent if you don't like the AI", I say that's fine if that's what you want, but I don't want, and I'm talking here about the solo-game and afterall, THIS THREAD IS ABOUT THE SOLO GAME!

Actually this thread IS NOT about the SOLO GAME!

This is the title of the thread:

Are quick battles vs the AI worthwhile?

This is the entirety of the original Post:

I know the playing the AI is just no comparison vs HTH, but I'm thinking quick battles must be particularly awful....

Thoughts?

This thread is about the possibility that Quick Battles are 'particularly' awful, probably due to 'inferior' AI.

The reasons for the AI being weak have been pretty clearly explained.

Playing with Humans is the only practical way to overcome the shortcoming of the AI, and will be for the foreseeable future.

If you refuse to play with humans, you refuse the only practical solution for improving the opposition quality.

I see very little room for complaint in that regard.

Your other complaints are largely personal taste, if you don't like it, do something else, or petition for change.

rotflmao sometimes you do have to state the obvious :D

Share this post


Link to post
Share on other sites

Streety,

Here's my answer to your AI concerns. Each section is preceded by a notional question.

"Chess AI is really good, so why isn't CMBN's?"

In relative terms, chess is trivial compared to ANY CM game. In chess, who can move what, where and how can be readily modeled and has now reached the point where I believe computer AI can get a draw against a Grand Master chess player.

The last time I looked (not recently), it took a powerful and specially programmed computer called Deep Blue to fight this mathematically elaborate battle. Chess AI can be loaded with every important chess game ever played, from which incredibly detailed If/Then logic can be created. "Oh, it's the XYZ gambit, from the (insert two Grand Masters) in 19XX. The proper counter to that is the LMN from (two more Grand Masters) in their epic clash in 18xx."

Armed with this depth and breadth of information, not only can the AI play chess, but it can employ predictive logic to look ahead several moves, thus, anticipating the opponent's moves, just as live Grand Masters do. The battle is essentially waged in 2D, with who can do what known a priori, thus bounding the problem. The number of possible moves is astronomic, but it's very much still finite.

"How does CMBN AI differ from chess AI?"

Even a simple CM QB scenario is a bear for AI to deal with. For starters, it's all 3D, a huge demand on processing just because of that. Nor are the paths of the participating units constrained as they are in chess. Then, there are dynamic LOS computations, sometimes for vehicles from both sides moving, in and out of terrain which blocks or degrades LOS. What is good for a movement path is not only a function of terrain, but weather, road state, cover, concealment, ground pressure, slope, morale state, fatigue, etc., etc. And that's just movement.

Where to move is not only a function of everything I just outlined, but known enemy positions, tactics and doctrine if known, a developing assessment of the IDed and likely opposition, military-technical characteristics and risk tolerance. In turn, that's the derivative of a host of hard and soft factors--all of which have to be either handled deterministically, by bounded probabilities, or as fuzzy logic. The outcomes of world historic battles can and do turn on something as simple as whether or not the commander slept the night before. Troop quality. Training. Fatigue. Hunger. Heat or cold. Thirst. All of these figure in the combat power equation.

The nature of warfare is such that what looks like suicide to one officer is a priceless opportunity to another. Rommel, for example, broke the British and with, say, 50 tanks, grotesquely outnumbered, drove on Cairo. He was stopped from taking the all but abandoned city by? His exhausted tankers passing out, utterly spent! So, how do you design AI that can tell the difference? This isn't pieces with printed combat values moving on hexes (stack those counters, roll 6 and wipe out one counter at 2:1 odds). How do you teach AI to a) recognize a threatening development and, if a), respond timely and effectively?

Ancient warfare isn't all that difficult to model (once at the battle imminent stage), the Civil War is doable, but more demanding. Jump to World War II, and the whole military analysis problem becomes not only exponentially more difficult, but exponents raised to a power. Weapon ranges are greater, weapon-target interactions vastly more complex, unit lethality much higher. Spacing is larger. The list goes on.

"What does the AI have to do as Attacker in a QB?"

Now, let's say the AI somehow can do all those wonderful things. (It can't). For a QB, the AI is tasked to field an appropriate force, so it has to have force selection logic, which has itself to be both smart and detailed. Those familiar with the extraordinary force optimization drills from CMx1 know exactly what I'm talking about. Kvetching about the AI's force choices has been with us from the beginning--with good reason. Precisely because of the insane levels of force optimization in CMx1, BFC made the problem much more tractable by forcing players to choose forces more typical of the unit organizations of the period.

