Off-map mortar inaccuracy and headless chickens

[Kwacker]

Just playing a quick battle, still trying to figure out the intracacies of CMBB after playing CMBO for ages.

I'm in a dificult situation, the AI has been given a bunch of guns and the terrain is fairly open. I have a few half tracks but only 1 81mm spotter to really hit back at the guns - I can't get anyone else close enough. Unfortunately the spotter is out of line of sight but I figure what the hell the guns are fairly dispersed so any innaccuracy probably won't matter and it's probably better to live with the firing delay than try to get into visual range and get shot up by some roving armored cars. Anyhow, I duly wait 4 mins keeping my head down for the mortar rounds to come in and then I can't even see them landing! Turns out they are landing off-map and nearer to the spotter than the target. Now, maybe this is a realistic representation of german mortar skills in 1944 - maybe one of you uber-grogs could confirm this. I don't seem to recall German mortars in CMBO being similarly afflicted - or maybe they just had their glasses recalled and can't read their mortar charts!

A second whinge. The same game I have a platoon of green recon infantry advancing to some tall pines to find cover and spy out the opposition. Unfortunately they run into the armored cars at ~400m so they hit the deck and begin sneaking. So far so good. All the squads get there but the HQ being a little behind the main body takes a casualty. Fair enough. Moments latter the HQ takes another hit and panics. Fair enough. However the HQ is now about 2m from the trees. What does it do? It sneaks 50m back into the open, the way it came, gets shot up some more and eliminated. Aaaargghhh!!!

Now maybe this is also realistic - maybe guys really do commonly do the stupidest things when their pants turn brown. However, that is not the impression I get from the accounts I have read. My guess is that, when under fire in the relative open the greenest or most elite troops will attempt to find some cover with more or less success. Some may freeze and many may attempt to bury themselves in the ground but I really can't imagine very many would attempt to repeat the exercise that just cost them 50% casualties when there is cover within 2m and plenty of other supporting units (ie alternate targets) all around them.

Strikes me that some tweaking of the AI is required here. I'm guessing, having dabbled a little with programming, that it would not be too hard to more heavily weight the response of panicked troops in this situation to find the nearest decent cover and stay there. For me (right or wrong) that would be a big improvement in the realism factor.

All that said CMBB is still the best game I've played and CMBO has had me hooked since the day I got it, about 40,000% longer than any other game, so thanks Steve and Charles!

[Bullethead]

I tend to agree that the direction and speed chosen by panicked troops is sometimes strange. They seem more like squirrels than headless chickens IMHO, however.

Squirrels often get run over because they have the "last safe place" mentality, which works like this: A squirrel decides to cross a road. When it starts, it's in a safe place. But then you come down the road and the timing is such that the squirrel is about 95% of the way across the road when your car gets close enough to panic it. When panicked, a squirrel runs for a known safe place, and the last one it knows of (and probably the only one it can remember at the time) is where it was when it started. So it turns around and tries to run back across the road to its starting point. This of course takes it directly in front of your car so it gets squashed.

As to your mortar inaccuracy... CM off-map arty of all types follows a pretty small set of simple rules. In CMBB, one such rule is that if the FO, of any type, doesn't have an LOS to the target point, the fire will always scatter off target by up to 400m. The scatter amount seems weighted on the long side as well, because my tests have shown it's somewhat more likely to be 400m off than dead-on, and the most common distance is about 250m rather than 200m.

Thus, the usual miss distance is greater than the N-S width of the normal impact pattern for guns and mortars. This means firing blind is almost always a complete waste of ammo.

NOTE: this assumes normal in-game fire at a non-TRP location. Blind initial barrages and blind fire at TRPs don't scatter, or at least not enough to make a difference.

[Kwacker]

Thanks for the off-map targeting info Bullethead. I'm fairly sure this is a change since CMBO so I guess I just have to relearn my SOPs. If it is a change I'm sure Steve and Charles have supporting evidence for this type of inaccuracy so who am I to argue.

