Can we make armor AI less prioritize to forward to city?

[exsonic01]

December 1991, fair weather day, start from ~15:00 PM. 15km x 15km (max size map), 50k vs 50k point meeting engagement, US armor vs USSR armor. Map is the area near city of Schneverdingen, so "battle of Schneverdingen" would be proper name. I can't recall the detail, but I purchased 8 M2A2 platoons, 6 of M1A1HA platoons, and 2 M3A2 cavalry platoons for US army. On the other hand, I gave USSR 18 BMP2 platoons + 6 T80U platoons + 12 T80BV platoons + 1 T80B platoon. Total 8 VPs, all maximum option for dummy. 4 in the city of Schneverdingen, and 2 north of city and 2 south of city.


I expected something to see armored battle around the city, and artillery + infantry battle in the city. However, what I saw was the first battle of Grozny. Look at that rushing T-80BVs.


I rushed all of my mech infantry to rush to the town, and preoccupied some important points. Without proper infantry support, or smoke or at least any pre-assault artillery, T-80BVs just rushed to my infantry + Bradley trap. I hardly get return fire, looks like T-80BVs have no idea where the shots are coming from. I think there were no recons in the city, though I gave them good amount of recon BRDMs and dismounted recon infantry. Well, I killed most of BRDMs before they reach to the VPs. But still, AI just rushed all tanks into the city, without knowing what is inside there.


Those tanks are dead already, look how they are sandwiched between HE and infantry + M2A2 line. All M2A2s were hiding in the building ruin tiles.


After the game is over, I figured out why there were no infantry support when tanks were pushed in to slaughterhouse. I find that all Soviet infantries dismounted from BMP-2 too early, not that far from the city but still some distance away. Then they attempted to walk across the town. Really?


Look from ISO view. I mean, what a view.
AI used tons of artillery, as I gave AI 3 artillery company (=36 tubes) and AI scored good amount of my Abrams using DPICM and Su-25 airstrike. Mi-24V scored a few Abrams and good number of my recon Bradley. And AI really loves to pour HEs in to the town. Some shells got a lucky kill on my Bradley, and some wiped out my squad. But still, it was not enough to clear all my boys from the town and eventually my defense line survived.


I think map + VP were the reason behind this behavior. You see, there are only 2 axis that Red army attempted to push. Center and south of city. Map has the big forest in the north, and there are 4 VPs in the middle of the city, 2 in the north of city, and 2 in the south of the city. I think those geometry and VPs forced AI not to choose northern forest route, and push half of his forces to the city.


This pic shows better axis of thrust of Red army in this game. Center to the city, and south of the city. I think those 2 corridors are kinda obvious, that anyone who played any wargame might be able to predict those routes without any difficulty. You can see I already deployed FASCAM to the major route. Geometry of this map has several choke points like this, make it easy to predict the assault. Maybe this is another reason why AI pushed his tanks into the city, because AI felt there is no where to assault.

But still, putting all armors into the middle of the city, without recon or infantry support or at least smoke or pre-artillery strike, is a big big mistake and shouldn't happened.




I think there should be a way to escape from those situations. Maybe I should have placed VPs outside of city, or just put only one or two VPs in the city and place other 6~7 as wide as possible through the map. Maybe I should've purchased more AI recon vehicles and recon teams. I purchased AI forces by my self, all of them were platoon, and I maximized all options regarding dummy VPs. Yet, I couldn't prevent those suicidal push of armors.

It would be great to share how to set up the game and other conditions (especially when there is a city in the map) to prevent such happenings. Because, I don't think this map was particularly disadvantageous to AI. City, town, woods, and several natural choke points like this map can be found in any maps in this game.

Why AI unloaded infantry from taxi far from the city? That I really have no idea. It was not "that far", but considering the speed of infantry in this game, that was too far from town. This is meeting engagement, not assault or defense. I mean, in the meeting engagement, AI should've moved units more rapidly and actively to preoccupy some important positions. I really don't know how to prevent AI from dismount his boys too early like that.

