The AI on the Attack (and food for thought for players too)

[blackcloud6]

I completed my first ASL scenario conversion to TOTH this morning, and the testing of my conversion got me thinking about how the AI attacks.

I suspect most of play TOTH as the attacker and not as much as the defender. From what little I've played as the defender I see feel that the attacking AI makes many mistake new ASL players make. I understand how difficult it is to get the AI to play well and cleverly on the offense in a computer game, but if the game designer can take into account these suggestions when trying to make the AI better, the game will become better and more challenging.

1. Shooting instead of moving. I notice the AI tends to shoot at the enemy, even at suboptimal ranges and conditions (through hindrances and into high TEM hexes). There is an old adage in ASL that: ‘if you are shooting instead of moving, you are probably losing.” In tactical combat it takes aggressiveness to win, and in these games that requires you to close with the enemy. Attackers need to keep moving towards and pressuring the enemy, very little fire should be done in the fire phase and that should be done by your key weapons (machines guns, ordnance), the squads need to be closing. I can tell that the attacking AI is going to probably lose the scenario when most, if not all, of his force is bogged down in a firefight. And many times I can figure out how to make it do so. So if there is a way to make the AI move more units towards the enemy, the attacking AI will get better.

2. Sarge says “don’t bunch up, one grenade can get you all!” New ASLers will tend to move in stacks. Those who move in stacks, die in stacks. Moving in stacks with a leader to gain is movement bonus should be done out of fire form the enemy or only, when under fire, where absolutely necessary. Believe it or not, you have more time than you might think to get the job don’t; don’t present to big juicy stacks for the HMG to eat up. The attacking AI seems to move, under fire, in stacks a lot. This needs to be looked at.

3. Move using cover. Moving in the open will get you killed. Moving in the open in stacks will get a lot of folks killed. Always use the terrain that will provide some protection to move in when heading towards the bad guy. The AI seems to want to take the fastest route many times and gets chewed up quickly doing so.

4. Assault move. This is a great tool to get next to the enemy. I’m not sure if the AI uses it enough.

5. Bait. Send those conscripts first and draw off first fire. But don’t send them on the route of your attack as you will then face residual firepower in crucial hexes. I don’t think the AI does this at all. It just starts moving units towards the objectives and gets good units shot up first. BTW, the flip side of this coin is important when the AI is the defender. The AI seems to take the first shot when it can at moving units. It does not wait until the enemy is in the open or at closer range. Thus is it is easy to bait the AI into first firing or final firing before you send in the real assault units.

6. Time is on your side…mostly. I find that the attacking players who win the most are the ones who spend their time moving units into proper position for supporting fire and to attack, and then at one crucial moment, the entire attack is launched and the defense stressed to the breaking point. Synchronizing the attack is crucial. Can we ever get the AI to do this? It is probably tough to do but it sure would be nice to see

[Paullus]

Nice list. We will work on the AI to make it even better on the Assault. We have some ideas but every contribution will help [:)]

[blackcloud6]

Nice list. We will work on the AI to make it better on the Assault. We have some ideas but every contribution will help

Cool, thanks.

[virtuali]

2. Sarge says “don’t bunch up, one grenade can get you all!” New ASLers will tend to move in stacks. Those who move in stacks, die in stacks. Moving in stacks with a leader to gain is movement bonus should be done out of fire form the enemy or only, when under fire, where absolutely necessary. Believe it or not, you have more time than you might think to get the job don’t; don’t present to big juicy stacks for the HMG to eat up. The attacking AI seems to move, under fire, in stacks a lot. This needs to be looked at.

The issue is, in order to improve this, we would need the crucial multi-hex firegroup capability. With the game as it is now, stacking might be the only way to gather enough punch to dislodge enemy in high cover terrain. So, before we can teach AI better tactics, we need the rules to allow them...

[blackcloud6]

The issue is, in order to improve this, we would need the crucial multi-hex firegroup capability. With the game as it is now, stacking might be the only way to gather enough punch to dislodge enemy in high cover terrain. So, before we can teach AI better tactics, we need the rules to allow them...

That is a very good point.

[iPhoneAppz]

ORIGINAL: virtuali

2. Sarge says “don’t bunch up, one grenade can get you all!” New ASLers will tend to move in stacks. Those who move in stacks, die in stacks. Moving in stacks with a leader to gain is movement bonus should be done out of fire form the enemy or only, when under fire, where absolutely necessary. Believe it or not, you have more time than you might think to get the job don’t; don’t present to big juicy stacks for the HMG to eat up. The attacking AI seems to move, under fire, in stacks a lot. This needs to be looked at.

The issue is, in order to improve this, we would need the crucial multi-hex firegroup capability. With the game as it is now, stacking might be the only way to gather enough punch to dislodge enemy in high cover terrain. So, before we can teach AI better tactics, we need the rules to allow them...

