Build 4.6.273 Feedback

[Arjuna]

Just to confirm I have not tweaked the cas rates from arty.

[dazkaz15]

Thanks Dave.

Let me know if you need a save of the exit being abandoned.

I'm off to bed now though, and work early tomorrow, so you won't get it till tomorrow night.

Good night guys.

[BletchleyGeek]

ORIGINAL: RangerX3X
There was some sketchy attack pathing going on (using the default settings for an attack on Breitfeld Crossroads the AI took an avoidance path way out of the way, almost 2km north and east of the objective).

Tim,

when the distance between the HQ and the attack location is greater than 3 kilometers or so, only flanking courses of action are considered. Besides that, heavily armored units - like the 51Bn - are very averse to cross open terrain where the enemy can project Anti-Armour firepower (as is the case, there's the 75mm PaK and the StuGs right between the 51st Bn and the objective. So in this case, the AI is working out a quite sensible path - avoid the enemy positions and attack from a flank.

Indeed, what the AI hasn't done is to first reconnoiter the approach to the FUP. That would be a very nice capability to add to our planning algorithms, but indeed, you can do that yourself.

Notice as well that plans are worked out according to what the AI knows, not what you do know.

[BletchleyGeek]

ORIGINAL: Mahatma

Arjuna or Bletchley,

When you tweak the code as you did with the assault and road column problem, does this also improve the AI we fight against? So that every slight improvement makes the game more challenging to the player, which is really very interesting because I'm guessing the ultimate goal is an AI that no human player could ever beat. Not even Dazkaz micromanaging every unit to death.

The AI that moves your units is made up of the same algorithms as the enemy AI. There are some SOPs - like the recent In-Situ Defend after Attack - which the AI isn't programmed to consider, or we haven't got the time to find out how to encode a clever usage into a workable heuristic algorithm.

Sorry, but the goal isn't to build Skynet. There are plenty of Doomsday scenarios already, and I reckon we don't need to add to them [:)]

[jimcarravall]

ORIGINAL: Bletchley_Geek

Notice as well that plans are worked out according to what the AI knows, not what you do know.

Something to consider.

In many instances, the terrain maps available to the human player are at a significantly improved level of detail than the RL forces had at the start of a battle.

How are differences between what was "recorded" or openly available to any individual prior to battle and what really "exists" once entering battle resolved?

FOW doesn't only apply to what is directly facing your force, but also to what your force has to address once if moves from a known position to an unknown location.

[BletchleyGeek]

ORIGINAL: jimcarravallah
Something to consider.

In many instances, the terrain maps available to the human player are at a significantly improved level of detail than the RL forces had at the start of a battle.

How are differences between what was "recorded" or openly available to any individual prior to battle and what really "exists" once entering battle resolved?

FOW doesn't only apply to what is directly facing your force, but also to what your force has to address once if moves from a known position to an unknown location.

We don't have terrain FOW, Jim. So the approach recon mission I mention is to ensure that there are no threats along the route. At the current level of detain for our terrain model, I think it would make sense to have FOW for fords, the state of bridges, and tracks - which are hardly marked on any maps.

Something like this would be technically feasible, but it would convey a significant programming time investment to rework our data structures for the terrain to 1) be different for each of the sides, 2) allowing it to change dynamically and 3) program the AI to handle such changes. One of the open technical challenges in the engine is to manage in a transparent - to the player - way the changes implied by bridges being blown or built.

At the moment we need to stop the show while we rebuild the relevant data structures - this is a multi-threaded real-time application, and we need to make sure that nobody is accessing terrain data while we're recalculating it. The problem is that the time required to perform this computations is quite big - depending on the map and your CPU taking anything between 1 sec and 1 minute or more. Before 4.6.259 we didn't handle this, and we had very nasty crashes which had Dave and me busy for a week.

[Arjuna]

ORIGINAL: dazkaz15

Thanks Dave.

Let me know if you need a save of the exit being abandoned.

I'm off to bed now though, and work early tomorrow, so you won't get it till tomorrow night.

Good night guys.

Yes please.

[decaro]

ORIGINAL: dazkaz15

Joe D
I have just managed to replicate your "TF ... abandoned its Exit task for lack of time,"

It happens after the initial orders delay if the unit is a long way from the objective i.e. more than 4 hours away via its chosen route, if you leave the end time at the default of 4 hours.
Incrementing the end time to a more realistic time solved this for me.

