CO2 BEDA FOMM Quick AAR

[john connor]

Beautiful, isn't it? All those lovely attack markers.... That's the AI putting together my attack. The Axis took Manhay briefly (without an attack - my own forces fled in the night) and now I'm having to attack to get it back. So I can't run the experiment there.

I might have to cobble together a quick test scenario to see if I can tempt the AI to put together a big attack.

 

[Daz]

There has been something else I have been wondering about and that is does the fog of war obscure the correct intentions of the enemy units?
From my experience I would say it does.

By this what I mean is, do we see a move indicator on the expanded info boxes on the side of the counters, where in actual fact they are attacking, but because of fog of war we are given false information.
When you see the enemy info boxes that are close and are rated as excellent Intel these we see as attacking, but more distant ones that have move indicators may well be part of that attack, but Intel on them is poor.

We know from past posts that after a surrender, fog of war is still in place to some degree, so are the move/attack indicator info boxes being updated to show the actual orders of the enemy units at the time of surrender?
If not then it would be hard to distinguish which units were part of an attack, and which ones are not.

In the Beda Fomm scenario I have seen many units that are indicated as moving, but are in fact in successive line (a favourite attack formation of the AI as I'm sure you are aware), and not in Road Column (a favourite formation of the AI for a move).

 

[john connor]

So here we go. I set up a strong collection of Allied forces in Fortified positions with a single objective on the map. An 11 hour scenario. I maxed out the Axis SS units - all 100%, leaders and units, no fatigue, 200% fuel, ammo etc. Didn't check command load, but I'm guessing it's ok as this is a Division. Divided VP points between the single objective and killing the enemy. Gave all the allied troops defend in situ orders. There are NO indirect fire units on either side.



Certainly looks like a strong combined arms attack, possibly at a Regimental level! At the very least there are 5 AI units attacking at the same time. And this was proceeded by enough ranged direct fire to already force the fortified defenders off the objective. As as aside all the defenders have had their supply blocked and suspended, for reasons I don't understand, since their bases and SEP are within that perimeter. But that's a different issue. (As is their evacuating their superb fortified positions when fired from 2km distance.) But the main thing is, the enemy AI is putting in a big (ish) co-ordinated attack! Hoorah! All is not lost.

The first time I ran this through, however, NOTHING happened. I mean nothing. The Axis got so stuck it didn't even move near.

So, I'm now more than ever thinking that Daz's theory about bugs affecting the AI's attacking capacity must be correct. Which means if they're fixed for the friendly AI then they will be fixed for the enemy and we might then be able to return to a more aggressive AI performance. I hope so. But it means, I think, that as well as everyone reporting all the CO2 UI issues, the lock up bugs and other behaviours affecting attacks etc are all noted and reported to Dave.

 

[john connor]

I'm sure you're correct about FOW, Daz. I wish I had whatever software Dave uses to be able to see both sides without FOW. It would make testing more accurate.

In the little test scenario I ran above you might think that with a full division immediately available to them, and on maxed out ratings, the enemy AI could do something even more aggressive. I do think that, actually. I think, if FOW was lifted, I might see that the AI was mounting a Regimental size attack.
But then again, it has to attack over completely open ground. But then, if a Division on maxed out aggression and stubbornness won't mount a powerful attack over open ground, then there's no chance once we put in realistic stats. And those kinds of attacks did happen. When I put together the Kursk scenario the struggle was to get the AI (as Soviets) to do just that, because the soviets did do that, in real life. But maybe the AI is more intelligent than the soviets were. We know what happened, after all...

I feel much happier about the AI having got it to attack as above. But I still think that it's not ideal that to get that to happen everything has to be maxed out. It means all scenarios should really be to play as one side only.

What I would have liked to have seen - I throw this out for consideration, because it may be that it's just unrealistic of me - in the above scenario, is the AI assemble the massive forces at its disposal somewhere out of range and sight (around Manhay, say) then put in a really co-ordinated massive attack. Then, if you start turning the stats down from 100% you should see more realistic and cautious behaviours. No?

I think it's really important that we try to get Dave to fix all the bugs that have been mentioned affecting our own attacks etc, as this will perhaps work wonders on the enemy AI.

 

[Daz]

This is a surrender pic playing as Allies of an Axis Bn sized attack.
Do you think 1.2 is part of the attack or is he just moving?
Its very hard to tell.

 

[Daz]

Another image exactly the same time before the surrender with the Current Intel filter selected.

 

[Daz]

Don't be discouraged in what you are doing here Peter, I think your doing an excellent job in bringing this to our attention.

I am just playing Devils advocate a bit here to put other possibilities forward.

If you remember right I did the exact same thing when you brought up the supply issue that we had in the game most noticeable in From the Meuse to the Rhine, and you were spot on with your assumptions (gut feeling) for that as well!

 

[john connor]

Well, I'm very encouraged that there was a 'large' attack in the test scenario I ran (5 posts above this one).

