Combat Mission AI and scenario design part 2

[Pericles]

Here is another example of unsatisfactory, immersion-breaking AI (or design) from the "Charge of the Stryker Brigade" campaign. These four enemy BMPs had not yet been engaged and this is only about six turns into the scenario... in less than a single turn (1 minute) they emerged from behind a hill and clustered together as seen in the picture. One of my javelin teams took one of these out later in the turn. 

This is highly disappointing. Are there plans to improve AI, or at least to release more competently designed products? I've never designed a scenario before, but this seems so amateur to me (to have a bunch of BMPs simultaneously crest a hill in an exposed position). Perhaps the designer is attempting to model how a panicked Russian force might react, but I simply can't imagine a bunch of Russian BMP drivers doing something like this, even when they weren't sure where the enemy was. 

 

[Vergeltungswaffe]

1 hour ago, Pericles said:

I simply can't imagine a bunch of Russian BMP drivers doing something like this, even when they weren't sure where the enemy was.

In combat, situational awareness sometimes becomes clouded and things happen.  These vehicles bunched up in Syria and paid the price.

and these 13 Churchill tanks of the 107th Regiment RAC (King's Own) managed to all get knocked out in a clump during a battle on the Orne in August 1944.

Sometimes things happen in combat and sometimes things happen that weren't what scenario designers intended.  But the first company to make an AI that approaches the challenge of playing against a human being in a tactical combat game will be unspeakably rich.

 

[Pericles]

The fact that you were able to post those real world pictures so quickly after reading this post is very impressive. But my point stands regarding the AI/scenario behavior in this particular tactical situation. 

You're right: the first company to make an AI that approaches the challenge of playing against a human being in a tactical combat game will be unspeakably rich. 

I think Battlefront can do better, maybe even become leaders in AI performance in the video game industry (most developers are resorting to exclusive multiplayer now anyway because AI is so difficult to do properly). Let's hope that they get more sales so that they can devote more resources to the improvement of AI and scenario design tools. 

[Bulletpoint]

It's not about AI, because there is no AI.

At least there is no AI that moves troops around.

It's about scenario design.

To me, it looks like the scenario designer gave a move order to a group of vehicles, but only gave them ONE square to move to. That means all the vehicles will try to move into that one 8x8m square, but there's not enough room for all of them. So they bunch up.

In the end, the designer needs to understand the tools he has to work with.

[Badger73]

22 minutes ago, Bulletpoint said:

It's not about AI, because there is no AI.

At least there is no AI that moves troops around.

It's about scenario design.

To me, it looks like the scenario designer gave a move order to a group of vehicles, but only gave them ONE square to move to. That means all the vehicles will try to move into that one 8x8m square, but there's not enough room for all of them. So they bunch up.

In the end, the designer needs to understand the tools he has to work with.

Agreed.  Yet Scenario Design remains an arcane art.  There are several very experienced designers who are very good at scenario design.  Maybe someday, an articulate, analytical, well qualified freelance writer could produce a BFC vetted scenario design document, much like Jon Sowden's "The Sheriff of Oosterbeek" design DAR/AAR with greater focus on programming the AI which might also incorporate the tips and tricks learned by the wizards of this craft. 

 

[Glubokii Boy]

4 hours ago, Bulletpoint said:

 

To me, it looks like the scenario designer gave a move order to a group of vehicles, but only gave them ONE square to move to. That means all the vehicles will try to move into that one 8x8m square, but there's not enough room for all of them. So they bunch up.

 

From the Nr 1 thread on this topic...A post by sgt Squarehead

"It may simply be down to an area being 'block painted' in an AI Setup or Move order.....Setup is always best done manually and I've found that if you paint one tile per vehicle in an AI group, it will often attempt to assign one vehicle per tile when carrying out those orders in game. "

I agree with this fully. Both that the cause for the problem also is most likely a 'block-painted' waypoint and the best qure for  it within the current game engine is to use multiple seperate tiles painted. The benefit of this is also that you will be able to specify the desired formation (line, vedge etc..) you want the AI Group to deploy in. In this particular situation paint 4 seperate tiles...chose the formation, the spacing and the excact location by painting the desired tiles...