Can you confirm that?

Yes I can confirm that, and pushing back that end date/time made all the difference, but here's the rub: I strongly suspect the AI is having the same Exit problems!

Playing against the AI Axis in the Battle of the Roadblocks scenario, I have been hard pressed to defend Noville and Foy most probably because the Axis Exit at Bertogne also becomes available to the AI for lack of time, so the AI sends all Axis units SW along the Noville-Foy route towards Bastogne.

What do you think?

[BletchleyGeek]

ORIGINAL: Joe D.
Yes I can confirm that, and pushing back that end date/time made all the difference, but here's the rub: I strongly suspect the AI is having the same Exit problems!

Playing against the AI Axis in the Battle of the Roadblocks scenario, I have been hard pressed to defend Noville and Foy most probably because the Axis Exit at Bertogne also becomes available to the AI for lack of time, so the AI sends all Axis units SW along the Noville-Foy route towards Bastogne.

What do you think?

I think you're right, Joe. If there's a problem with the timings estimation for you, there should be the same problem for the AI (especially after Dave unified how timings are handled all across the board).

[Renato]

ORIGINAL: dazkaz15

Ref the orders cancelation as I understand it, please correct me if its inacurate:

When you place a move order (note I am only talking about a move order, not attack, defend, or exit order, which are handled differently) on the map, the end time of that move order is estimated to take into account the command delay, the time taken to move to that location at normal speed, and a small contingency time is allowed to take into account minor holdups.

If your move order should encounter enemy, on route it starts to move more tactically depending on the weather conditions, the slower it will move in close proximity to the enemy.
This will alter the time it takes for your unit to get to the ordered location, and it may run out of time.
Move orders do not slip like attack orders.
This may have been implemented like this to let you know that the move has run into a problem, because when the order expires you get a message saying move complete, and you can also filter the units to see which ones have not orders, or it may have been to prevent a move from running into problems but making no progress until the end of the scenario.
Would need dev confirmation to confirm why it does not slip indefinately, because I don't really know myself, these are all just guesses.

Anyway this problem can easily be solved, all you need to do is make sure you increment the estimated end time of the move to whatever you want, before you want the unit to give up.

I personally just increment it a few days, as this is just a few clicks in the day box.

Joe D
The same sort of thing happens for an exit order, except for some reason its just gives it a set time of about 5 hours to reach the exit.
All you need to do it increment the end time for the order to the end of the scenario to ensure it never expires.

Now there is a quirk to this.
If you place a move order then increment the end time, if you then place another move order to extend the distance of the move it will re-set the end time again, so you will need to increment it again to a time that you are happy with for it to expire.

All this is off the top of my head, from experience, I have not had time to do tests to confirm what I have written here.
I will do so as soon as I am able.

dazkaz15,
I replayed the Hoefen Ho-Down scenario incrementing the end time of all move orders to the end time of the scenario, and I had just one cancellation near the end.

I think you have really find something here, even if I find it very strange.

Certainly I didn't expect anything less from the hero of Elsenborn Ridge, who attained an almost incredible decisive victory! [:)]

[BletchleyGeek]

ORIGINAL: Renato

I replayed the Hoefen Ho-Down scenario incrementing the end time of all move orders to the end time of the scenario, and I had just one cancellation near the end.

I think you have really find something here, even if I find it very strange.

Okay, now I get to understand a bit better what's the matter with this "orders cancellation" issue. Thanks, Daz.

Renato: as far as I can recall we only automatically move forward end times for attacks. The End time is assumed to be a 'hard' deadline - it can't be postponed. Back in May 2013 we had a discussion about this topic, and we concluded that we needed at some point to add the ability in the UI so the player can indicate what is his intent regarding the end-time explicitly, for any kind of task. We also reckoned that the implications of such a change would be too big to handle within the timelines we were considering for ending development of CO1. So we took the compromise of enabling the slipping of attack tasks - that came out with patch 4.6.262.

If we hadn't looked back at this is because the frequency of complaints regarding this issue went down considerably after rolling out this change.

I know it's busy work to go over the End Times and adjust them further, as it introduces micro-management we would rather not to have. But that's the way things are going to be until CO2 comes out, I'm afraid.

