Is this game not best played in 3 minute rounds?

[blackcloud6]

Soviet doctrine of the time was to attack in mass support by lots of artillery. Give the AI mass and watch what it will do to your human NATO force. And make sure you give it a good amount of recon forces too. That was also Soviet doctrine: leading with robust combat recon patrols (that is why in the troops selection portion there are tanks platoons in the recon section.)

The Soviets were going to attack in echelon knowing that the first echelon will take heavy losses but cause enough attrition to make the second echelon able to make the breach in NATO lines and third echelon to exploit the breach.

[Phoenix100]

Thanks, blackcoud6. And whilst doing all that they made no attempts whatsoever to use covered approach routes etc? They simply charged out and took needless losses? Really? Why not use the obvious covered approaches, if they're there? In the game above the recon elements had already spotted enough to realise, I would have thought, that for the follow on elements to charge across open space as they are doing they will take completely unecessary losses. In the game above my forces are Finnish, not Nato, by the way, using, I guess, mostly Soviet kit. I lost, of course, because I gave the AI a 30% force advantage, because I've learned that if I don't do that then I will win easily by simply using ambush tactics, even playing as Finns with the same - or less good - kit. Fair enough - you might think - the AI attacks with odds in its favour (as you should) and wins, so what's the issue? Well, that attack looks like WW1, no? And I was actually thinking that maybe the AI right now would be good for WW1, or even Napoleonic combat?

Of course, if what you are saying is that, yes, really, the soviets did plan to attack just like that, then I take it all back. BUt I'm guessing that if I switch sides then the US AI will do the same, no? Will have to try that.

[blackcloud6]

Well, the only evidence of actual Soviet doctrine used in combat was the Arab Israeli wars and in 1973 in the Golan, the Syrians attacked in mass across open terrain. They got hammered but almost broke through. The Egyptian got across the canal and achieved their objective of planting themselves in the Sinai.

Looking at your map example, the Red forces don't have much of a choice but to cross the open ground at the road. So the AI is going to do so. Yes, it could try to go through the southern woods, and it might if you replay it. But then, I would have that covered by a TRP and make it real slow go for them...plus they couldn't put their whole force there anyhow. And it looks like the AI tried to use the more covered approach in the north and got hammered but used the center and had more success.

Please understand, that in modern combat, the attacker is always going to take casualties as there systems are too lethal. Numbers may well matter. This is why the Sherman tank in WWII has an apparent bad rap due to losses, but it was on the offensive for almost all of its combat life. The attacker has to move, exposed at time against well emplaced, dug in and hidden defenders who usually know the terrain better.

The only thin the game might add for a scenario designer is "pace of attack" where the AI is told to go a certain speed. A lower speed may allow them to stop and find cover and fight which sounds like what you really want it to do.


Oh, and by the way, if you going to design and play scenario where the player is the Soviet, or Soviet trained and equipped forces, I would say keep most of the forces as company sized units as this will force the player to try to stick to Soviet doctrine.

[CapnDarwin]

Phoenix, in the picture above, granted it's hard to see elevation, how would you have moved your forces forward through that open gap knowing that you are in enemy contact? The only issue I see, is maybe a lack of smoke and HE being dropped to take away some of the obvious Blue fire lanes/positions.

[blackcloud6]

"Thanks, blackcoud6. And whilst doing all that they made no attempts whatsoever to use covered approach routes etc? They simply charged out and took needless losses? Really?"

The Soviets during the Cold War were about mass and speed. The idea was to pressure NATO across the line with quick drives to break the line knowing that somewhere they will get through. The would stay in March Formation as long as possible. With just 60 miles to Frankfurt, the idea was to "go for the Rhine" and try to do so before NATO could dig in and reinforce. Losses be damned...

NATO doctrine actually was to force the WP forces to deploy out of march and thus slow their pace, channel them into kill zones and hammer them with airtlley, air and direct fires. Then when weakened, NATO would counterattack that weakness in Corps strength.

[CapnDarwin]

@blackcloud6, how much time or effort would the Soviets invested in city fighting in the opening few weeks of WW3?

[kevinkins]

the game might add for a scenario designer is "pace of attack"

I was thinking along similar lines using rudimentary waypoints. I remember a game from the 90's called "Wargame Construction Set - Tanks" and "Age of Rifles" used objectives to coax the AI along the map. Not a highly, but a very lightly scripted AI. I believe the objectives were very simple i.e. attack, defend and hold. If I recall, the designer could release formations from the hold objective after a certain length of time. This might work well to sim a Soviet style attack. I think a credible bounding overwatch could be programed. If anyone has the manual, I would love to confirm my memory.

[Veitikka]

ORIGINAL: Phoenix100

Picture below. All I would ask is would that happen in real life? Really? I'm no military expert, so I don't know. Maybe soviet doctrine of this period still included the 'human wave' type attack. If I were the AI in this case, though, I would be approaching much more cautiously, using covered lines of approach. I'm afraid seeing this kind of thing is quite off-putting. Maybe others are not so bothered, in which case I'd be interested to know why.