Like the sgt mentioned...by doing this the AI will usually asign 1 vehicle/tile...most of the times atleast 

 

 

[Pericles]

So the consensus here is that this bunching behavior is a symptom of incompetent scenario design. Extra disappointing because this is a campaign sanctioned by BF ("The Charge of the Stryker Brigade", part of the $10.00 Battle Pack). Clearly it was rushed. 

Just played another turn and sure enough, now the enemy tanks are bunching up unrealistically:

 

[Sgt.Squarehead]

22 minutes ago, Pericles said:

Clearly it was rushed.

Not sure that's entirely fair, there do seem to be one or two issues with the briefings in this campaign, but the behaviour you are seeing may very well just be the TacAI making bad calls.....As the discussion above suggests there are several possible causes, but without decompiling the campaign and taking a detailed look in the editor (we'd need to know what plan the AI was on too) it's impossible to determine for sure the precise cause of the bunching.

Based on your new image I'd guess 'block-painted' AI orders are the cause, but without knowing the composition of the AI group it's unfair to criticise the designer's use of that policy.....The single tile per vehicle technique can go a bit squiffy if you have infantry in the same AI group (there are workarounds, but they are map dependent and definitely easier on MOUT maps).

FWIW

[ASL Veteran]

A couple of points about when vehicles enter a new waypoint location.  If the designer places a rectangular block of say - eight wide by four deep - then the vehicles will normally tend to distribute themselves evenly throughout the painted block.  However, there are occasions when, for some unknown reason, the AI will attempt to place two or more vehicles onto the same block and there is absolutely nothing that the designer can do about it.  This is most frustrating when you want the vehicles to unload troops at a specific waypoint.  When this behavior occurs the first vehicle to arrive will occupy the spot and then the subsequent vehicle (usually it is only one other vehicle when this happens, although more than one can do this on rare occasions) will stop next to the location it wants to occupy.  The vehicle that is attempting to enter the occupied location will then sort of .... move in place a bit and turn ever so slightly which prevents troops from unloading since the vehicle never fully 'stops'.  It does this until the vehicles move to the next waypoint.  Sometimes this can be a side effect of a waypoint that has been painted too small, although it can also happen with a waypoint of any size so when it does happen there really is nothing the designer can do about it other than keep repainting the waypoint until it doesn't happen during testing.  I don't know for certain what is going on in the image, but it could either be the game or the designer.  Once again, unless the designer chooses to join the thread or someone looks at the scenario file itself we won't know what's going on in this particular scenario.  The different vehicle facings do suggest that either the waypoint is too small or the AI vehicles are attempting to occupy the same spot since they would normally all be facing the same direction (the next waypoint or a spotted enemy)

Keep in mind that the when the designer puts in the AI plan he has no idea where the player will be with his forces at any given point in time.  Triggers help some, but if the player does something completely unanticipated the AI will not react to what the player is doing since the designer is essentially creating a plan to fight an opponent while completely blind as to what the player is doing.  Imagine fighting a scenario by laying out all your plans before the scenario starts and then not adjusting or altering anything no matter what happens during the course of the battle.  That's how an AI plan is created.

[Erwin]

14 hours ago, ASL Veteran said:

Imagine fighting a scenario by laying out all your plans before the scenario starts and then not adjusting or altering anything no matter what happens during the course of the battle. 

Ironically this may quite realistically portray Soviet rigidly preplanned  tactics - at least during early war era.

[ThePhantom]

4 or 5 vehicles in one AI Group has a tendency to create group clustering. I've tried every possible way to avoid it.... However, it happens a lot. The answer is, bring the vehicle per AI group down to one or two. The key to success for the Combat Missions editor is to add more AI groups in the editor or create small scenarios. 

[sburke]

On ‎2‎/‎12‎/‎2018 at 12:32 PM, Pericles said:

So the consensus here is that this bunching behavior is a symptom of incompetent scenario design. Extra disappointing because this is a campaign sanctioned by BF ("The Charge of the Stryker Brigade", part of the $10.00 Battle Pack). Clearly it was rushed. 