[Arjuna]

OK I have looked into the EndTime issue for player created tasks. I have modified the code so that Exit tasks will default initially to scenarioEnd. however these will be overridden later in the code once the AI knows a bit more. At that second stage it will endeavour to find a nearby side objective within its radius and if it does and the objectiveEnd > timeNow then it will set the endTime to the objetiveEnd. Eg if you have an Exit objective that ends D2, 2000 - ie is no longer active - but the scenario doesn't end till D3, 0600, then the end time for your order will be set to D2, 2000.
 
I have also changed the default durations for other task types. Defend, Delay, Deny Crossing and Construct Bridge default to scenarioEnd. Rest and Reorg now default to 6 hours duration while other tasks default to 12 hours duration. Further I have made sure that Withdraw tasks are treated like Moves in that it will calc a route and use its duration to determine end time.
 
This I believe should rectify most of the issues relating to order timings. I will put out a new hotfix tomorrow but this will definitely be the last.

[Mahatma]

'...will definitely be the last.' haha didn't you say something similar a few days ago?

I've played a few scenarios and overall the patch is damn good. But if I may make a suggestion or two:

a. Units wake up at 0800, losing the benefits of darkness. Why not have them wake at 0600? At present it's up to the player to micromanage the rest parameter for every order if s/he wants an early morning or night attack as units by default sleep 2200 to 0800 or thereabouts.
b. I have a suspicion that the AI units recover much much quicker than friendly units after routing. A friendly unit will sometimes rout for an hour and be in rout recovery for 6 hours. But AI units will rout and recover like they did in previous patches, i.e. quickly. This is probably due to the new 10% rule regarding casualties because units that begin the scenario at ~20% strength don't seem to have a problem when at 19% strength, even with only 19 men. But a unit which goes from 100% strength to 80% or 200 men to 160 men will sit and be in retreat recover for literally hours but an enemy HQ rout and recover in 1 minute flat (yes I've seen this.)
c. A unit that is in retreat recovery will not process any orders at all it seems. Even when the order is to withdraw to a safer location. So in effect they'd rather do 'retreat recovery' than 'retreat' when the order is clearly a good idea.

[Arjuna]

Mahatma,
 
The AI doesn't have faster recovery times. For me to track this down I need a save taken before the AI controlled unit changes from rout recovery to normal and another after they have done so ( or a sceeenshot showing the time they changed ). Ideally I would like a save when they went into rout mode as well.

[Arjuna]

c. A unit that is in retreat recovery will not process any orders at all it seems. Even when the order is to withdraw to a safer location. So in effect they'd rather do 'retreat recovery' than 'retreat' when the order is clearly a good idea.

Well all unit process receive orders, regardless of rout status. What they do however after getting the orders is another thing. Are you telling me that if you give say a withdraw order to a retreat recovering un it that its Info box doesn't go pink (indicating it is processing orders)? Or that it does go pink but the unit stays recovering and doesn't move?

Let me say this that while a retreat and retreat recovery is a controlled action it's not one from which you can simply say stop all that and move south instead and expect to get an immediate response. We are talking WW2 here where at best you may have a walki talki down to Platoon level. But by and large where most instructions were delivered face to face and where commanders had either get their subordinates to come to them or they went in turn to where their subordinates were. All this takes time and when those subordinates are involved in a stress action like a retreat its difficult for them to then also go around and give their orders to their subordinates. Retreats occur because the force is engaged with the enemy and soldiers naturally have a tendency to focus on the immediate threat.

To model this we do impose a certain time for units to recover from retreat before they can get moving again. I think that's reasonable.

[Mahatma]

Arjuna,

Thanks for the explanation. In my head all my companies have a radio guy like in Vietnam so it's nice to have my ignorance removed. I suspect part of my problem is due to inexperience and low morale of my infantry.

Question for other players:

How do you find the retreat mechanic? I'm sort of going out of my mind about it but no-one else has mentioned this apart from Repsol, making me think it's all in my head.

So, do your units recover from retreating or routing in decent time when they lose more than 10% of the starting number of men? What about 20%? Many of my companies, once crossing the 10% threshold, will retreat for hours and hours.

