Panzer Corps AI discussion

[robman]

We aren't playing with the Imperial Japanese army, after all.

But hopefully we will be soon...? (hint, hint)

[comradep]

My biggest gripe about the AI is the retreat choices it makes for both my units and the enemy.

In almost all cases that I've seen, the unit retreated to the hex on the other side of the unit compared to the hex I attacked from. I don't recall seeing the AI retreat in the direction of my units if there was a valid retreat possibility "behind" the unit as seen from the direction my own unit attacked from. To put it in numerical terms using the layout of the keypad: if I attack from 4, the unit (at 5) usually retreats to 6.

[Rudankort]

In almost all cases that I've seen, the unit retreated to the hex on the other side of the unit compared to the hex I attacked from. I don't recall seeing the AI retreat in the direction of my units if there was a valid retreat possibility "behind" the unit as seen from the direction my own unit attacked from. To put it in numerical terms using the layout of the keypad: if I attack from 4, the unit (at 5) usually retreats to 6.

Yes, that's correct. If a unit can retreat to the hex opposite to the attacking enemy, it uses that hex. Then, it tries two hexes next to it. And if those are unavailable as well, it will try to retreat to one of the hexes next to attacking enemy.

[OldGiants]

In almost all cases that I've seen, the unit retreated to the hex on the other side of the unit compared to the hex I attacked from. I don't recall seeing the AI retreat in the direction of my units if there was a valid retreat possibility "behind" the unit as seen from the direction my own unit attacked from. To put it in numerical terms using the layout of the keypad: if I attack from 4, the unit (at 5) usually retreats to 6.

Yes, that's correct. If a unit can retreat to the hex opposite to the attacking enemy, it uses that hex. Then, it tries two hexes next to it. And if those are unavailable as well, it will try to retreat to one of the hexes next to attacking enemy.

Would that that were true.

My description of the AI retreats continues to happen. Units with no one 'behind the lines' will plunge straight forward deeper into enemy units rather than go in the opposite direction to safety.

I'll try to take a screen shot next time it happens.

[El_Condoro]

What sort of terrain is in that 'open' hex? If the terrain is impassable to the unit, for whatever reason, it can't retreat into it. The screen shot will help, too.

[OldGiants]

Here's a simple example of what I mean. The Brit infantry was attacked in the town. Instead of retreating to the southwest, an open terrain terrain road not next to any German units, it went North next to the captured Somua, the only unit left that could attack it, and was destroyed.

This happens a lot, not always, but a lot.

http://www.flickr.com/photos/75917869@N08/7314779932/

[Tarrak]

Here's a simple example of what I mean. The Brit infantry was attacked in the town. Instead of retreating to the southwest, an open terrain terrain road not next to any German units, it went North next to the captured Somua, the only unit left that could attack it, and was destroyed.

This happens a lot, not always, but a lot.

http://www.flickr.com/photos/75917869@N08/7314779932/

In this case the unit is behaving exactly as Rudankort wrote a few posts above. It is first trying to retreat opposite of the attacking unit but that field is blocked by the artillery so it tries the two hexes next to it. It either picks one at random or simply scans them in some order (probably defined by the coordinates). Of course one could implement a check to prefer a hex without any enemy unit neighboring but it wouldn't be always the better move as well. In the end there are so many variables you need to consider it becomes impracticable to check them all.

[OldGiants]

Survival is always the better move.

The hex to the SW would allow for reinforcement and survival.

How hard is it to code a preference for no contiguous enemy units? Even I could do that in a few minutes.

[deducter]

Survival is always the better move.

The hex to the SW would allow for reinforcement and survival.

How hard is it to code a preference for no contiguous enemy units? Even I could do that in a few minutes.

Not necessarily. You are not considering all possible factors. There are ways to exploit the situation you outlined. The main one is that you can select the retreat path even more carefully now, and use a unit with movement points left to finish off the weakened unit anyway.

I can think of situations that arise where you can force a retreat into more favorable terrain, so that a unit you have with movement can go up and finish off the weakened unit. For an advanced player, this means making it even easier to force the AI to retreat onto rivers, close terrain, etc. where it becomes even easier to pick off the weakened unit.

[OldGiants]

Survival is always the better move.

The hex to the SW would allow for reinforcement and survival.

How hard is it to code a preference for no contiguous enemy units? Even I could do that in a few minutes.

Not necessarily. You are not considering all possible factors. There are ways to exploit the situation you outlined. The main one is that you can select the retreat path even more carefully now, and use a unit with movement points left to finish off the weakened unit anyway.

I can think of situations that arise where you can force a retreat into more favorable terrain, so that a unit you have with movement can go up and finish off the weakened unit. For an advanced player, this means making it even easier to force the AI to retreat onto rivers, close terrain, etc. where it becomes even easier to pick off the weakened unit.

I'm considering more factors than the AI currently considers. That would be an improvement in the spirit of this thread.

Your 'consider all' is much further from the current situation, and not a standard that need be met to make a large improvement.

[deducter]

I'm sorry if I sounded rude, but my point is that your proposed change wouldn't necessarily make the AI smarter. I would argue your proposed changes is not better than the current system, because I can still take advantage of it in other ways.

[OldGiants]

I'm sorry if I sounded rude, but my point is that your proposed change wouldn't necessarily make the AI smarter. I would argue your proposed changes is not better than the current system, because I can still take advantage of it in other ways.

I don't view your valid point of 'take advantage' as an exploit to be avoided because it requires planning and application of reserves to make it happen.

If the unit had gone to the SW, it would have required me to have a unit in reserve which could move into attack position on that turn. If the player has such a unit, it is good planning, not an exploit.

By moving adjacent to my unit which could no longer move but had an attack available, the AI made it simple for me to destroy that unit on that turn without any planning on my part (I had moved that unit up to get the multi-hex side attack benefit--my choice instead of keeping a reserve).

I don't want to the AI to make me gifts of extra attacks, or to plunge weakened units into areas where they can be more easily destroyed when an alternative route might save them--or at least cause me to allocate movement and/or attack resources to complete the job.

And I don't understand how you could 'take advantage' of my suggestion. An example would help me comprehend that.