Attack rearmost - how does it really work?
 

It must be a popular idea to use your cavalries outflanking your opponent, kill his commanders and drive the troops to rout. But unfortunately the outflankers will often refuse to overcome the front line and attack them instead, which makes this tactic meaningless.

I remember someone said there should be a morale check for passing each hostile squard, but can anyone here explain me how does it actually work in algorithm level? Has the own morale of the outflankers influence on it? And if the enemy squard is mixed from different types of units?

Any comments will be greatly appreciated.

Re: Attack rearmost - how does it really work?
 

It will never work fantasticaly.

Try breaking your flankers into as many small groups as you can. Place tiny groups of them on the outside of a large group. Your chances improve of getting one of the small groups to slide well past the large group before choosing their target.

I also remember something about using a single commander with the units on guard-commander but I dont remember the results.

Snag the BattleSim map and try various formations and scripting.

Re: Attack rearmost - how does it really work?
 

Forgive me since I am more intending to ask the algorithm and the tread name is not so accurate.

If the guard-commander trick works, I will assume it does depend on the flankers' morale.

Thanks for the quick answer, could you please tell me how to find the BattleSim map? I have just checked pinned treads in the map forum but found nothing.

Re: Attack rearmost - how does it really work?
 

Re: Attack rearmost - how does it really work?
 

I don't know how the algorithm works, but pg. 74 of the manual makes it sound like ZOC could be a consideration. The edge of one formation passing by another one may get tangled in its zone of control, and perhaps that almost entirely tangles the formations. I've seen battlefield behavior that could be explained this way.

It could also be attack rearmost troops of nearest enemy formation rather than rearmost formation. I've seen behavior which looks like that as well....

Just speculation - not an algorithm or any sort of methodical knowledge - but might be stuff to look for during experimentation.

Re: Attack rearmost - how does it really work?
 

You'd think it would make sense to just tell them to go around the back, fighting anything along the way, then attack the closest thing.

That way if you put them on the flank they'd go forward in a straight line until reaching the end of the field, then turn around and attack the nearest stuff (aka the rear of the enemy).

Re: Attack rearmost - how does it really work?
 

I don't think its Zones of Control. I've seen small groups move down the side, then make a sharp turn and close almost to the center to reach a target.

I've also seen larger groups have one flank get involved in a fight when they come in contact with the enemy, while the rest of the group continues towards the rear.

Does the debug log shed any light on this? I've never checked.

Re: Attack rearmost - how does it really work?
 

As I understand it, 'attack rearmost' checks for each enemy formation they pass. If the check fails (?) they attack that formation.

Try this command with fliers, they will consistently attack enemy rear formations. (although 'attack archers' works even better, most of the time)

Re: Attack rearmost - how does it really work?
 

It would do if most fliers weren't utter dodos.

Re: Attack rearmost - how does it really work?
 

Fliers don't always make it to the rear either. I suspect the same check applies.

Re: Attack rearmost - how does it really work?
 

Yeah, I've also seen both of those behaviors as well.
Running with "-dd", the debug output doesn't show too much about how targets are chosen. It does mention that the squads are following the "Attack rear" orders, so at least we can be fairly confident that those are not being changed for some reason or other.

Anyway, I used Gandalf Parker's BattleSim map to run some experiments. There are still more to do, but here is the setup and preliminary results.

Setup
(1) To (hopefully) remove morale as a factor, only Morale 50 troops were used.
(2) To make battlefield movement instantaneous (and thus reduce number of checks and distractions due to battlefield dynamics), only flying troops were used.
(3) Only troops that could not damage each other were used.
IOW, squads of Air Elementals were a logical choice.
(4) All troops were commanded by Adepts of the Silver Order (mldr: 70) set to (hold)*5, "Cast" orders.
(5) All squads were set to "Hold and Attack Rear", except for one squad in only one of the armies, which was set to "Attack Rear".
(6) Squad placements and sizes were varied across experiments.

Results
(1) Squads ignore commanders when determining rearmost targets. If Silver Adept is behind all squads, he will not be targeted but a squad in front of him will be.
(2) Squads actually do target rearmost squads, given the conditions of the experiment.
(3) Squad size does not seem to affect targeting, given the conditions of the experiment.

More experiments need to be done with non-flying units, large squads jostling flankers, etc....

Re: Attack rearmost - how does it really work?
 

I appreciate the testing.

There used to be a command for "attack commanders" in the game but it was too powerful so it got removed from our options. Since that was a separate command, I would not be surprised if the command that was left in was specifically something other than that.

Re: Attack rearmost - how does it really work?
 

This highlights something interesting, though.

If you use assassins, to force your opponent to put bodyguards on important commanders, there will be squads of 5 units completely in the rear that your flankers can target, conveniently adjacent to those expensive mages you wanted to kill all along. ;)

Re: Attack rearmost - how does it really work?
 

Good point. It sounds well worth checking.

Re: Attack rearmost - how does it really work?
 