Oh and artillery! OMG horribly effective (not a complaint). Try this: take your own artillery, a sizable amount so that the bombard value is around 500-1000, and bomb your own units for fun. I've seen bombard=600 cause 20 casualties to an entrenched infantry company in 20 minutes, which with the threshold mechanic means that that company will now do almost nothing useful for the next 6 hours. Imagine how effective this is for breaking up infantry assault in the open! 500 bombard + ROF max for 10 minutes vs a moving target is probably enough to retreat almost any infantry unit (I've not tested this). And that's all the Fuhrer gave me to take the bridges east of St Vith.

[dazkaz15]

I have still not been able to download the latest beta from the members section, so not played the latest one yet.

I could see from the last one that this was probably going to be a problem though.
The artillery has a vastly disproportionate killing power compared to any other unit. I have read somewhere that is was responsible for 80% of casualties.

The problem is, that its very hard to decide if its too powerfull, because every historic report form the Battle of the Bulge that I have read has said how instrumental it was in the outcome of the battles.
Especially the Allied artillery.
It even says on the WIKI about the German troops refusing to leave cover, even when ordered to because of it.

There is not much mention about how effective the Axis artillery was though.
This may be because the Axis were on the offensive, the Allies were in good defensive positions a lot of the time.
Also their fire procedures were not as effective.
Another reason might be that the Allies had a lot more AFV's than the Axis, when they did switch to the offensive, so had some protection from shrapnel when they did get caught out in the open.

[Arjuna]

I uploaded to Matrix earlier today a new hotfix build 4.6.274. I reviewed and tweaked the retreat code. In fact I rewrote the Retreat(). I have tested it a bit and it looks good to me. But there are always going to me anomalies. It will never be perfect. AI is never perfect. It can always be improved.
 
As to artillery. Yes it was referred to as the queen of the battlefield because in chess the queen is the most powerful piece. Those stats about 80% really refer to shrapnel rather than just limited to arty. A tank fire a HW round and you get a lot of shrapnel. A grenade produced a lot of shrapnel. So 80% is too high a proportion. But one think I have been doing a lot of is paying close attention to cas rates and their causes in the game. If you run the tutorial through say 20 times and take a look at the AAR cas rates you will see that arty accounts for around a third on average, sometimes more, sometimes less. If you think it is brutal well your dam right. The German SS attacks on Elsenborne Ridge were stopped dead by the massed US arty on the northern shoulder. After the first wave was almost annihilated it took a particularly brave or foolish soldier to put their hand up for a repeat performance.
 
The main reason the German arty was less effective against the US counter attack was the bulk of their heavy guns were immobilised back in their starting locations, they were no way close to be being able to resupply their guns like the Americans could and they lacked the doctrine for rapid coordination of massed arty. They could do it for set piece operations but they couldn't match the US arty in mobile ops. I might add that the British really were at the top of the pile when it came to rapid massing of fires in WW2.

[dazkaz15]

(Written by Maj. Charles E. Rousek, Exec. Officer of the 38th eff. Aug 8, 1944)

At 0535 a tremendously heavy artillery and rocket barrage hit the "C" Troop positions and the Squadron
CP area, and then rolled back to hit "B" Troop and "F" Company on the Mutzenich Hill.
This barrage was followed at 0600 by the attack of the 1st Battalion, 751st Regiment, 36th VG Division on "C" Troop's right
and the 2nd platoon, "F" Company, S/Sgt. Bielicki commanding, defending the Roer Valley in Monschau.

The spearhead of the attack was completely smashed by canister fire of the tank platoon delivered at a
range of 50 yards. A "C" Troop Patrol under Tec. 5 Van Order broke up other attacking elements, as they
emerged from Menzerath, with artillery and mortar fire which he called for and adjusted by radio.

When "F" Company's mortar illuminating flares revealed more enemy approaching up the valley, Sgt. Bielicki
requested artillery defensive concentrations which broke up this attack. However, the enemy reorganized and attempted to infiltrate into the tank positions.
Sgt. Messano seized an anti-aircraft light machine gun and placed it in position, dismounted, to block for good this attempt.
Continued "C" Troop mortar fire and artillery fire finally broke up concerted enemy activity in the valley, except for sniper fire all day.

Meanwhile, the enemy launched another attack at 0830 down the winding road, which led from Imgenbroich to the north end of Monschau, known as the "snake road".
This attack was first brought to a halt by machine gun fire from "C" Troop's 2nd and 3rd platoons, and then smashed by fire from the 62nd
FA Battalion and "E" Troop (M5 75mm Howitzers).