Forces now in hand, it'd be nice if the AI could come up with a decent attack plan (the AI Holy Grail), but military planning is an enormously complex matter. In CM, many real world considerations aren't even modeled, yet attacking's still a nightmare. The AI has no memory, nor does it have, say, the military history of the world broken down as chess is. Nor does it learn. It simply does what its logic tells it to do, and that logic, sadly, isn't very smart.

It's one thing to push forward a phalanx, quite another to feint right, fix center and deliver a left hook, using infantry, armor and fire support in a coordinated, realistic and logically definable matter. And still another to dynamically assess and respond to what the enemy does. The AI has to figure out what to do, how to do it, when--while under fire. The AI has to constantly assess where it is, what it can see, what it can shoot at, maneuver options, whether to go for kills or drive on the objective. And what if the objective's occupied? Another set of computational demands. Of course, the larger and more complicated the forces, conditions, terrain variability, fire means and many more--the vastly worse become the demands on the AI.

"Is this all that needs consideration?"

Everything I've detailed above is by no means exhaustive, atop which here is an acute and fundamental resource problem. BFC is NOT Bungie. It doesn't have several million dollars and hundreds of programmers to create gaming magic. Last I checked, there are two whole programmers, who do ALL the programming tasks. If they're working on AI, they can't do something else, and there are production schedules to meet, playtesting, bug fixes, game patches for earlier releases and a plethora of other things they need to be doing.

"How does the AI in CMBN differ from the AI in CMBO?"

The more intricate the game, and I conservatively estimate CMx2 to be at least an order of magnitude greater than CMx1, the more pieces the poor AI not only has run, but also dynamically assess what's going on, what is likely to happen and what needs to happen in order to execute the combat task. So, if we could put the logical smarts in the CMBN AI (running at, say, 3 GHz on 4 GB RAM) into CMBO, which, recall could run on a 200 MHz computer w/o a graphics card, that might somewhat impress you, but the AI in CMBN simply isn't equal to what you'd like it to be able to do.Does it suck? No. Does it represent the best which could be had, given the sum of all the constraints I listed, I think so.

I hope this wasn't too long or exhausting to read, but you asked an exceedingly tough, nontrivial question. My answer fully reflects this.

Regards,

John Kettler

Share this post


Link to post
Share on other sites

I really appreciate your effort, John Kettler, but - FOR THE FOURTH TIME - I am not trying to equate a wargame computer AI *TODAY* with a chess game. I was making an analogy with chess computers of 30yrs ago and the advances in computer opponent AI since then. Please reread all my posts. And to prove the concept that such advances are at least plausible, CM1 did not provide too unreasonable an AI opponent, though it still had some weaknesses. And I'm saying that for the solo gamer its a shame that for CM2 the priority was to go to 1:1 scale and new graphics etc, rather than to improve on those remaining weaknesses of the AI-opponent of CM1 gameplay. And if they only have, as you say, two guys to do it, then my point persists in that they got their staffing priorities wrong.

Yes the tac-AI in CMBN attempts to be far more complex and I've not disputed that. What I'm saying is that for the solo gamer all these benefits are more than outweighed by the disadvantages of CM2 having a less independent enemy AI and (perhaps even more importantly for solo-gamer replayability) the doing away with truly random maps. And its a shame more couldn't have been done regards solo-campaigns. And even though they straight-jacket the CM2 computer enemy in this way, as an opponent it acts little better (or no better at all) than it did in CM1. And others on this thread have found the same issue. Which is why I posted on this thread to say the same. And this is why, along with all my previously posted reasons, I've said that on-balance, for the large-battle WW2 solo-gamer, CM2 is a backward step.

I've said all that I'm going to say, and I'm not going to rephrase them all yet again for the many who have not read my argument fully/properly. So I'm now done on this thread. But thanks to the few who have read it all and understood my thrust.

Share this post


Link to post
Share on other sites

Streety,

I'd crafted a reply to you minutes ago, but the computer burped, vanishing the post. I'll start out by apologizing for the chess AI redundancy, but this was because I was head down answering you from before. Let's see if I can respond cogently and succinctly.

I understand your point about the AI and am firmly convinced that with today's computational horsepower, memory and special processors, wonders could be worked if applied to CMx1. Were the AI given SOPs, battle drills, basic military principles, force coordination principles, the ability to set a flag to look again for a target previously sighted or engaged, better threat engagement prioritization (obviously has some, given how it goes after antitank guns and antitank teams) and a few other things, it'd be something to reckon with.