Also, it would be great if AI armor could recognize city and city boundary, forest and its boundary, and field. Then, make AI do not go into the city, unless there are "good number" of recons and infantry are already inside the city boundary. Or, maybe it would be possible to analyze the total area inside the city boundary, and sum of area of all LoS of infantry and recon inside the city, and compare them. If area of LoS of infantry in the city is greater than some number, let's say 40%, when compared to area inside city boundary, then allow AI to push to the town. AI tanks need to learn how to wait until it is ready.

Plus, it would be great for AI to try unexpected routes, like northern forest of above map. I know terms like "unexpected routes" are not possible for AI to understand neither recognize, but sometimes forest is good way to catch a player off guard. It would be great if AI consider the places out of natural choke points as axis of thrust. Decision of "natural choke point" seems not easy as well, but there should be a way. If it is able to write a geometry / map analysis script, and calculate the area and position of forest and forest boundary, it would be helpful for AI to which is the forest and not, and might be helpful for AI to decide assault via forest.

At the same time, I think weapons like RPO, flamethrower (I know AB just began to introduce those units), chemical weapons, realistic WP, and TOS-1 would be great for such situations. We need more options for thermobaric, napalm, WP, and chemicals. In addition, if area fire (or command fire on empty tile) is not possible, at least allow units to perform return fire against the muzzle flash region. Limit the fire rate and total amount of ammunition for such "fire against muzzle flash" to prevent "gone Winchester".

[nikolas93TS]

I see there were at least three VP in the town itself. So no surprise AI tried to take it. Also, flanking attacks might have exposed the force to defilade ATGM attack in open fields north and south of the town.

[exsonic01]

ORIGINAL: nikolas93TS

I see there were at least three VP in the town itself. So no surprise AI tried to take it.

Yeah I also thought the same, too much VP inside the town make AI rush to the town... Then, how do you recommend to put VPs in above map? 1 VP in town and displace others as wide as possible?
Also, it seems that the place of VP plays very important role on AI behavior, so it would be great there would be a rule of thumb or guideline of VP placing. For example, place them symmetric, or don't put too much VPs in the town or etc...

Also, flanking attacks might have exposed the force to defilade ATGM attack in open fields north and south of the town.

Hmm that is why I gave AI 4 mortar teams and 36 tubes, lead AI to use smoke and use mortar to snipe the ATGM vehicles. But strangely I didn't see any smoke field by AI, and mortars were relatively quite. Maybe because I killed all recon vehicles quickly.... If you see above pics, second from bottom, AI don't spread out his recon infantry, it seems AI didn't try to infiltrate. That might be a reason behind the poor reaction of AI.

[nikolas93TS]

AI has a primary goal to take objectives, so having 3 VPs in the town and so forward on the front-line will most likely imply enemy troops will appear en-force in town first. It might however opt for a different plan, but scenario has to be replayed several times to see how AI react each time.

[exsonic01]

ORIGINAL: nikolas93TS

AI has a primary goal to take objectives, so having 3 VPs in the town and so forward on the front-line will most likely imply enemy troops will appear en-force in town first. It might however opt for a different plan, but scenario has to be replayed several times to see how AI react each time.

Yeah, I tested by myself last night in the same map, and watched them via dev mode. You are right, 3+ VPs in "big" city is not a good idea, I tested just 3 games, USSR, FRG and DDR, and all games AI prioritize the city and rush everything AI have to city, which makes the game like the first battle of Grozny. I think 1 VP is kinda OK, but 2+ is no good.

I think, it would be good to introduce & share to new players to how can we set the AI in this game more challenging or interesting, in game tutorial or manual. Here's my list so far.

1) When select the map, try minimize the tile of big city or big woods for AI deployment zone.
2) Try to purchase platoon for AI forces instead of company.
3) Set dummy object chance and dummy area maximum.
4) Try not to put VPs inside the big city. If place, place just one VP only, and set other VPs in different places.
5) Proper distance among VP is required.
- If VPs are located too far in random manner, AI try to disperse his forces, this makes AI units become an easy target, which can be crashed part by part.
- But if multiple VPs are located too narrow in small region, this makes AI to concentrate to corresponding region. Route can be differ, but player can easily predict their approaching way. Game can be too easy.