Just played another turn and sure enough, now the enemy tanks are bunching up unrealistically:

This is pretty rude. 

One thing you may not be aware of (and as others have noted) is when you give a group an order to move from point a to point b, they won't necessarily follow good logic to get there.  For example if you have a group of pixeltruppen in a row of houses and you paint the row of houses across the street and order them to move there, they won't just all go straight across the street to the nearest house.  You'll see some go for the far right house all the way to the far left house across the street.  Big scenarios are difficult to manage which is one reason I avoid them.  In your case I don't actually know if the designer did anything wrong at all.  The TAC AI may have decided to criss cross the vehicles in going to their next waypoint.

Does it occur every time the battle is played or did it occur just this instance while you were trying it?   I can appreciate the disappointment when stuff like this happens, but to publicly call out a designer whom you don't even know and call them incompetent is pretty crappy and an example why a lot of folks won't take on design or won't release it.  It has been hard enough getting folks to take a dip in the water and create new stuff, don't piss in the pool.

 

Edited February 14, 2018 by sburke

[MikeyD]

A question, has 'Pericles' done any scenario AI orders design, himself? Its all right there in the editor, its part of the game. Can you get the AI-controlled enemy to do what you want? And if not, why not?

[Sgt.Squarehead]

Harsh.....But fair. 

[Pericles]

6 hours ago, sburke said:

This is pretty rude. 

True. My charge that the scenario was designed "incompetently" was too harsh. It was borne out of frustration. 

No I have never attempted to design a scenario. 

Overall, pointing out problems with AI/scenario design is constructive if the problems are real. I provide two examples in this thread of enemy bunching in the first 10 minutes of a scenario in a stock campaign. This is sufficient empirical evidence that there is a problem. It does not matter if I play the scenario over again and find that the AI does not bunch. The fact is, the AI bunched in two separate instances in the first 10 minutes of the scenario.

The problem may be with the AI ("TacAI"), it may be with the scenario design tools, or it may be with the scenario designer's decisions. Given the comments so far,  it seems I was wrong to conclude that the problem is solely with the decisions made by the scenario designer. It is all of the above. Perhaps the designer could have split the enemy into smaller groups. 

Overall the designer for this campaign has done well. But bunching like this really takes the wind out of my sails. 

[MOS:96B2P]

8 hours ago, sburke said:

 It has been hard enough getting folks to take a dip in the water and create new stuff, don't piss in the pool.  

+1.  This.  The editor can be very rewarding and the more scenario designers that will release their scenarios the better. 

On the positive side this thread has encouraged some designers to talk about various problems and solutions when working with the TAC AI.  If this could be a thread about tips, tricks and advice when designing scenarios it could be a useful conversation.  Like a military AAR you want to be honest about problems and learn lessons but you also want to find ways to improve things going forward.   

[Bulletpoint]

While Pericles was a bit blunt in the way he phrased his frustrations, his point still stands. A campaign made by Battlefront should ideally be quality checked to a higher degree than a user-made campaign.

I'm not saying the editor is easy, and I definitely understand how errors can creep in. I made a good deal of mistakes, myself, and I wouldn't call anyone incompetent. But when I make mistakes, people generally forgive them, because they know I spent several weekends of my spare time providing a scenario for free.

If I sold my scenarios for real money, I would also have to accept that people would get more frustrated with my mistakes. In that sense, I think Pericles has a fair point.

 

[A Canadian Cat]

1 hour ago, Bulletpoint said:

While Pericles was a bit blunt in the way he phrased his frustrations, his point still stands. A campaign made by Battlefront should ideally be quality checked to a higher degree than a user-made campaign.

And they are. BFC provided scenarios are tested at least an order of magnitude more than any 3rd party made scenario I have ever seen. The only thing that comes close is when a scenario is featured at theBlitz for the scenario of the month but that does not exercise the AI - 'cause PBEM. Oh and BTW one of the best ways to find a more challenging opponent is to find a more challenging opponent - a human. Check out the "New to Multiplayer" section of this post

Some thoughts on the subject of testing: https://dojo.ministryoftesting.com/lessons/why-didn-t-you-find-that-bug PS those thoughts are from a tester's point of view on how to try to get better at testing - that's how I try to think. I am sure I am not perfect either.