It would also be nice to see a range of commander traits be selectable: bold, by the book, timid for starters. Stress reaction: recoils and goes doggo, rattles but regroups, unfazed by whatever comes. Casualty tolerance (combination of strategic, operational and tactical factors). Replacements. Materiel and ability to replace it. Reinforcements available. Instructions from higher commander: ironclad (seize at all costs Objective X by time Y, using the following approach and units; Auftragstaktik (from this asset pool, figure out a way to take Objective A by Z time; Lee at Gettysburg re Little Round Top ("Take, if practicable, Hill B"); prepared attack "Tell me what you need to guarantee the seizure of Objective C in order to provide a jumping off point for the battalion at time W").

These are some ideas I have to show what might be possible. I believe that an AI with such features could potentially be something of a terror, particularly since the AI never daydreams, doesn't get sick, isn't afraid, never has an off day or can't think on its feet because it got no sleep and hasn't bathed in weeks.

Unfortunately, this is a thought model--unless and until BFC releases the code and a host of brave souls, possessed of impressive coding skills, takes a shot using what I've set forth or better to develop a formidable AI.

Regards,

John Kettler

Share this post


Link to post
Share on other sites
Are quick battles vs the AI worthwhile?

No.

But in CMx1,they are.

Now I will disagreee with this to some extent.

First, If CMx1 AI can give you a good challenge, You either have to know how to tweek it to make the odds pretty unfavorable or you are just not a very good player. Because I found the AI doing terrible things with its troops in movement, thus the reason it was only semi good on defence. But even on defence it had a tendancy to want to move its troops from good placement locations to stupid positions.

Now , just for giggles. I played a QB vs the CMX2 AI last night just to see if it would be as terrible as I like to point out it is.

Without spending a ton of time to set up and play, could I get a decent battle.

So of course I selected to be the attacker, since the AI does better on defence. I selected the map and I let the AI pick its forces and I selected my forces and gave myself no arty to help improve the challenge and a nice short time frame so that i could not take my sweet time disecting its non- mobile defence.

For armor I selected a company of cronwells, that should make it interesting and a company of infantry minus about a platoon & the heavy support weapons. So I had a few light mortars plus infantry.

I can gareentee you I had a better battle vs the AI for the first 20-25 minutes, playing Real time on elite than I likely would ever get from the old system.

First the enemy was not on the objective, it had made a good battle line between it and me. Second it held its position well and even at one point in the game counter attacked with a stug to get itself in a flanking position to the other flank in which i was pushing.

I lost two tanks before I could even get a spot on a enemy location. (Surely something that would never happen in he old game). Which included taking out my one and only firefly.

I lost two more tanks one to a unseen unit as i was pushing the flank and the last was a cronwell that I rolled down the enemy flank once I had control of the battle, he was trying to mop up and take out any remaining armor that might be on the battle line. I knew a stug would have its flank exposed to it. As I moved it in position to take it out. I opened the hatch to take a better spot and My commander was killed by a hidden infantry unit i had missed in a area i had mopped. The tank lost command control so i watched helplessly as it rolled into a spot that likely could be spotted by the stug and was and watched my tank go into flames.

In the end. I lost 3 cronwells and the Sherman, The AI lost 4 Stugs and a Pv4.

It did not have any infantry to say, just a few scouts that I located and keep me busy early on and a FO with arty that tried to hit me with 3 fire missions within the game.

Now if I knew how (which I dont) give the enemy more units than the preselected point value. Because i had a 2-1 advantage almost in that game. So if I could have tweeked it to give it 50% more units, it might have been a real challenge. Even so, I did not win the battle, I played a battle that the outcome was given since I had the advantage and it would show. I felt the AI played me pretty even and for its limitations, thats not bad.

The things that gave it the ability to do that to me was the spotting and fog of war aspect that is within the CMX2 engine that you will never see in CMX1.

Either you like it or you dont. But it adds the unpredictability of battle (Of course i played the game in a way to make sure that the enemy might get the game engine advantages of that part of the engine)

But I set up played and finished the game in about a hour and a half.

So even with all the complaining, I will still point out, it is a game, learn to use the tools and you might find it is useable after all.

I do smile how we all love to complain about things the game lacks and what it should have, but it is up to us to make the most of what we have also.

Share this post


Link to post
Share on other sites

The more I mess with the QB feature vs the AI. I will totally disagree with the comments that it is worse than CMX1

This is not how i play the game generally, but this thread made me wonder how well is the AI working in QB's since the first few I tried the set ups were junk and it did not perform relistically at all, back when the game was first released.