It's difficult to set things up to test my gut feelings (that not enough happens) when (a) there's FOW, (b) there are genuine tactical considerations to contend with.

BUT. It's the most sophisticated AI (friendly and enemy) out there and so, my feeling is that it's designed to try to imitate reality - at least that's what Dave is trying to do.

So, it SHOULD be possible, with the Kursk scenario I did, for example (and only as an example), to have the soviet side with overwhelming numbers in men and materiel (as in reality), but with their quality overall poor, and the Axis side with a much, much smaller but higher quality force, and the actual historical objectives set, and the AI will very roughly, then, do what the individual sides did in reality. But right now that just doesn't happen.

Or, with Beda Fomm (to take another example) it should be possible to have an overwhelming superiority in men and materiel, of a generally poor quality (the Italians) contested by a much smaller force of higher quality (the Allies), with the actual historical objectives set (exit points) and the result to be that the AI can do what the Italians did - put up a brutal fight. The Allies won, but it was close and brutal. The stock scenario is roughly set up along these lines and what happens appears to be that the AI as Axis becomes crippled and unresponsive.

That is also what happened when I set up the Kursk scenario along historical lines. To get different results the stats for the AI side had to be maxed out and a lot of thought put into trying to script attacks through the back door, as it were, by clever placement of objectives.

I think that this kind of sleepy AI is also what I have seen (and posted about) in, for instance, Maas-Rijn (where it should be possible to have the Axis (as AI) completely crush the paras in Arnhem (if the player just sits back, at any rate), but where what you end up seeing is masses of Axis forces surrounding the paras and then never using their strength to mount effective, strong, coordinated attacks at, say Regimental level. There are attacks, but they're too small to make a diff.

I saw this also in Spearhead v Reich, where, on surrender (and at the end) there were huge assemblies of Axis forces that were just not doing anything.

So that's what leads to my gut feeling that something might need tweaking.

 

[Daz]

Pete, could you put together a Beda Fomm Scenario, where you replace most of Italian infantry with period Axis Armour Coy's, as I mentioned in a post earlier?
I would do it myself but I can't be bothered to learn how to do it at the moment, as I'm more interested in continuing my St Vith AAR once a few more of the major bugs are fixed.
I would be incredibly interested in results.

As I was saying earlier when I read about the historical battle, it mentioned the importance of tanks in an attack over terrain like this and I was just wondering how things would differ.
Also what difference it would have made if the tanks were committed to the battle much earlier, so enter them right after the recon Bn if you can?

My biggest fear is the AI will rush them straight to the exit objective, leaving everything else to fend for itself.
So maybe give the Axis the same objectives as the Allies to ensure they keep the road open for the rear echelon guys?

Something else that has been bothering me is the amount of emphasis the AI puts on exiting.
I was wondering if instead of the exit objectives removing the units from the map, there should be an Assembly Area objective where the number of men and equipment inside this area is added up at closing time of the objective.
This way I think would be a bit more realistic as instead of the unit suddenly disappearing they will hang around to help defend the area for others to exit.
This isn't a problem for us as players because we can put in place a rear guard to keep it open and/or continue to keep the force in play until the last minute, but the AI will rush them off the map as soon as it can substantially weakening the forces left on the map.

 

[john connor]

It would work out better, Daz, I'm sure. But it's a little more complicated to transport and place a new OOB onto the map and I think I'll have to call a halt to messing around with scenmaker now to exercise these issues. Dave has no time anyway right now to do anything but sort the UI. With the time I have available I thought I would try playing a few more of the new scenarios instead, so I've just started with Brazen Chariots, which, in any event, looks ideal to test what you're saying as it's Axis armour against Allied with just one objective. I'll let you know how it goes!

Peter

 

[john connor]

This isn't the place to be placing these pics, of course, but I started here, so I'll continue, for now...

New Build - .12. I ran the little test scenario again and look at this:



That is no less than 10 Axis units attacking all at once!!!! I'm really pleased. Looking good for this build, like there is certainly a possibility of an improvement to AI attacking behaviour. I'll test more.

 

[Daz]

Great work Peter.

I managed another hour playing Manhay Crossroads as Axis last night and I’m very pleased with the game at the moment.
It’s been very challenging, even when placing enough individual orders to keep the command load maxed out.
I have been having a real 'ding dong' battle over Manhay.
The Allies seem desperate to take it and it has changed hands several times.

My main criticism and I know you and Dave will both hate me for saying this, but I don't think the Allies are using their artillery enough.
They have a massive superiority in artillery but I have received very little in the way of bombardments, and the ones I have had, have not been sustained, taking away a huge advantage for the Enemy Allied AI side.

