AI surrenders too easily

[c2yeung]

I have played a long MG scenarios, 1 hour+ long but don't remember which one. Allied Versus AI. Allied took control of 3.5 of 5 objectives. I looked at the map after AI surrendered (there is still a lot of time left) and saw a few HQ units, some remnants of a squad scattered plus a full company guarding the last objective which is still a way to get to. Considering the strength of my force, I am confident I can get the rest of the objective within the time remaining. It is not going to be a walk in the park but I just feel that AI has given up the fight too early. Does anyone know how AI determines when to throw in the towel?

[Holman]

My rough sense is that the AI generally surrenders when nearly all of its remaining units are broken, shaken, or panicked. (I don't know if there's an algorithm that weights an HQ's panic as worse than a rifle team's, etc.)

I don't think the AI ever surrenders on the basis of tactical situation (or score) alone. I believe it's purely morale based.

I wouldn't say it surrenders too easily in general. It certainly fights past the point where many real-life commanders would withdraw. Obviously, though, some scenarios are easier than others.

[Collingwood]

And on the other hand I was attacking a small town and had generally wiped out all traces of resistance. I kept waiting for a surrender, then stumbled upon a HQ unit hiding in a courtyard toward the rear. They bravely tried to fight against overwhelming odds but I wiped them out. Still no surrender. I then destroyed one section left in a building, and got a surrender.

I kept expecting a surrender much earlier and was thinking WTH, these guys must be super hard core elite fanatics. All their buddies are dead, they're surrounded... no they declare, we've still got a pistol and one rifle left!

I think Martyr's right - our pixeltruppen aren't thinking "har, we still control the church, we will get a draw" - instead it's "hmm, there's only a couple of us still alive, perhaps it's over for us".

[db_zero]

I haven't play much against the AI, but h2h things tend to get pretty bloody-perhaps too bloody. Maybe the AI has it right for the European theater in the Western Front.

The Eastern Front and Pacific Theater is a different story.

[Doug Williams]

That's odd, OP, I usually have the opposite experience, with the AI, IMO, waiting too long to surrender. I've played some "small" QBs against the AI where it didn't surrender until it had less than half a squad left, and nothing in the victory location.

[Pak40]

c2yeung,

In all honestly the AI doesn't surrender enough. If you read 1st hand accounts of battles in Normandy and especially Market Garden, there were lots of soldiers surrendering on both sides. This leads to bloated unrealistic casualty figures, especially KIAs.

In CMBN, you might occasionally get a soldier or two that surrenders in a battle. Also in some rare cases where the global moral was low to begin with, the AI might surrender the entire battle (like in your case). The enemy may have had a company's worth of units left but it's quite possible that they were low in morale to begin with and your previous success has pushed them over the edge.

You could possible open the battle up in the editor to figure this out.

[MikeyD]

A driving factor might be ammo. A long hard-fought battle implies depleted ammo stocks, broken command links, casualties, suppression. Conscript and green units just can't survive such abuse. Veteran and elite forces, though, are liable to keep fighting with a broken table leg if they've run out of ammo. Those guys usually prefer to die on their feet.

[YankeeDog]

Yeah... I don't really know but I think global surrender has a lot to do with the "Condition" of AI units still on the map. I don't think the global surrender trigger has much (if anything) to do with who controls any objectives.

So if the AI units are high experience and/or have morale bonuses, you have to take out nearly all of them to trigger a surrender. OTOH, with a low experience and/or negative morale modifier AI force, it's somewhat easier to force a global surrender.

Regardless, IME the AI is pretty much pantsed by the time the global surrender happens.

[c2yeung]

c2yeung,

In all honestly the AI doesn't surrender enough. If you read 1st hand accounts of battles in Normandy and especially Market Garden, there were lots of soldiers surrendering on both sides. This leads to bloated unrealistic casualty figures, especially KIAs.

In CMBN, you might occasionally get a soldier or two that surrenders in a battle. Also in some rare cases where the global moral was low to begin with, the AI might surrender the entire battle (like in your case). The enemy may have had a company's worth of units left but it's quite possible that they were low in morale to begin with and your previous success has pushed them over the edge.

You could possible open the battle up in the editor to figure this out.

If AI based its surrendering decision on the morale, leadership, quality, strength, ammo and the like of its force, what is it comparing them to? AI doesn't know with any certainty the conditions of my force. Maybe mine is just as bad. Having said that, I have read some historical accounts in WWII that a few infantry men bluffed their way and succeeded in capturing enemy force many times their size.

Perhaps, scenario designers have the answers. Aside from capturing objectives, enemy casualties and friendly conditions etc..What perimeters, if any, are used to determined victory condition or surrender the whole battle.