Well, I created two more battles. First one I selcted a probe for the attacker at night and decreased my force by 30%, figuring it would reduce my advantage as the attacker to very slim. I selected my forces and let the AI auto select from mixed. It was a town med battle and it played very well.

Map was ok, nothing to write home about but better than a random generated one.

I had a company of engineers and 2 churchills. The defence had infantry, plenty of armor cars and a Stug.

Again it held a battle line and multible times it acted like it was counter attacking, either to plug a hole in its line or to swing up on my flank.

It gave me a decent fight. nothing too challenging. but it reacted much better than the older games.

I played a second game in CMFI, this time as the defender. I set the attacker in a assault game to receive additional forces, I think 10% and another medium size map, these were played on elite.

This time I was plesently suprised. The map was fantastic. Felt close to any scenario map you might come across. I was way outnumbered, I had a company of Armored infantry, which had plenty of AT guns, I had some 81 off board mortars and 2 Shermans to use to help stop any breaches.

The enemy had approx. 300 infantry, again a ton of armored cars, A Stug and a Marder and also some off board Arty.

I assume that these maps might have some human programmed AI battle plans. But what I saw was the AI attack, split its forces and attack in two locations. Thus forcing me to play somewhat honest and making me hold my troops across my front. it forced me to move units to secure the breaches it made at the two concetrated attacks and caused me to thin my lines below decent levels. Now it is not human and if it was it could have adjusted its units at the breach points and overrun me, but being a machine following simple orders it had a tendancy to keep moving troops to the same locations. But once in awhile a few would get off that path and create a little excitement.

but again, though nothing fancy and it appears it is not going to do more than just its programmed paths. It still is for sure not worse than what we had before and given a little battleplan human programming. could make for some good challenges,

Note: I played these both in one evening. I set the game for 45 minutes, real time and spent some time to select my Units.

I will rate my games as a 7 out of 10 compared to maybe a 5 out of 10 for how something like this played in CMX1

Share this post


Link to post
Share on other sites

I assume that these maps might have some human programmed AI battle plans.

AIUI, to qualify as a QB map, there have to be at least two plans: one for each side. Otherwise the AI's forces will just get dumped in the setup zone and won't know where to go.

Share this post


Link to post
Share on other sites
I really appreciate your effort, John Kettler, but - FOR THE FOURTH TIME - I am not trying to equate a wargame computer AI *TODAY* with a chess game. I was making an analogy with chess computers of 30yrs ago and the advances in computer opponent AI since then.

This is true, but it's not true for computer games generally. No computer game has a meaningfully better AI than it did 15-20 years ago. The problem is not the power of the computer; the problem is that no one really knows how to make a computer think like a human.

Chess programs got better not because computers got better, but because programmers got better at explaining the rules of chess to the computer. None of this has much application to actual AI, which is why Big Blue wasn't really a success from that an AI POV.

Even if BF had 100 times the staff and budget, they couldn't make the AI better. Billion dollar AIs in robots have problems with fundamental things like telling a window from a door - they are just not very smart.

Please reread all my posts. And to prove the concept that such advances are at least plausible, CM1 did not provide too unreasonable an AI opponent, though it still had some weaknesses. And I'm saying that for the solo gamer its a shame that for CM2 the priority was to go to 1:1 scale and new graphics etc, rather than to improve on those remaining weaknesses of the AI-opponent of CM1 gameplay. And if they only have, as you say, two guys to do it, then my point persists in that they got their staffing priorities wrong.

They could not have improved the AI. No one could have.

Yes the tac-AI in CMBN attempts to be far more complex and I've not disputed that. What I'm saying is that for the solo gamer all these benefits are more than outweighed by the disadvantages of CM2 having a less independent enemy AI and (perhaps even more importantly for solo-gamer replayability) the doing away with truly random maps. And its a shame more couldn't have been done regards solo-campaigns. And even though they straight-jacket the CM2 computer enemy in this way, as an opponent it acts little better (or no better at all) than it did in CM1. And others on this thread have found the same issue. Which is why I posted on this thread to say the same. And this is why, along with all my previously posted reasons, I've said that on-balance, for the large-battle WW2 solo-gamer, CM2 is a backward step.

I've said all that I'm going to say, and I'm not going to rephrase them all yet again for the many who have not read my argument fully/properly. So I'm now done on this thread. But thanks to the few who have read it all and understood my thrust.

For the record, I don't think that CMx2 is a backward step at all. The details is much better, and the AI in scenarios is also much better, due to triggers. The troops in CMx1 never counterattacked or set traps, for example.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

×
×
  • Create New...