On the other point squirrel, or headless chicken, I think it is debateable. In the same QB (I was not having a good day ) I had a squad hiding behind a hedge rout and run laterally towards some tall pines - away from a small building that was considerably nearer - thus exposing themselves to the threat for much longer. They got shot-up again, then ran back to where they originally routed, got shot-up a third time and then ran directly away from the enemy, again in the open, by which time, fortunately for them, the enemy had found better targets to shoot at. Perhaps brainless chicken would be a better description as panicked troops seem to eye-ball a patch of cover at random, fairly independant of it's effectivenss, and leg it in that direction.

In my opinion nearest cover would be best in more cases and this is what the AI should aim for. On the western front I'm sure this would have been the case - if all else fails the troops would have surrendered once cut-off or under very heavy fire in cover. On the eastern front maybe they would have been more inclined to get out of there - even if it mean't exposing themselves to more risk.

[Michael Emrys]

Originally posted by Kwacker:

In my opinion nearest cover would be best in more cases and this is what the AI should aim for.

IIRC this is how it was in BO and there were many protests that troops were routing to the nearest cover even when it was occupied by the enemy shooting at them. You wouldn't like that would you? Well, neither did anyone else. So, when it came time to tweak the AI for BB, it was decided to weight the decision of which direction to route to towards the friendly map edge. This has resulted in the behavior you have just witnessed.

This is one of those areas where it is especially tricky to program. Things that are obvious and self-evident to a human brain turn out to be very complicated for an artificial program to duplicate. I'm confident that such a program can be written, but I fear that the string of logic is apt to be a long one.

This is just one more reason why the engine rewrite needs to be given first priority above graphic improvements, IMO.

Michael

[Kwacker]

Hi Michael,

Do you know for sure that the AI behaviour has been tweaked in this respect cf to CMBO? I'm wondering if the behaviour is just more obvious as infantry tends to get pinned / routed more (at least when I'm playing )

And no I don't want panicked troops rushing towards the enemy and I don't want 'em often running away from "safe" cover either. Guess I want it all - and one day we may get it thanks to the guys who made this game. By the way CM is not a real time game so frankly I'd rather sacrifice turn calculation time for good AI behaviour anyday - even if an average turn with a bunch of panicked units took several minutes to determine.

I agree that coding AI is never easy but maybe this or similar is / could be the logic / procedure:

To choose destination for units when under fire and panicked / routed:

Ignore anywhere more than 200m away.

Break potential area into a grid (grid resolution 2.5 x 2.5m??) and estimate starting "figure of merit" (FOM) for each destination on grid based on cover provided at that point e.g. woods = 25%, out of LOS of all enemies = 100%, road = 0% etc. This uses existing parameters / tools in the present code.

Modify FOM as follows:

1) Determine path maximizing cover to each grid point using search routine (PCs are very good at these!). Multiply FOM by mean cover of path. e.g. all wooded path = x 25%

2) Multiple FOM of each destination point by 200m - distance from starting location / 200m or some similar function e.g. a point 10m away *190/200, a point 150m away *50/200

3) Multiply by some trigonometric function favouring movement along the direct path to friendly map edge over lateral directions over directions towards your enemy. Would simply require calculation of the angle between a line from present location to the destination and a line from present location to the friendly map-edge along a path normal to the map edge. Using a look-up table the angle could be converted to some value between 0 (straight to enemy map edge) and 1 (straight to friendly map edge). This way running towards the enemy is not ruled out - just disfavoured.

4) Pick the destination with the highest FOM after these calcs

Naturally the above would require a lot of testing and tweaking to get right but my experience of similar problems is that the above simple rules or similar can give decent AI results. For example you might want to weight distance over direction to avoid the issue I originally whinged about - you get the idea.

One other complication is assesing when the above chosen destination should be reassesed - to avoid brainless chicken syndrome. My guess is that it should only be re-evaluated when a new unit starts shooting at the fleeing unit - not when there is incoming from the present targeter as it appears at present.

As a final thought the FOM of the present location will often win out - nothing wrong in that result I guess.