Perhaps if you zoomed it out so we could see the big picture, and also terrain elevation, we could perhaps understand why the AI picked that approach.

If you mean that these mech companies should have advanced in forest, well, in many cases that's definitely not the best option. In forest the vehicles can be ambushed or immobilized and move much slower than in open, or when following a road.

[Veitikka]

ORIGINAL: Phoenix100

And whilst doing all that they made no attempts whatsoever to use covered approach routes etc?

If the AI opponent has the 'infantry' force type selected then it favors covered routes.

[Phoenix100]

Mmm. Maybe I'm being unfair. It's true that the AI tried approaches (with armoured recon, I think - tanks, at any rate) down both flanks, and, as you point out, blackcloud6, were hammered in tank ambushes. So maybe I'm criticising the AI for doing what my defence tried to get it to do - be funnelled into the open. That wouldn't be very fair.

What would I have done, Capn Darwin? Got killed. Lost the battle hands down (in real life, or even in real life gaming, as I'm not much good...), but, being ideal, I would have stuck to the forest and least open routes to get recon elemnts slowly into place (after all, there's no time limit, as such) in positions where they could detect my forward recon, then bring up limited assets to eliminate my forward recon and try as much as possible to assure myself of a more covered approach to the belt of town/forest/undergrowth in the middle of the map. Then so on further down the map. Try to detect him first, then do something about it. Slower, I guess. Which is what people are saying might be good. But what people are also saying is that is not soviet doctrine. What the AI may have done in that game is detected both forward recon that I had, eliminated one of them, flanked the other, ran into an ambush, retreated. Then came down the middle anyway. I had an HQ further back with eyes on that middle ground, and hence there were many AI losses to arty.

But these are all great and interesting points people are making. That said, I believe from Vietikka's points, made more than once, in this thread and elsewhere, that the AI is perhaps not so capable of finesse. I believe Vietikka is saying that 'if Infantry then cover', 'if Armour then out into the open'. Which would suggest it doesn't do anything as sophisticated as assess and reassess then choose between trees and fields. Sounds like he means if it's tanks it goes for fields, if it's feet they go for trees. Regardless? Though, in my example above I have - as you can see - lots of infantry caught out in the open too.

I realise these things are complicated. I will mess around more. I'm assuming that it's nothing to do with soviet doctrine, in fact, but that the US AI will behave the same. I will have to try that out.

[nikolas93TS]

I, as human player, personally tend to avoid covered routes and try to bypass such bottlenecks or strong-points when using mechanized forces.

[blackcloud6]

ORIGINAL: CapnDarwin

@blackcloud6, how much time or effort would the Soviets invested in city fighting in the opening few weeks of WW3?

Well, I don't think the Soviets wanted the war to go a few weeks, especially from the late 70s on. I think they wanted to blitz the Rhine and then call for cease fire by threatening further action and nuclear war. Remember, the French openly stated that an existential threat to France results in a launch of the nukes from the Plain of Albion.

The Soviet strategy is similar to Egypt's in 1973. Modern war is too lethal for prolonged conflict at high intensity and risks nuclear destruction.

That said, in the opening days, with eyes on securing the Rhine, the Soviets would by-pass the large cities and leave them for follow up forces to mop up or starve out.

[CapnDarwin]

Kind of what I was thinking. Thanks for the information.

[Veitikka]

ORIGINAL: Phoenix100

I realise these things are complicated. I will mess around more. I'm assuming that it's nothing to do with soviet doctrine, in fact, but that the US AI will behave the same. I will have to try that out.

There's no 'faction AI' in the game. Actually, when I was doing a lot of research for the AI I mostly used ideas that were based on American sources about the Soviet doctrine.

Indeed, it gets complicated very fast. What is instantly obvious to a human eye means nothing to the computer. I think what some players expect is that the AI should, iteratively and in real-time, to probe and 'see' its environment (elevation, terrain types, bottlenecks, enemies, objectives), make a data bank of all this input, make a list of possible actions based on the goals, threats, casualties and available resources everywhere on and off the map, and finally assign all its units, formations and artillery to their optimal roles based on all this available information.

[Phoenix100]

Yes, Vietikka. That's hit the nail on the head. That is exactly what I expect!!

We can aim for something like that, at least....no?

Does it presently do ANY of that?

[Veitikka]

ORIGINAL: Phoenix100

Yes, Vietikka. That's hit the nail on the head. That is exactly what I expect!!

We can aim for something like that, at least....no?

Does it presently do ANY of that?

Basically it does, but rather crudely. Perhaps one day the latest AI technology out there will make it to this game, but I cannot promise having it in the near future patches. You might even have to wait for an 'AI DLC'. Well, that was a joke, but I believe that even big corporations are struggling to make an AI that's so human-like.