 

1 hour ago, Bulletpoint said:

If I sold my scenarios for real money, I would also have to accept that people would get more frustrated with my mistakes. In that sense, I think Pericles has a fair point.

IMHO rudeness is never fair.

[Bulletpoint]

1 hour ago, IanL said:

Quote

If I sold my scenarios for real money, I would also have to accept that people would get more frustrated with my mistakes. In that sense, I think Pericles has a fair point.

IMHO rudeness is never fair.

Well if I shouted "IT'S F...ING RAINING!" you could call me rude, and you'd be right.. but you might still want an umbrella 

[A Canadian Cat]

2 hours ago, Bulletpoint said:

Well if I shouted "IT'S F...ING RAINING!" you could call me rude, and you'd be right.. but you might still want an umbrella 

Yep. The thing is I avoid people who are rude - even those that might be correct. Rude is pointless, rude is crass, rude is unnecessary. My preference is to surround myself with smart, confident people who speak their mind and are constructive and thoughtful. Rude people are kicked to the curb. I don't really care how damn smart you are rude = outta my sight. Which means if I am the boss you are fired if I'm not I'll ignore your rude butt. Life is too short to bother with rude people.

Oh and I can tell if its raining pretty well myself thanks. See no value in rudeness.

[Pericles]

Can two instances of bunching within the first 10 minutes of a stock campaign be a sign of poor design? Maybe all the thorough testing before release didn't pick it up? I highly doubt it. If it was tested a lot before release, this would have been noticed and corrected. I believe this because it happened so early on in the scenario, before the "TacAI" could have got rolling so to speak. Like I said, I'm interested in constructive discourse. Throwing an "incompetent" or "kicked to the curb" in here and there makes things more colorful as long as empirically supported claims are being made. 

And +1 for Scenario of the Month, great initiative. 

And now for some bunched smoldering wreakage in a large open map:

 

[Erwin]

40 minutes ago, IanL said:

...I avoid people who are rude - even those that might be correct. Rude is pointless, rude is crass, rude is unnecessary. My preference is to surround myself with smart, confident people who speak their mind and are constructive and thoughtful. Rude people are kicked to the curb. I don't really care how damn smart you are rude = outta my sight. Which means if I am the boss you are fired if I'm not I'll ignore your rude butt. Life is too short to bother with rude people.

 

Xnt sentiment, Ian.  I have copied that statement for future reference.  Rudeness rears its ugly head way too often on these forums.

[Bulletpoint]

10 hours ago, IanL said:

12 hours ago, Bulletpoint said:

Well if I shouted "IT'S F...ING RAINING!" you could call me rude, and you'd be right.. but you might still want an umbrella 

Yep. The thing is I avoid people who are rude - even those that might be correct. Rude is pointless, rude is crass, rude is unnecessary. My preference is to surround myself with smart, confident people who speak their mind and are constructive and thoughtful. Rude people are kicked to the curb. I don't really care how damn smart you are rude = outta my sight. Which means if I am the boss you are fired if I'm not I'll ignore your rude butt. Life is too short to bother with rude people.

Oh and I can tell if its raining pretty well myself thanks. See no value in rudeness.

We agree, Ian. I was just trying to say that sometimes, focusing on whether someone is rude or not can be a red herring. I think what Pericles showed was interesting - and I also would be disappointed if I saw that bunching up behaviour in my game. But, here's the interesting part - I never saw that happen. Never.

And regular readers will know I'm probably the biggest stickler for details on the forum. Yet I never saw that problem happen. So I'm thinking it's not about how the game TacAI works, but an issue with the way the AI plan was put together for this particular scenario.

Let's calmly discuss that

[A Canadian Cat]

6 hours ago, Bulletpoint said:

And regular readers will know I'm probably the biggest stickler for details on the forum. Yet I never saw that problem happen. So I'm thinking it's not about how the game TacAI works, but an issue with the way the AI plan was put together for this particular scenario.

Let's calmly discuss that

Yep. Carry on discussing it