Multi hex firegroups are probably the most important thing that the game needs.

[DoubleDeuce]

Another thing, and maybe I am the only one seeing it as it hasn't been posted elsewhere,(at least that I can find), is vehicles racing way ahead of the infantry to attack objectives. And I would second the ability for fire groups, stacking by the AI seems to be one of its biggest vulnerabilities when its on the offense.

[pzgndr]

ORIGINAL: iPhoneAppz
Multi hex firegroups are probably the most important thing that the game needs.

This and MG penetration factor are important.

[idjester]

ORIGINAL: iPhoneAppz

ORIGINAL: virtuali

2. Sarge says “don’t bunch up, one grenade can get you all!” New ASLers will tend to move in stacks. Those who move in stacks, die in stacks. Moving in stacks with a leader to gain is movement bonus should be done out of fire form the enemy or only, when under fire, where absolutely necessary. Believe it or not, you have more time than you might think to get the job don’t; don’t present to big juicy stacks for the HMG to eat up. The attacking AI seems to move, under fire, in stacks a lot. This needs to be looked at.

The issue is, in order to improve this, we would need the crucial multi-hex firegroup capability. With the game as it is now, stacking might be the only way to gather enough punch to dislodge enemy in high cover terrain. So, before we can teach AI better tactics, we need the rules to allow them...

Multi hex firegroups are probably the most important thing that the game needs.

Sorry I do disagree, it is slightly "harder to dislodge enemy units" but I really don't have much trouble doing this. The loss of multigroup firegroups isn't critical. It just makes you stack your units in kill stacks more. Instead of units in 2 or 3 hexes you just place those units in one hex. And it makes the game quicker to play since I am not moving each unit by themselves, instead I am moving stacks with leaders and a few single units. Quick play and the LOS and rules for multihex firegroups would be a nightmare to add. Is that worth the time and effort for such little bonus, that would be the main question.

[Peter Fisla]

Multi-hex fire groups is difficult to implement from interface and AI perspective, this feature is not in my future scope at the moment.

[Gerry4321]

Sorry but stacking and not spreading out is totally ahistorical. Imagine crossing a wide stretch of open ground while all bunched up.

[BeamMeUpScotty]

I have to agree with everyone else...

[Rakkasan101]

Being able to focus fire on targets of your choice is also historical[;)].

[Gerry4321]

Not at all. Concentration of force on points was normal. Now we can only do that by stacking.

[idjester]

ORIGINAL: Gerry

Sorry but stacking and not spreading out is totally ahistorical. Imagine crossing a wide stretch of open ground while all bunched up.

That's not what we are talking about. I can move up one unit, than another, then another... I can move them all one at a time.

It's also ahistorical to think that a leader in one hex is going to be able to control the units in several other hexes..

Its a game mechanic that simulates something that can't be simulated correctly.

[Gerry4321]

I never said a leader will affect the fire from a multi-hex firegroup.

How are you going to get a lot of concentrated firepower if you don't stack? And stacking is against basic military principles given the obvious deadly effect of MTRs, etc. on such units.

[Monkie]

I've seen the AI use the advance segment to move from cover into open ground for no tactical reason whatsoever. Maybe that can be tweaked a bit.

As a side question in regards to fire groups, it's been awhile so I have forgotten but does SL or ASL have rules for multi-hex fire groups?

[UP844]

Some of my experience and some rants [:)] about the AI. This is not a criticism of Peter's work: I think he has made a great work with this game; I'm having great fun with it and yes, I am one of the ones who were waiting for it in the last 35 years. I just want to highlight some of the weak spots of the AI opponent.

1) The AI looks a bit shy; I think it should be somewhat more aggressive.

I just finished two games of "Rehearsal for Crete" (one as the British and one as the Germans).
The German AI left its Sd Kfz 251 sitting in the open in its entry hex under fire by a 2 pdr AT gun, for several turns, until it was finally destroyed; the 222 armored car stood still in the hex where it entered, while nearby infantry was slaughtered by long-range British HMG fire [&:] Some Fallschirmjager squads never moved during the whole game, even though there were plenty of British units visible elsewhere on the map. In the meantime, my M3 GMC ran wild across the map destroying German infantry with ease.
When I played as the German, I never saw the British AFVs, and the same is also true for the British elite squads entering from the south.
Maybe this timid approach is due to Victory Conditions of this scenario (no VP for objective hexes, only for units destroyed) even though I think this should make the AI more eager to hunt enemy units.

The same also happened many other times: I played "Hill 621", several times as the German, but I never saw the SU-122 and the SU-152 (I didn't miss them, especially the SU-152s [:D]); playing as the Russians, I only saw the Pz IV F2s until I crossed the ridge: all the remaining German AFVs were sitting there (the Duke of Wellington would have loved this reverse slope defense ).

