Testing UPDATE5 - 1.0.88 - ALPHA

Post bug reports and ask for game support here.

Moderators: Peter Fisla, Paullus

User avatar
UP844
Posts: 1669
Joined: Thu Mar 03, 2016 9:10 pm
Location: Genoa, Republic of Genoa (occupied by Italy)

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by UP844 »

Just for the sake of change, some additional whining about the AI [;)].

At the start of the Rout Segment, the AI had a leader and three squads in the Russian-controlled VP hex and three other squads in the adjacent hex NW of it. All the units were broken.

The first group routed in the hex NE of its starting hex (move D), as it was its only option.
The second group routed away (move A) in the only available hex and suffered interdiction, then routed again (move B) following the only available route path and suffered interdiction again, and finally (move C) moved in a wooden building.

So far, so good, but... the three squads that routed first moved in the hex initially occupied by the other group, exhausted their movement allowance and finished the Rout segment adjacent to a German stack. Score three kills for the Germans. I wonder why they moved there (and the leader that was stacked with them didn't).

Does the AI evaluate the status of possible rout hexes every time it is about to move? If this is the case, I can (barely) understand the reasons for its move: considering the German units that are known just before the second routing move (move E) in the hex where they were they could be fired by three squads behind the AFVs, a StuG, an Elite squad and a sIG 33. In the adjacent hex, the last two units cannot fire, so it could appear a desirable move, but... doesn't the AI "remember" there were enemy units adjacent to that hex?

Some other times, I have noticed the AI loses units remaining adjacent to enemy units because it does not rout when its only option would imply being subject to interdiction, which can be dangerous but in my opinion beats certain death [;)]

P.S. Of course, I have no saved game: I'll try to provide one, even though it's hard to tell when a similar situation will occur again.




Image
Attachments
AI_rout.jpg
AI_rout.jpg (1.25 MiB) Viewed 236 times
Chasing Germans in the moonlight is no mean sport

Siegfried Sassoon

Long Range Fire (A7.22)........1/2 FP
User avatar
Peter Fisla
Posts: 2574
Joined: Fri Oct 05, 2001 8:00 am
Location: Canada

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by Peter Fisla »

UP844, Rout Segment is very CPU demanding because there is a lot of LOS checking going on and also once a broken unit routs and discovers an enemy unit in LOS and then it routs away, it cannot go back to the hex were it was as known enemy units are tracked. So yes, Rout Segment is very abstracted...
User avatar
UP844
Posts: 1669
Joined: Thu Mar 03, 2016 9:10 pm
Location: Genoa, Republic of Genoa (occupied by Italy)

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by UP844 »

ORIGINAL: Peter Fisla

it cannot go back to the hex were it was as known enemy units are tracked.

If known enemy units at the start of the Rout Segment are tracked, cannot the hexes adjacent to them be defined as "off limits" for routing units? This would spare many unnecessary losses among AI routed units.
Chasing Germans in the moonlight is no mean sport

Siegfried Sassoon

Long Range Fire (A7.22)........1/2 FP
User avatar
Peter Fisla
Posts: 2574
Joined: Fri Oct 05, 2001 8:00 am
Location: Canada

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by Peter Fisla »

ORIGINAL: UP844

ORIGINAL: Peter Fisla

it cannot go back to the hex were it was as known enemy units are tracked.

If known enemy units at the start of the Rout Segment are tracked, cannot the hexes adjacent to them be defined as "off limits" for routing units? This would spare many unnecessary losses among AI routed units.


sorry, I don't follow...can you please explain
User avatar
UP844
Posts: 1669
Joined: Thu Mar 03, 2016 9:10 pm
Location: Genoa, Republic of Genoa (occupied by Italy)

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by UP844 »

I was in a hurry and I didn't explain what I meant in detail.

I did not understand the criteria the AI applies to decide if a specific hex is eligible as its next destination during the Rout Segment.

In figure 1 below, the routing units have no choice: they must rout away from their current hex since they are adjacent to enemy units, and two of the three hexes they could rout into are not eligible as destinations since there are adjacent enemy units. The only viable option, therefore, is to move North-East (move "D"). Here, the units could stop (that's what the leader did and, in my opinion, it the wisest move), but they choose to continue.

Figure 2 shows the next possible move for the routing units. This time, there are six possible hexes the units can rout into:
"no-1" is not eligible as it is the starting hex;
"no-2" is not eligible as it is adjacent to the visible DC-armed German unit (the stack with the broken and wounded leader is not visible at this time);
"no-3" is not eligible as it is adjacent to three visible German units: the DC-armed squad, the Elite squad and the sIG33;
"no-4" is not eligible as it is adjacent to three visible German units: the StuG, the Elite squad and the sIG33;
"yes-1" is eligible (but dangerous)
"yes-2" is eligible if the AI does not "remember" the presence of the German stack with the Elite unit on top which caused the route of the other broken units (see Figure 3). This appears to be the case, since the units rout again ("E" move) and are subsequently eliminated.

From your reply, I understood the AI "remembers" the situation at the start of the Rout Segment, i.e. what hexes have known adjacent units and are, as a consequence, not eligible as destinations for routing units (Figure 4). I was unable to understand why the routing units moved adjacent to an enemy stack that was known at the start of the Rout Segment.

I had the vague (very vague) idea that "marking" all the hexes adjacent to known enemy units as non-eligible as destinations for routing units might be a significant impact on CPU use: perhaps marking in such a way only the hexes from which broken units had to rout could have a lesser impact on CPU use?

