Testing UPDATE5 - 1.0.88 - ALPHA

[UP844]

Hi Peter, I have one, but I do not remember at what point I saved.

Should I save the game at a given segment? (e.g. at the start of the Movement Segment)

[Peter Fisla]

Yeah, it’s important to have a save game before you see some challlenges with lone AI leaders. Yes, saving before AI movement segment is what i’m looking for. The when you go through the AI movement segment and notice something isn’t right then you have a saved game that I can look at.

Thanks

[UP844]

Aye-aye, Sir!

[UP844]

The attached game has been saved at the end of the German Admin segment of the American player turn #10.

The leader in 15,25 moves to 16,25 and then returns to 15,25.
The leader in 18,23 moves to 19,24, then to 18,24.

The latter could have moved:
from 18,23 to 19,23, then to 20,22 (where there are some broken units), or
from 18,23 to 19,24, then to 19,25 (and then it could have advanced in 20,26)

Let me know if you need more saved games.

[Peter Fisla]

Ok great, I'm looking into the issue...

[Peter Fisla]

Hi UP844, ok I believe I have improved the AI lone leader movement, take look at it and let me know with further testing.

thanks!

[dox44]

also thanks for the update!

[UP844]

In the attached ZIP file there are two saved games, one saved at the very start of the Movement segment, the other just before the Advance Segment.

Lone leader movement in the movement segment seems to work fine: here we have two lone leaders that head for a stack of broken units in 20,5:
leader "X" moves first from 22,7 to 21,5 and then to 21,4.
leader "Y" moves second, from 18,7 to 19,7, to 20,6 and finally to 20,5.

In the advance segment, howewer, leader "Y" moves alone out of the hex containing the broken units to 21,4, while leader "X" moves into 20,5.

In such a situation, I thought the leader who already was with the broken units should stay there, while the other should advance towards the next eligible friendly stack.

[Peter Fisla]

ORIGINAL: UP844

In the attached ZIP file there are two saved games, one saved at the very start of the Movement segment, the other just before the Advance Segment.

Lone leader movement in the movement segment seems to work fine: here we have two lone leaders that head for a stack of broken units in 20,5:
leader "X" moves first from 22,7 to 21,5 and then to 21,4.
leader "Y" moves second, from 18,7 to 19,7, to 20,6 and finally to 20,5.

In the advance segment, howewer, leader "Y" moves alone out of the hex containing the broken units to 21,4, while leader "X" moves into 20,5.

In such a situation, I thought the leader who already was with the broken units should stay there, while the other should advance towards the next eligible friendly stack.

Ok, I have just uploaded 1.0.91, I believe I have fixed the lone AI leader issue in the Advance Segment. I'm still looking at improving Movement Segment AI lone leader behaviour. I will let me know when I'm done.

[UP844]

Downloaded and installed 1.0.91.

On the first turn, a good order leader routed with a broken squad and DID NOT advance in the subsequent Advance Segment [:)].

One swallow does not make a summer... but so far the issue appears to be fixed.

More on this later...

[Peter Fisla]

It’s important with the build 1.0.91 not to play old saved games, please start only new scenarios.

[UP844]

The leaders behave quite well: they managed to rally a significant number of units. I have not taken notes about the number of broken units remaining in the rear during the game, but I have the distinct impression their number has remarkably decreased in version 1.0.91 [&o]. In fact, apart from a few units that were broken by a OBA in the very first turn, all the other broken units were ahead of the leaders, not behind them.

During the whole game I only noticed three instances when the leaders appeared to behave "strangely". I have no saved games because these are the only three times when the leader "misbehaved"; I hope the screenshots I took are clear enough.

In case A, the leader left a broken squad in the movement segment and headed for another hex containing a MMG-armed broken squad (red arrows). Perhaps the AI "thinks" an MMG-armed squad is more important of a squad without a SW? (or is this wishful thinking?)
In case B, the leader left a broken squad in the movement segment and headed for another hex containing two broken units (cyan arrows). Does the AI "think" it is better to rally two squads than one?
In case B, the unit indicated with a red arrow was broken by German fire and routed up to the hex at the bottom from which the leader departed. In the advance segment, the leader advanced towards its starting hex [:D]. Am I correct when I suppose it advanced towards the closest of two equal (i.e. containing 2 broken units) hexes?

Case C is the instance that puzzled me more: the leader started the advance segment with a pinned unit and advanced alone towards the enemy positions. The leader wasn't "alone in the hex" at the start of the Advance segment: is this the reason it didn't advance towards the broken units?

[Peter Fisla]

always save before the Fire Segment (in case you have no control during a Fire Segment), so that you can then follow what happens in Movement segment

[UP844]

I didn't save because the three cases above occurred in the last turns of the game and so far the leaders behaved as expected.

If the hypotheses I made about the reasons for the behaviour of those three leaders are correct, however, I think we can consider the "lone leader charging" issue solved. I will make other tests (with other scenarios) to see if any other inconsistencies occur.

[UP844]

I made a quick game of "Buchholz Station" (my first attempt at making scenarios) and the leaders behaved properly [:)].

[Peter Fisla]

ORIGINAL: UP844

I made a quick game of "Buchholz Station" (my first attempt at making scenarios) and the leaders behaved properly [:)].

Nice to hear, thank you

[Peter Fisla]

UP844, do you feel like sometimes the AI personnel units are not advancing fast enough ? If you have an example (saved game), I can take a look...

[UP844]

Quite the opposite: I feel sometimes the AI moves too quickly.

By way of example, I noticed sometimes the AI uses Double Time and then stops after moving 1 or 2 hexes [&:], or a Normal Move to move 1 hex only. Is this due to the AI re-assessing the situation and the risk/benefit ratio before moving additional hexes?

The situation I wrongly reported about was due to the presence of two exit hexes in the scenario I was tinkering with: in such cases, the AI always starts a mad rush to the exit hexes (but I think such behaviour can be corrected - at least for a defending AI - by making use of the "DEFEND HEX" option).

The fire issue is what is most bothering me: besides the long range issue (more on that later), I think the AI often does not fire for no apparent reason. I noticed that, in the Defensive Fire segment (where the AI does not have to choice if it should fire now or move later), not all the AI units that are eligible to fire actually fire: I cannot see why they shouldn't. In the last scenario I played saw a Canadian (British) squad stand still at a range of 3 hexes from 2 SS squads in a stone building without ever firing, not even in the Fire or Defensive Fire segment. I acknowledge its chances to inflict some damage were pretty slim (DR of 2=1MC, 3=NMC, 4=PTC), but since it didn't perform no other action at all, why shouldn't it fire?

As regards to long range/reduced effect fire: giving medium/heavy MGs the ability to fire at long range has made the AI a more dangerous opponent, but this only stands true for scenarios with very long visibility: in most scenarios, visibility is set to 12-16 hexes, i.e. within the normal range of MGs.
I think the possibility of allowing long range fire for all units should be slowly and cautiously explored.

By the way, I would like to hear some other opinion about AI issues.

[Peter Fisla]

When an AI personnel unit has move command and moves, it will check for danger situations. If you have saved games where the AI should fire obviously and it did not then please let me know, I can take a look. Generally, the AI evaluates fire conditions.

[UP844]

Thanks Peter, your statement explains why the AI sometimes stops in its tracks!

I will try to provide a saved game showing AI units not firing when they should have do so.

Would it be feasible to add an autosave feature, just for such debugging purposes? It could also come handy for those rare instances when the game freezes.