2) Leaders don't care enough about rallying broken squads.

All too often I see a leader rally himself and charge straight towards the enemy, leaving scores of broken units behind: this afternoon, I played "Canicatti High Ground" and two US leaders advanced alone towards the German-occupied ridge while a dozen broken squads were left on their own .

3) Target selection and threat assessment

I am sure there is a logic behind target selection, but sometimes the AI ignores the most threatening units and shoots on some distant and harmless target (Example: a Pz IV F2 fires its MA to a single Russian infantry squad while there is a T34/76 a few hex away)

The AI also does not seem interested in minimising damage: I repeatedly shoot some tank in the rear and the IA only rotated the turret in reaction. This is great if you are chasing Panthers with Stuarts (Defending Meximieux scenario), but I think no tank crew take more than one shot in the rear without turning towards the attacker.

My two €cents .

[decaf]

*** This is a repost from March 12 in the "Requests" thread. ****

I'm reposting it here since is is on-topic and you may have missed it.

= = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =

Dear Peter,

I'd like to put a request in for Multi-Hex Firegroups. I'd like lots of other enhancements, too. But, my number one request
is "Multi-Hex Firegroups".

Please allow me to expand on "how" and "why".

How?
My handle, javafiend, is because I like programming in Java (not because of caffine). I'm familiar with the software machinery
needed to support a UI, such as events and tasks. So, I'm sensitive to the amount of work a UI change can entail.

Right now you have a MouseListener that recognizes LMB and ctrl-LMB for your Unit control Panel (lower left).
You also have a MouseListener that recognizes LMB for your Tactical Map.
I propose you extend the Tactical Map MouseListener to include ctrl-LMB, and that this mouse gesture will signal the selection
of multiple Firegroups.

The implementation flows from this extention. Simple LMB operates as before -- clear Unit Control Panel and refresh with the
new unit list. Now, ctrl-LMB will append more units to the Unit Control Panel. Only allow an append if the selected hex
is Personnel, not a solitary Leader, and adjacent to one of the other hexes in the Unit Control Panel list. User may repeat
ctrl-LMB on multiple hexes as desired. Not adjacent, no append, maintain current Unit Control Panel list.

Once the multiple location list is finally selected, then the user can select and deselect one or more units as is currently
done in the Unit Control Panel. The only difference is that the Unit Control Panel contents can be longer than before, and
encompass multiple locations. Of course, it will be necessary for the Unit Control Panel to shift to a scrolling mode as it
fills up. I note that you already have this function implemented in your Action Log Window.

Combat algorithms should be pretty much unchanged. It will be necessary to account for variable range in the subroutine.

The advantage of this approach is that it retains your current UI look-and-feel. It merely provides an extra, optional,
mouse gesture, ctrl-LMB, in the Tactical Map. This does not provide any burden to player's current play style. It can
easily be ignored as a feature if a player desires -- just don't press the ctrl when mousing the Tactical map.

Why?
The rest of this section, Peter, you are probably familiar with. Please skip if you like.

I go back to the old ASLML days. "Those who know" would *always* advise ASL newbies, "Don't stack!" For one example:

>>McGrath's Tactics 101, Mon, 27 Dec 1999,...
2. Don't Stack. "Always spread out and form firegroups. It is
better to have one squad in a stone building and another in the
woods than to have both... in the same location in a stone
building... By not stacking, the most your opponent can hurt
with a single attack is one squad. My rule of thumb is one squad
with a '-1' leader and two squads with a '-2' or better leader."

For you TOTH-ers who've never done ASL; you've seen terrible, horrible things happen to your stacks from just one shot.
The computer rolled a "2" on it's fire attack, and the badness hits every unit in the stack (followed by much wailing
and gnashing of teeth).

Now we see the TOTH combat experience depart from ASL. You have to stack to get a good firepower -- because TOTH
only has Single-Hex Firegroups.

(This is *not* designed to restart ASLML discussions on how much to stack. My view -- tactical situations will vary.
I really want dispersed fire for some of those situations.)

What about AI?
Well, Peter, for starters, you might want to leave the AI alone regarding how much it stacks, and leave it doing
Single-Hex Firegroups. See what user experience has to say. If the players start smashing the AI because the players
disperse and the AI stacks, then add some new AI behaviors.

Closing...
Thanks for listening, Peter. I know that bugs come first, etc... Just wanted to offer some thoughts on how to
tame the "UI beast" regarding Multi-Hex Firegroups. Hope this helps.

[blackcloud6]

As a side question in regards to fire groups, it's been awhile so I have forgotten but does SL or ASL have rules for multi-hex fire groups?

ASL does.