Image
Attachments
Rout_options.jpg
Rout_options.jpg (1.17 MiB) Viewed 236 times
Chasing Germans in the moonlight is no mean sport

Siegfried Sassoon

Long Range Fire (A7.22)........1/2 FP
User avatar
rico21
Posts: 3034
Joined: Fri Mar 11, 2016 8:05 am

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by rico21 »

UP844 is busing with "Check you 6", so I continue modestly his piece.I downloaded 1.097 but,
With 1.095, I've seen:
1-In AI Phase Movement, tanks move then Infantry move then tanks move, I've been nicely surprised or I dreamed?
2- The stack of units not automatly with leader above the stack.
3- An AI squad with malfunctiuned weapon don't fire then it always should fire like an AI squad with no weapon.
4-Autosave is planned?

Thanks Peter for your great work.
User avatar
UP844
Posts: 1669
Joined: Thu Mar 03, 2016 9:10 pm
Location: Genoa, Republic of Genoa (occupied by Italy)

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by UP844 »

Another case of AI "avoidable losses".

A British leader and two squads (all of them broken) started their Rout Segment in the VP hex indicated by the red arrow.
They first moved south-east, in the only eligible hex (the sole hex not adjacent to unbroken enemy units known to them) and there they "saw" the Pz II. This should force the units to continue routing.
The leader and one of the squads, in fact, moved south-west (again, in the only eligible hex), took interdiction from the German 1st line squad 2 hexes away (no effect) and ended their rout in the wooden building.
The remaining squad, however, did not rout with the other units [&:]. It remained in the stone building and at the end of the rout segment was eliminated for being adjacent to an enemy unit.

Why didn't the second squad rout, considering that the alternative was certain elimination? There was room for it in the wooden building where the other units routed (it only contained a squad and a LMG), so it could have followed their route.

Image
Attachments
norout.jpg
norout.jpg (468.89 KiB) Viewed 236 times
Chasing Germans in the moonlight is no mean sport

Siegfried Sassoon

Long Range Fire (A7.22)........1/2 FP
User avatar
UP844
Posts: 1669
Joined: Thu Mar 03, 2016 9:10 pm
Location: Genoa, Republic of Genoa (occupied by Italy)

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by UP844 »

I played Rico's last work, "The Red House", with v.1.0.98.

The AI does not appear to be especially trigger-happy: on the contrary, it appears to be extremely reluctant to fire, especially in the Advancing Fire Segment.

The screen below was taken when I played with the Germans: the AI has three stacks that can fire on the German units:

Group A (cyan) consists of a B* leader, a 1st Line SMG squad, another 1st Line SMG squad (pinned) and an Engineer squad. Only the two Good Order squads fired, with an overall FP equal to (5/2 (FRD) + 1) + (6/2 + 1) = 7 FP. The stack fired on the easier target, i.e. German 1st Line squad in the wooden rubble, with a net +1 modifier (-1 leadership, +2 TEM). The DR required to have some effect is 7 or less (58.33%).
The chances to affect at the broken 2nd line unit are identical, but I would also have fired on the Good Order Unit.

Group B (yellow) consists of a C* leader, a 1st Line SMG squad, another 1st Line SMG squad (pinned) and an Engineer HS armed with a DC. It didn't fire, even though it had two possible targets: a C leader with a squad and a MMG in the fortified building (DR required 5 or less = 27.77%) or the OOC 2nd line squad in the stone rubble (DR required 6 or less = 41.66%).

Group C (red) consists of a 1st Line SMG HS armed with a DC, an Engineer squad armed with a LMG and an Engineer HS armed with a MMG (that cannot fire after having moved). It didn't fire, even though it had two possible targets: the OOC 2nd line squad (DR required 5 or less = 27.77%) or the OOC 2nd line squad in the stone rubble (DR required 5 or less = 27.77%).

I would have fired with both Group B and C: these units have nothing better to do, anyway. End even though their chance to actually hurt the Germans are pretty slim (let's say 50% of the probability to obtain an effective result), why waste an opportunity? The AI, however decided not to fire.

For the record, the game ended with a record 25:1 loss rate for the Germans (1 HS lost vs. 12.5 squads).

Image
Attachments
trigger_happy-2.jpg
trigger_happy-2.jpg (772.39 KiB) Viewed 236 times
Chasing Germans in the moonlight is no mean sport

Siegfried Sassoon

Long Range Fire (A7.22)........1/2 FP
rmmwilg
Posts: 228
Joined: Tue May 01, 2018 11:01 pm

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by rmmwilg »

Hi guys,

Updated to v5 (1.1.09) but no joy on playing with it playing in 4k resolution (3840x2160); however I see prev. comments about lowering the zoom level, This Lenovo zoom defaults to 300% at that setting. Is that the inhibiting factor?
I did try changing the config file for a windowed setup with 4k resolution, but still no joy. At the mo, the highest it runs on is 2560x1600, in which this Lenovo zoom defaults to 200%
Paullus
Posts: 1097
Joined: Tue Jun 09, 2015 1:41 am
Location: Sweden

RE: Testing UPDATE5 - 1.0.88 - ALPHA

Post by Paullus »

Try the new alpha build Peter has put in this thread. Zoom out to 100%. This should work. After that try zooming in incrementally until it stops working.
For my part, I shall do my duty as a general; I shall see to it that you are given the chance of a successful action. /Lucius Aemilius Paullus
Post Reply

Return to “Tech Support”