Anyhow enough of my idle thoughts, but I sincerely hope this gets improved in CMII - I'm sure it will as it is one of the few major flaws in present human v human play.

[Michael Emrys]

Originally posted by Kwacker:

Do you know for sure that the AI behaviour has been tweaked in this respect cf to CMBO?

I'm going on a definite recollection of all the complaints and a much less definite recollection that someone from BTS said something about this as a fix. Keep in mind that I am in an advanced state of senility.

As for your suggestions in the remainder of this post, I can't guess how they would work within the framework of all the other code that goes to make up CM, but they do strike interest in my fevered brain.

Michael

[Bullethead]

Kwacker sai:

Thanks for the off-map targeting info Bullethead. I'm fairly sure this is a change since CMBO so I guess I just have to relearn my SOPs. If it is a change I'm sure Steve and Charles have supporting evidence for this type of inaccuracy so who am I to argue.

We had quite a discussion about this. Basically, it comes down to a choice between accuracy and delay time. In real life, if an FO can't see the target, the whole spotting round procedure is either impossible or complicated to a greater or lesser extent. "Complicated" means there are ways around it that work in some situations, but with the cost of increasing the FO's delay time by a greater or lesser amount, and some additional number of spotting rounds, depending on the situation. "Impossible" means there's nothing for it but to take your best guess and hope for the best, accepting whatever inaccuracy you get. And depending on the situation, the FO might decide to just shoot in the interests of speed.

The problem is, CM's engine isn't good with all the shades of gray the above implies. When is a situation "very complicated" as opposed to "impossible", and when is it only "slightly complicated"? In addition, the current arty system can't handle some of the real world corrective measures that are used in "complicated" situations, like coordinated illumination and using smoke and airburst spotting rounds.

So, longer delay time or lots of inaccuracy? CMBO used the 1st system, CMBB uses the 2nd. Neither is perfect; in fact, both are almost equally wrong. But the CMBB method fit in better with some of the other changes to arty in the new game, and people had been complaining about a lack of scatter, so that's what we ended up with.

Steve has said that totally revamping arty is one of the major goals for the new engine. I really hope so because the current system will never be very realistic. Changes toward realism in 1 area create equal and opposite reactions elsewhere. But for now we have to wait.

[Kwacker]

Interesting comments - thanks Bullet head.

[Goanna]

Thanks for your comments Bullethead. I was waiting for you to comment on an arty thread prior to having an unsupported rant. I wanted to see if the delay times are realistic, because they seem not.

For instance. I have LOS with a 105 spotter to a large building to the north with a field behind it and east that has vast hordes of Russians sneaking through it in August prior to harvest. I can see the buggers go in since I am on a hill about 1 km away, but not once they get in there. Since I can see the big building I figure I will have the spotter target that, then get on the blower and have battalion walk the rounds 100 m away into the field after they are landing and I have adjusted for accuracy. Sound realistic?

Unfortunately, the CMBO engine says that when I try the walk move, I have to wait a whole 7 minutes (!) for the adjustment even though the spotter is right there on the phone, can see the initial rounds land and all I want is a shift of no more than 100 m.

It's experiences like this that make me question the efforts that have been made to make the game more realistic. Basically, as I see it currently off-map arty is only good for prep barrages.

This and a number of other "added realism" effects have seriously made me consider what dalem's was on about in another thread. What I think he was trying to say is that it may be more realistic, but is it more fun? Maybe I should stop upgrading my CM and just stick to CMBO because in many ways, I think that game was more fun to play.

[Bullethead]

Goanna said:

For instance. I have LOS with a 105 spotter to a large building to the north with a field behind it and east that has vast hordes of Russians sneaking through it in August prior to harvest. I can see the buggers go in since I am on a hill about 1 km away, but not once they get in there. Since I can see the big building I figure I will have the spotter target that, then get on the blower and have battalion walk the rounds 100 m away into the field after they are landing and I have adjusted for accuracy. Sound realistic?