I think artillery is always going to be a hot issue, and Dave has had to adjust things both ways due to player pressure in the past. I'm sure he must be getting fed up with it, but at least he should know where in the code to adjust it now, with all the experience he has had in doing so

Having siad that it's no big deal, it just makes it easier for me to kick AI butt and I have been having a blast doing so

 

[john connor]

Great, Daz. Another test for the arty balance would be, I think, for you to play as Allies and see if the Axis AI can get anywhere near the historical position, which was that they got into both Manhay and Grandmenil and along almost to the Erezee exit area and managed to hold these positions for various periods, at least 24 hours, before being forced off, mainly by Allied arty!!! I haven't tested Manhay as Allies since the latest build yet, but I think it would be a good test for the AI. Manhay as Axis seems to play more or less historical to me - that is I have to play quite carefully to achieve the positions I've just described, but it is certainly possible. With more care and micro-management you can do better than historical. That's a good balance, I think - no? But from the Allied side I have worried about the Axis AI being able to do as well as the real life Axis soldiers. I will get round to trying it with this build, but I'm now doing that beta testing thing we were invited to do and have been trying to set that up today. You joining that?

 

[john connor]

Very quick run through of stock Manhay scenario with new build, playing as Allies and playing sloppily - a few obvious clicks here and there.
A pic from day 2 18.30:



It's tough for the AI, I think. There are a lot of objectives relative to the forces available. That has led to the armour being dissipated, perhaps - individual companies, as you see, have been tasked with different objectives. I would have a query about the groups of forces under the 3 yellow circles. These are significant concentrations that are doing nothing. One of them is holding a very minor objective at Oster, the others seem to be in reserve. But I would have thought the time for using the reserve, or most of it, is long gone. Still, there's nothing clearly terrible about the AI effort. As it stands it will head for a draw like that. What more could you ask, you might say?

Well, I used to play this scenario as Allies and in the beginning - the first day certainly - it was hard. the Axis ALWAYS took Manhay and Grandmenil before I could move and my base that was in Manhay ALWAYS got destroyed before I could get it out (as happened historically). So something has changed. I suppose, to replicate the position I used to be happy with I would only have to increase the Allies orders delay penalty by another couple of hours, then the Axis could attack before the Allies could even react (in a C&C sense - as happened historically). But that does mean that something has changed, I think. The Axis did attack Manhay and Grandmenil in this run through, but not with the kind of coordinated attacks that would be necessary. I think that's probably because the AI has too many objectives to consider. It may be that to get this scenario to work as well as it used to I would just need to remove the more peripheral objectives from the AI.

 

[Daz]

From the historical briefing, it says that the Germans started their attack on Manhay at about 21:00, just brushed aside two roadblocks and were in Manhay attacking the HQ column not long after that by the sounds of things.

If its a historical outcome you are after then I would say the Axis start to far back, and are hampered by the inability of the Motorized infantry to dismount and attack through the woods.

In my play through those two roadblocks took hours to brush aside, and Manhay was well and truly consolidated by the time I got the main road (red road) cleared to allow motorized units through.
If its that difficult for me to do playing carefully I imagine the AI has a very hard time of things.
Its actually quicker to get to Manhay via Oster.

 

[Daz]

I didn't make many saves, but this is a surrender image from early on in the scenario. D1, 22:30.
As you can see my attack has only just got underway. This was after the Coy I was using for Recon came up against resistance and I then moved the rest of the Bn forward to put in the attack on the roadblock.
The Allied HQ, and Base are well settled into Manhay already.

It will take several more hours to clear the road of them Roadblocks.

I must admit I never used my Panzers to assist though as they did historically.
They are resting in reserve waiting for the infantry to clear the road.

 

[Daz]

Another one at D2 01:00 two and a half hours later.
As you can see the roadblock is still there.
Hardly a case of brushing it aside!

 

[Daz]

I will get round to trying it with this build, but I'm now doing that beta testing thing we were invited to do and have been trying to set that up today. You joining that?

I think I am going to pass on it for now mate.
I would love to do it as it sounds very interesting, but I want to try and finish the St Vith AAR, and I don't have time to do both I'm afraid

 

[Daz]

Wow I must say I have just had a real eye opener as far as my understanding of the game goes, especially with regards to the capabilities of the latest build

I decided to run the game fast speed placing just 3 orders on the map to see what would happen.
As you can see they are very basic move orders, with allow attacks and one set to Bypass.
All the waypoints were placed to encourage the formations to advance up the main road.
I will post the situation at D2, 01:00 in a bit so you can see how they got on compared to my deliberate and well managed moves commanding at a lower level of command in my posts above.

 

[Daz]

This is at D2, 01:00 having given no new orders at all.
It was a delight to watch, with the AI bypassing through the gap in the forest, and putting in attacks where needed to brush aside the Roadblocks.

Look at the surrender casualty figures!
Incredibly light and no Tanks were destroyed at all even though the it is dark close terrain, probably because they were constantly supported by the infantry that were putting in attacks at the same time.

Quite an incredible performance really, compared to my more managed effort at this stage.

I did however focus on more objectives to the west in my game, and occupied what I considered to be key locations for the ensuing fire fights come daybreak.

Note how the AI has put in no bombardments at all for either side, reflected in the surrender casualty figures.
A very poor performance by the AI artillery.