[CCIP-subsim]

As someone who's for years tried to tell machine learning and neural network AI enthusiasts to "hold their horses" and check with cognitive scientists, psychologists and humanitarians on whether their assumptions on whether their models of intelligence are actually representative of, well, intelligence... this is a familiar discussion indeed [:D]

Yeah, it's always a difficult problem - complicated additionally by players having a bit of a "superhuman" information and decision space in a game like Armored Brigade, which honestly by any measure is extremely user-friendly in a way that real war could never dream of being.

Personally, I don't really expect an AI that will work based on the various field manuals and other doctrine documents. What I'd rather see is an AI that can take units at its disposal and try to work out a reasonable use for them given the situation. Even that is really really difficult at the moment - it's a bit easier in "pure" games like Chess or Go, and even there it requires a massive amount of computing power to be human-like... but in "messy", context-rich games like AB, it is near-impossible to have an AI that genuinely "thinks" and plans in a manner that a human player would. Inevitably you have to resort to tricks and shortcuts that will work most (but not all) of the time. Players will still find ways to exploit those things that don't quite work, however.

I'm all in favour of a more "generic" AI though - which assumes that rather than there being a fundamental difference in ways of thinking between Western and Soviet forces, the main difference is in the resources (i.e. units) at its disposal. A Soviet AI will play differently from a NATO AI not because it has different priorities; rather, both want to win, but one has T-72s at its disposal while the other has M1s. And from that view, the best AI is the one that can successfully "understand" those differences and use them to gain an advantage, regardless of which side it happens to be on.

[CapnDarwin]

Fun topic and here is some food for thought. If you take the Red Force situation above and have 100 new players plan a strategy and execute it we will have 100 different plans with a bell curve distribution where a few do splendidly well and a few have Custer-like disaster and a whole bunch of folks in the middle with various levels of success and failure. Now let them plan again and run and again and run and over time the bulk of the bell will shift forward to success. If we take 100 PCs and let the AI go at it with its deterministic logic, we will see all 100 do pretty much the same thing within the confines of the fuzziness in the deterministic system. Run the AI group a million times and the results aren't going to improve. They will just fuzz about the deterministic centerline. Why? Simple answer is humans learn. We have memory. We generally remember with great clarity what worked well and what was a disaster and therefore we can adjust the planning the next time the same set of circumstances are seen. Even the best AIs with a lot of fancy if-then/neural constructs are still short on this learning and memory concept. While we have learned that it can be very bad in many situations to race tanks across open ground in face of an unseen enemy, the AI only has whatever fixed parameters or weighted outcomes that a programmer can give a certain situation. Where a human is almost effortless in looking over the map and its complex variables and can think and remember on the fly, the poor AI is running ones and zeros, if-thens, and rolling some dice for a weighted variable situation. Even by chance, if the AI does something well, we may perceive it as a dumb move based on our own way of resolving the situation. Call it an intellectual observation bias. AIs (and I hate calling them that by the way since they are just fancy decision engines today) in games is a very difficult thing to work for a number of the factors discussed. No doubt the team will make improvements over time, but as players we need to keep out intelectual observation bias in mind when looking at these situation.

Party on! [8D]

[22sec]

I think within the parameters of AB the AI deals with the same issues as we humans do when it comes to managing forces. It is a challenge to maneuver formations across a battlefield. I think some kind of system is needed to influence the AI in which formations to use as well as which paths (covered, shortest, quickest) to use. The other suggestion I would like to see considered is side specific no-go zones, which are not visible to the opponent. For example I design a scenario, and I create a series of no-go zones for the attacker, thus helping to influence the AI’s plan. The current no-go zones do that, but they apply to both sides and are visible to both sides.

[TarkError]

ORIGINAL: Phoenix100

Which is what people are saying might be good. But what people are also saying is that is not soviet doctrine. What the AI may have done in that game is detected both forward recon that I had, eliminated one of them, flanked the other, ran into an ambush, retreated. Then came down the middle anyway. I had an HQ further back with eyes on that middle ground, and hence there were many AI losses to arty.

Well, a Soviet commander might as well use a difficult covered approach if the preservation of his force is more important than time for the success of his mission. I remember reading in Chris Donnelly's Red Banner where a NATO officer who thought that a Soviet division would be channeled along valleys while would be in for an unpleasant surprise if he considered that the wooded hills were an obstacle "on account of their lack of roads". There's more to Soviet tactics than the stereotypical mass attack.

The Syrian attack on the Golan Heights in 1973 was about as far from Soviet doctrine as possible...they pretty much ploughed straight ahead with no tactical maneuver at all, with ineffective fire preparation+support, and the Syrian high command failed to redirect their efforts from the failed attack at the Valley of Tears to the south (a reinforcement of failure).