[Childress]

In all honestly the AI doesn't surrender enough. If you read 1st hand accounts of battles in Normandy and especially Market Garden, there were lots of soldiers surrendering on both sides. This leads to bloated unrealistic casualty figures, especially KIAs.

Great minds think alike. CM games (particularly CMSF) would enhance their sim cred with more surrender events- like unhorsed tank crews, for example. An additional causative factor may be the excessive resilience of the troops which some posters allege. But coding a plausible surrender algorithm gotta be hard , it's so... situational.

[MikeyD]

This might be a 'be careful what you wish for' issue. I recall in the past there being lots of complaints about the AI suddenly giving up the fight just as the player is set to spring his clever trap, to the player's great frustration. This happened a lot in CMSF where oftentimes you were facing green and conscript troops. It happens less in CMBN because we play less with green and conscript troops. Some players boast of only picking crack/elite forces for QB games. Heck, you'll never see an auto-surrender if you're playing with elites!

[Skwabie]

Interesting...

So far almost all of my wins are AI surrenders. And 90% of those is when AI lost 50% men.

There were exceptions like a decision scenario where you choose the course of the campaign by occupying A or B zones. Other than that the 50% rule seems to apply.

[womble]

Interesting...

So far almost all of my wins are AI surrenders. And 90% of those is when AI lost 50% men.

There were exceptions like a decision scenario where you choose the course of the campaign by occupying A or B zones. Other than that the 50% rule seems to apply.

I wish the AI would surrender at 50%, sometimes. Most of the time I seem to have to reduce them to shreds of broken remnants of shadows of their former units...

[Michael Emrys]

I wish the AI would surrender at 50%, sometimes. Most of the time I seem to have to reduce them to shreds of broken remnants of shadows of their former units...

Maybe start them out with really low morale?

Michael

[womble]

Maybe start them out with really low morale?

Michael

That would generally mean either persuading my HvH opponent to pick low morale troops, or editing the campaign or scenario, losing any FoW regarding the opposition forces (and being a lot of work in the case of a campaign).

[Holman]

If AI based its surrendering decision on the morale, leadership, quality, strength, ammo and the like of its force, what is it comparing them to? AI doesn't know with any certainty the conditions of my force. Maybe mine is just as bad. Having said that, I have read some historical accounts in WWII that a few infantry men bluffed their way and succeeded in capturing enemy force many times their size.

I don't think it needs to compare them to anything. Broken is broken. Panicked is panicked. By definition, units that are so totally spent won't be making a calculation about whether the enemy is equally bad off. Such units will probably tend to overestimate enemy prowess instead.

As for bluffing, it's perhaps possible for a few plucky hold-outs to overcome a force much larger than themselves, but those bold units will (by definition) not be broken or panicked.

Perhaps, scenario designers have the answers. Aside from capturing objectives, enemy casualties and friendly conditions etc..What perimeters, if any, are used to determined victory condition or surrender the whole battle.

I don't believe scenario designers have control over total surrender, just AI plans and victory conditions. If the AI somehow holds all the objectives but is now completely broken, it will still surrender. Just consider them to have made the calculation that the territory is no longer worth the cost of their lives.

As with all things CM: you'll encounter some outlier results here and there, but sustained play will reveal that the game gets things pretty satisfyingly close to what we would expect from real life.

[womble]

I have a distinct recollection that I've had the enemy surrender on a turn where I didn't fire at them, even, but did put the last VL into "contested" state. As a result, I believe that VP do have an impact, but the general state of the AI side does to, and to a much larger degree.

Of course, I might have missed some shooting, or some other effect might have caught up with the opposition (if a distant platoon only just dropped a morale state because news of their comrades' destruction only just reached them).

[George MC]

I don't believe scenario designers have control over total surrender, just AI plans and victory conditions. If the AI somehow holds all the objectives but is now completely broken, it will still surrender. Just consider them to have made the calculation that the territory is no longer worth the cost of their lives.

As with all things CM: you'll encounter some outlier results here and there, but sustained play will reveal that the game gets things pretty satisfyingly close to what we would expect from real life.

Scenario designers can arrange for reinforcements to come on after the scenario has finished i.e. set the reinforcement group to come in after 2 hours when the scenario finishes at 1 hour. This will tend to make the AI fight on as it "think's" it has stuff coming on to help it.

I used this a lot in CMSF to slow down the AI from surrendering. Although TBH I don't use this in any WWII scenarios I've designed and not sure if anyone else still does?

[Apocal]

As with all things CM: you'll encounter some outlier results here and there, but sustained play will reveal that the game gets things pretty satisfyingly close to what we would expect from real life.

Uh, battalions being reduced to around reinforced company strength before abandoning their mission is not what you'd generally expect in real life.