In real life, the FO would all for fire on the wheatfield directly and this would NOT be a blind shot for him. This is because the shellbursts stick up WAY higher than the grain, so he can see exactly where he's hitting in relation to where he wants to hit. It matters not that he can't see the dirt below the grain, or the soldiers crawling along down there. He can see the top of the grain, he can see the shellbursts above the grain tops. That is no different from seeing shellbursts above the short grass of open ground tiles.

This is one of the most basic problems with the CM's arty system: units can't "see" shellbursts in the same way they can see other units. This might not seem like a big deal, but it is. The entire system of spotting rounds and all that is totally dependent on the FO seeing where the shell lands in relation to the desired target point. So if that visual process isn't modeled in the game at all, there's no way to make an accurate simulation of FO-controlled artillery. All you can do is come up with a set of abstractions which of course don't work in all situations.

Unfortunately, by not modeling actually seeing shellbursts but basing all FO LOS and delay/scatter determinations on whether he can see the ground, the abstractions in CM cause a lot of unrealism. Like your example. Or say the enemy is in a patch of trees 2 tiles by 3, surrounded by 200m of open ground all around. You can't target on the patch of trees, you have to aim at the near edge and hope the spread of the pattern reaches back far enough, knowing you're going to waste at least 1/2 your ammo on the open ground in front. This problem also makes it impossible for the game to do things like coordinated illumination missions at night, or using smoke and/or airbursts for spotting rounds over large forests or behind hills.

So yes, there are a whole bunch of inaccuracies in CM's handling of arty. The increased delay times for lack of LOS in places where that doesn't matter in real life is one such area. Unfortunately, there's no hope of fixing these things until the engine rewrite. I sure hope that does the trick--it might not, you know.

It's experiences like this that make me question the efforts that have been made to make the game more realistic. Basically, as I see it currently off-map arty is only good for prep barrages.

In WW2, arty WAS mostly useful only for barrages. Most arty fire in WW2 was pre-planned, either barrages or on-call at preregistered points (what CM calls a TRP). The US was the major exception to this, but no US on the Eastern Front . So making arty mostly useful for barrages and TRPs, especially on the Eastern Front, is a big realism enhancement. It might not go about achieving this in a tidy, perfectly realistic way, but it does achieve it.

So get used to it and change your tactics and unit purchase habits. Do what they did in real life. Don't get so many arty FOs and use the few you get for barrages. Use assault guns and IGs for suppressive fire on point targets that pop up during play.

it may be more realistic, but is it more fun?

It's not fun if you keep using CMBO tactics and continue holding CMBO opinions on what you should buy and how you should deploy it. That will get your ass kicked muy pronto. OTOH, if you adapt to your new environment, you'll find CMBB at least as much fun as CMBO.

[JonS]

Originally posted by Bullethead:

... Most arty fire in WW2 was pre-planned, either barrages or on-call at preregistered points (what CM calls a TRP). The US was the major exception to this, ...

Grr. You know better than that Jim

[Goanna]

Yeah, very good point Bullethead. I have already started changing my buying tactics (buying mostly on board mortars for suppression) and only using FOs for prep or timed barrages. But with all the changes, I just find CMBB much more frustrating to play, and hence am less likely to keep playing it like I did CMBO. Maybe the game as a whole is just growing old for me or something like that. Anyway, thanks for your your always valuable FO perspective.

[Cpt Kernow]

Goanna

CMBB id frustrating if you choose to play it like CMBO. I at first had similar experiences untill I realized I was playing an entirely new game. I would never consider going back to CMBO now as it does not provide the same SATISFACTION when you pull of a succesfull attack. CMBB is slightly more cerebral than CMBO but still provides all the visceral moments of excitement and tension.

Keep pushing and your perservernece will pay off.

[Pascal DI FOLCO]

Originally posted by Bullethead:

Goanna said:

</font><blockquote>quote:</font><hr />For instance. I have LOS with a 105 spotter to a large building to the north with a field behind it and east that has vast hordes of Russians sneaking through it in August prior to harvest. I can see the buggers go in since I am on a hill about 1 km away, but not once they get in there. Since I can see the big building I figure I will have the spotter target that, then get on the blower and have battalion walk the rounds 100 m away into the field after they are landing and I have adjusted for accuracy. Sound realistic?