German patrols were observed and fired on all parts of the Squadron line during the rest of the day.
When an observer reported that all the enemy pillboxes were being loaded with extra men, it was plain that another and heavier attack was due to come.
Reinforcements were requested of Group Headquarters, which procured Company "A" 146th Combat Engineering Battalion, Captain Ball commanding.
One platoon was given to each of "B" and "C" Troops to bolster their line, while the 3rd platoon was dug in on the hill behind Monschau to cover the front line positions with overhead fire and to
protect our right flank. Six machine crews from the 186th FA Battalion, and attached TDs were also put
into the line with "F" Company's hard-pressed 2nd platoon.

The night of the 16th and 17th was marked by much activity.
As many as five enemy searchlights lit up the Mutzenich Hill and a score or more large low-flying planes flew westward over the Squadron line.
These were the planes which dropped LT. Col. Von der Heydre's Parachute Battalion behind the Squadron line At 0400 a "C" Troop listening post under Tec. 5 Allen Anderson on the Merzerath Hill detected a force of 60 or 70 enemy approaching their post.
They called for and received fire from "E" Troop (M5 75mm Howitzers) within two minutes, which broke up the enemy endeavor.

At 0615, another heavy barrage struck the Squadron line, even reaching back to the "C" Troop mortars this time.
At 0630, the enemy launched his initial attack to pierce the line in "C" Troop's left.
The enemy, having assembled in the draw east of Mutzenich during the night, attacked "F" Company's 1st platoon at the railroad track, but were finally driven off with machine gun fire, and punished severely by artillery fire as they retired into the draw.
At 0800, the full weight of the entire 751st regiment struck the Squadron, as one attack coordinated with the initial attack just mentioned was made down the "snake road" while the main effort was made in an all out assault on the Mutzenich Hill.
The attack down the "snake road" was met by machine gun fire from "C" Troop's 3rd platoon, but kept coming.
It was finally stopped by tank and TD fire from "F" Company's 3rd platoon whose weapons overlooked the "C" Troop position.
Artillery fire further broke up this attack.

As you can see Artillery and mortar fire was instrumental in breaking up the attacks historically in the Hofen scenario.
You can read the full account here:
http://117th-cav.org/History%20of%20the%2038th.pdf

This link was originally posted by Simovitch here:
tm.asp?m=3546563

[dazkaz15]

ORIGINAL: Mahatma

Arjuna,

Thanks for the explanation. In my head all my companies have a radio guy like in Vietnam so it's nice to have my ignorance removed. I suspect part of my problem is due to inexperience and low morale of my infantry.

Question for other players:

How do you find the retreat mechanic? I'm sort of going out of my mind about it but no-one else has mentioned this apart from Repsol, making me think it's all in my head.

So, do your units recover from retreating or routing in decent time when they lose more than 10% of the starting number of men? What about 20%? Many of my companies, once crossing the 10% threshold, will retreat for hours and hours.

Oh and artillery! OMG horribly effective (not a complaint). Try this: take your own artillery, a sizable amount so that the bombard value is around 500-1000, and bomb your own units for fun. I've seen bombard=600 cause 20 casualties to an entrenched infantry company in 20 minutes, which with the threshold mechanic means that that company will now do almost nothing useful for the next 6 hours. Imagine how effective this is for breaking up infantry assault in the open! 500 bombard + ROF max for 10 minutes vs a moving target is probably enough to retreat almost any infantry unit (I've not tested this). And that's all the Fuhrer gave me to take the bridges east of St Vith.

I don't know if you have seen this yet, so here is the link.
tm.asp?m=3360428

You won't like me for this, but you get much better results if you micro manage the troops individually into good cover.
Concealment is even more key than good cover I have found.

If they can't see you they can't bombard you.

Attacking with soft infantry only at night allows your units to be to close to the enemy for them to be able to use artillery against you.

Also managing your own artillery to suppress any units that are able to observe you during an attack makes their spotting of artillery fire less accurate (I did read that somewhere a while back, but would need Dev confirmations to be 100% sure).
Never attack a dug in unit across open terrain during the day, unless its heavy fog, you will just get annihilated.
If you have to attack during the day, you need to put in as many diversionary attack in at the same time as you can manage, to dilute the effect, and hope that the enemy AI does not realise which is your main effort [;)]
Good luck mate.