I guess there should be a rule of thumb for VP setting to make game more interesting and fun. I'm mostly playing "big games" only these days, max size map with max size points, so I usually set 7~8 VPs. I'm searching for more ways to set VP to help and improve AI.

[Veitikka]

ORIGINAL: exsonic01

Hmm that is why I gave AI 4 mortar teams and 36 tubes, lead AI to use smoke and use mortar to snipe the ATGM vehicles. But strangely I didn't see any smoke field by AI, and mortars were relatively quite. Maybe because I killed all recon vehicles quickly.... If you see above pics, second from bottom, AI don't spread out his recon infantry, it seems AI didn't try to infiltrate. That might be a reason behind the poor reaction of AI.

The AI opponent doesn't use mortar HE shells against vehicles. This can be modded by changing the 'vsVehicle' attribute in the database for these shells.

I think currently the AI never uses mortar smoke shells, and artillery smoke is used only when the player is defending. What you had was a meeting engagement.

[exsonic01]

ORIGINAL: Veitikka
The AI opponent doesn't use mortar HE shells against vehicles. This can be modded by changing the 'vsVehicle' attribute in the database for these shells.

I think currently the AI never uses mortar smoke shells, and artillery smoke is used only when the player is defending. What you had was a meeting engagement.

Thanks to let me know. Soft skin vehicles like ATGM launchers can be denied by mortar, but chances are not that high. Yet, it would be worth to try. I wish if AI can distinguish the soft/hard skin vehicle.

It would be great if you could make AI use smoke more frequently in any mission type. Defense, meeting engagement, and offense, all of them. Or at least for meeting engagement.

I guess similar artillery smoke algorithm for assault can be applied for meeting engagement, smoke all AI assault/approach route, or use smoke when AI forces are in the open field. For defense, it would be great if AI use artillery smoke on units being attacked by my units.

[Veitikka]

ORIGINAL: exsonic01

ORIGINAL: Veitikka
The AI opponent doesn't use mortar HE shells against vehicles. This can be modded by changing the 'vsVehicle' attribute in the database for these shells.

I think currently the AI never uses mortar smoke shells, and artillery smoke is used only when the player is defending. What you had was a meeting engagement.

Thanks to let me know. Soft skin vehicles like ATGM launchers can be denied by mortar, but chances are not that high. Yet, it would be worth to try. I wish if AI can distinguish the soft/hard skin vehicle.

It would be great if you could make AI use smoke more frequently in any mission type. Defense, meeting engagement, and offense, all of them. Or at least for meeting engagement.

I guess similar artillery smoke algorithm for assault can be applied for meeting engagement, smoke all AI assault/approach route, or use smoke when AI forces are in the open field. For defense, it would be great if AI use artillery smoke on units being attacked by my units.

I think it's a good idea to allow the AI to use smoke screens in meeting engagements. The artillery smoke algorithm was improved some time ago, and it should work well in this. Perhaps it will help the AI in situations where the player rushes to the objectives and waits for the AI there.

Firing smoke to the AI's own positions is more debatable. If not done carefully, the AI can make its own units blind, allowing the player to advance freely.

[exsonic01]

ORIGINAL: Veitikka
I think it's a good idea to allow the AI to use smoke screens in meeting engagements. The artillery smoke algorithm was improved some time ago, and it should work well in this. Perhaps it will help the AI in situations where the player rushes to the objectives and waits for the AI there.

Firing smoke to the AI's own positions is more debatable. If not done carefully, the AI can make its own units blind, allowing the player to advance freely.

For meeting engagement, I guess it would be better for AI to deploy smoke filed in AI half of the map, or only up to the center line of the map. Open field, approaching route, and VPs of AI half or center line region... Because, if AI field the smoke on player half, that would screen and help players. But there should be an exemption for AI unit in friendly half of map.

Or, maybe it wouldn't be good idea to recognize smoke area by AI half or player half. Anyway, I wish your algorithm could smartly cover meeting engagement