In real life, the FO would all for fire on the wheatfield directly and this would NOT be a blind shot for him. This is because the shellbursts stick up WAY higher than the grain, so he can see exactly where he's hitting in relation to where he wants to hit. It matters not that he can't see the dirt below the grain, or the soldiers crawling along down there. He can see the top of the grain, he can see the shellbursts above the grain tops. That is no different from seeing shellbursts above the short grass of open ground tiles...

</font>

[Bullethead]

JonS said:

Grr. You know better than that Jim

OK, OK, the Brits and CWers were an exception, too. Happy now?

Anyway, when do you want to start discussion CMMC arty rules changes in like of CMMC #1 experience and CMBO?

[Bullethead]

Pascal DI FOLCO said:

Maybe there is a simple solution to that FO-LOS problem : if the engine was tweaked to establish FOs LOS some 3-5 meters ABOVE GROUND, instead of ON the ground, it'll "realistically" reflects the LOS to a SR. The FOs will then be able to target across a wheatfield, but still not across a wood or hill.

Yeah, I've been thinking of something along those lines, to keep from having to actually do LOS checks on shellbursts. My idea is as follows:

Each HE shellburst would have a height number assigned to it. Smoke shellbursts already have this for the existing LOS system. Then, when the FO tries to target a point, the game assumes the actual height of the target point is whatever the gound elevation is plus the height of the shellburst. If the FO can trace an LOS to that point and that height, his fire would be as if he had an LOS.

This could be a 2-step process. Say that the FO's HE bursts aren't tall enough, then the game checks the taller height of the smoke bursts. If the FO can see the top of the smoke, then it's still like he has an LOS for scatter purposes. However, the spotting rounds are fired as smoke instead of HE, and the time between spotting rounds has to be at least equal to the time it takes between shell impact and smoke plume growth.

[Sitting Duck]

Originally posted by Kwacker:

...(M)aybe guys really do commonly do the stupidest things when their pants turn brown. However, that is not the impression I get from the accounts I have read...

Perhaps this is because the fellows that did do the stupidest things did not survive to write an account of their actions.

[Michael Emrys]

Originally posted by Pascal DI FOLCO:

Maybe there is a simple solution to that FO-LOS problem : if the engine was tweaked to establish FOs LOS some 3-5 meters ABOVE GROUND, instead of ON the ground, it'll "realistically" reflects the LOS to a SR. The FOs will then be able to target across a wheatfield, but still not across a wood or hill.

Not a good approach IMO. Then the FO would be too easy to spot and shoot at in return. Bullethead is on the right track that shell bursts should be modeled at their correct height.

Michael

[Kwacker]

Originally posted by Kwacker:

...(M)aybe guys really do commonly do the stupidest things when their pants turn brown. However, that is not the impression I get from the accounts I have read...

--------------------------------------------------------------------------------

Perhaps this is because the fellows that did do the stupidest things did not survive to write an account of their actions.

Well the vast majority of combatants did not write about their experience so I guess they must all be dead

[Pascal DI FOLCO]

Originally posted by Michael emrys:

</font><blockquote>quote:</font><hr />Originally posted by Pascal DI FOLCO:

Maybe there is a simple solution to that FO-LOS problem : if the engine was tweaked to establish FOs LOS some 3-5 meters ABOVE GROUND, instead of ON the ground, it'll "realistically" reflects the LOS to a SR. The FOs will then be able to target across a wheatfield, but still not across a wood or hill.

Not a good approach IMO. Then the FO would be too easy to spot and shoot at in return. Bullethead is on the right track that shell bursts should be modeled at their correct height.

Michael</font>

[Michael Emrys]

Originally posted by Pascal DI FOLCO:

Ok, the full modelling is certainly better ...

Not to mention conceptually free of the unhappy metaphysical entanglements.

Michael