Well, Jim, my thoughts were much in the same direction. But, then I tested....:)

Setup
(1) Defender army given two squads of Mechanical Men.
(1a) One squad co-located with commander and set to "Guard Commander" orders.
(1b) One squad either at front or at rear of defender's side of the field. Set to "Hold and Attack Rear" orders.
(2) Attacker army given a squad of Air Elementals at rear of attacker's side of the field. Set to "Attack Rear".
(3) Sizes of defender squads varied.

Results
(1) Squads guarding commanders are ignored during squad targeting. :shock:
(2) Non-guardian squad always chosen no matter its size or position on battlefield.

I know the above assessment is probably going to be questioned. Maybe I'm overlooking something, but it seems to be a fairly clean experimental setup. I would certainly welcome independent verification of this.

My guess is that units with "Guard Commander" orders are specially flagged as guardians (for the purposes of assassination attempt resolution), and this flag must be getting considered by the squad targeting logic. (I've noticed that there is a #bodyguards map mod command which is distinct from #units, so this might lend further credibility to the hypothesis.)

Perhaps this isn't new to anyone, but another interesting observation is that there appears to be two kinds of targeting: squad targeting and individual targeting. A squad will attack another squad with all units initially, but after a round of combat, any non-engaged units of that squad will individually look for other nearby targets, including members of other squads. I think one sees this when flankers are trotting down the field and some get peeled off into nearby scrums, but it is very obvious in the experiments with fliers.

Re: Attack rearmost - how does it really work?
 

I'm not surprised that guards get ignored for purposes of Attack Rearmost, since they get ignored for purposes of Stay Behind Troops. Thanks for testing, though. Your findings with squad/individual targeting are also interesting.

-Max

Re: Attack rearmost - how does it really work?
 

I've generally had good results with attack rearmost with Hector and his elephants (if you couple them with a powerful attack force in the front), and with insane flying units, like a Gorgon. Hector generally bypasses indie front line troops and does attack the commanders in the rear, and a Gorgon seems to think front line troops are not worth the trouble, and immediately hits the rear. Other units... meh. They might attack anything on the way.

Re: Attack rearmost - how does it really work?
 

Hmm, I've always had problems with my mages and guards in the rear being attacked by fliers. At some point I placed a decoy squad behind the mages which seemed to help.

I've now tested this with three squads of Caelum infantry, two set guard commander with a commander on (hold)(hold)(attack rear) and one set to (hold and attack rear). One archer squad was also there.
Defender were four squads of Marverni infantry and one druid with guards at the very rear.
Result:

http://forum.shrapnelgames.com/attac...1&d=1232280979

The behindmost squad was those that was set to (hold and attack rear).

Re: Attack rearmost - how does it really work?
 

I did my testing with one squad of fliers set to immediate "Attack Rear" orders rather than "Hold and Attack Rear". Maybe there is some difference because of that, though it seems unlikely...?

Also, I haven't experimented with squads that are guarding a commander set to "Attack Rear". It is conceivable that commanders will seek out other commanders on the battlefield. Is this what you are implying?

Thanks for looking into this.

Re: Attack rearmost - how does it really work?
 

OK, Illuminated, I think I see what is happening in your test case. The number of attacking and defending squads seems to matter. In my earlier tests, I had 1 attacking squad with 2 defending squads to choose from. When I added a second attacking squad, then the defending bodyguards were targeted as well. So, I guess I would say target preference is given to non-bodyguards, but that bodyguards are not ignored.

http://forum.shrapnelgames.com/attac...7&d=1232294679
Attacker Squads (nearest to farthest): Air Elementals (size 2), Silver Adept + Mechanical Men, Air Elementals (size 4)

http://forum.shrapnelgames.com/attac...8&d=1232294896
Defender Squads: Mechanical Men, Silver Adept + Mechanical Men

http://forum.shrapnelgames.com/attac...9&d=1232294943
Two Attacking Squads (Attack Rear) Targeting Two Defending Squads (Hold and Attack, Guard Commander)

http://forum.shrapnelgames.com/attac...0&d=1232294999
One Attacking Squad (Attack Rear) Targeting Non-Bodyguard Squad (Hold and Attack)

Re: Attack rearmost - how does it really work?
 

I know that the combat replay is just a visual representation of a combat which has been decided by formula, and the visuals might not have actually been factored in. As far as "attack rearmost" and "hold and attack, rearmost" I think they are different but it might just be an impression that is worth testing.

My favorite is to use attack rearmost from one flank (usually left flank, upper screen), and hold and attack rearmost from the other (right flank). The fact that the enemy charges forward and seems to shift slightly toward the first attack appears to make a difference in what the second attack hits when I watch the battles. But I havent bothered to try and track numbers on the tactic. It seems to hit the back back squads more often, or at least the truly backside of the enemy frontlines.

Another thing Ive noticed (at least as far as AI) is the tendency to place archers to the top side of the screen (their right flank). Placing tramplers or cavalry with lances on the bottom side of my screen (my right flank) with orders to charge archers tends to take them across the entire enemy army on their way to the archers. Is that just visual impression or effective?

And thanks to you both for such great exampling.