AGEOD Forums

In a thread about whether or not to use an Army as a primary combat stack, moni kerr wrote this:

There is a downside to this configuration. During combat an enemy force will not target the Army if there is an active Corps in the region.

A wily opponent would attack the Army stack and let the Corps mtsg in support. Once that Corps enters the region and joins in the combat, it's units will get all of the attention of the attacker and will not gain the benefit of the fortifications. The results can be quite devastating.

Wanting to test this, I turned off the AI in the Shiloh scenario, and assembled and maneuvered ASJ and his Corps into a line around Humboldt, TN. I then assembled a single stack under Grant, and sent him to attack ASJ’s stack stack at Humboldt. Holding Grant’s stack and behavior constant, I varied the CSA’s defensive setup to test Army behavior, defensive MTSG effects and overall battle engine behavior.

I organized Grant’s forces intentionally into a single large stack to test Army combat behavior, and sent him in as a single ~3000 PWR, 4 Division stack with 9 stack artillery (25 CPs). The Union has several more divisions available. The CSA has five potential Divisions but in all three scenarios some must MTSG to participate.

I have included the [ATTACH]26936[/ATTACH] if anyone wants to set up the forces differently and run a few iterations themselves. Unzip directly to CW2->Saves, and it will appear on the in-game Load screen as Testbed Z. (I am on 1.03 and use fred zeppelin’s roundel battle-flag mod; hopefully that doesn’t affect compatibility.)

[ATTACH]26937[/ATTACH]

The stacks are a little disorganized and not optimal. If I were doing this again, I would take the time to organize everything better before testing the battles.
I checked out splitting up Grant’s forces with Wallace and synch-attacking, but didn’t include it in the analysis because the force was slightly larger (they were somewhat more effective than Grant alone). I spent a lot of time looking at the round-by-round screens on the battle report trying to figure out how the Army stacks were behaving.

Continued...

General observations:
-This line from the AACWiki is simply not true for CW2 (or is the wrong way to explain what actually does happen):

It is important to note that a lone army HQ stack will never initiate combat by itself.

Grant marches into Humboldt in O/O posture and initiates combat every single time.

-MTSG does NOT happen on the first round of combat. MTSGers begin showing up on the second and later rounds.

-In the cases where Grant retreated he often took 35-50 pursuit hits. When the CSA retreated they took closer to 20, even though the CSA has 11 vs 15 cav for the Union. I can think of possible explanations, (Forrest is in ASJ’s stack while Grant does not have a Cavalryman for one) but Pursuit was not the focus of this testing.

-The second and later rounds usually resulted in ~50 hits for either side depending on who fought, etc. The first round was more variable. In scenario 1 it was 60-100. In scenarios 2 and 3 the main defensive stack had three divisions and first round damage was often well over 100, twice that of the later rounds, and would have been more if the range were longer (Humboldt = Woods).


Scenario 1 Results:
It is a close run thing for the CSA, and they take heavy casualties, but usually end victorious. Humboldt is Woods and in most battles two Divisions on the CSA side and two-and-a-half on the Union side were causing each other heavy casualties. I ran over a dozen iterations, but picked out these five for illustration (one was duplicated on accident).

[ATTACH]26938[/ATTACH]
On the first round ASJ’s 2 Div Army does 75-100 damage to Grant, but takes 30-50 hits in each division. On the second round when Breckinridge and/or Bragg arrive ASJ’s divisions are weak and damaged, while the MTSGers are not. Grant targets them with his fresh divisions and there is an even combat depending on who MTSGs.

ASJ’s stack is often not participating at all on round two since Grant’s stack is targeting the fresh MTSG divisions; and if/when ASJ returns in later rounds he no longer has the entrenchment icons on his Divisions and it is mainly his artillery that participate (based on the lack of cohesion hits his divisions take though the icons are lit.) This behavior could to be a combination of his Divisions’ reduced size/injured status (lack of combat element engagement) and Army combat behavior (stack participation in “support” of friendly elements).

It is important to note that not being entrenched does not disadvantage ASJ much: only the artillery are engaging, and then only in fire phase (Grant’s Divisions are assaulting other divisions) so entrenchments don’t matter. In fact, it may be that the reason the entrenched icons do not show up is because none of his participating elements used entrenchment (they are all artillery in fire phase) that round.

Still, his participation in the battle after rd 1 is minimal, meaning that the attacking divisions ARE turning to fight with the MTSGers after the first round, as moni kerr indicated. I think that if all the stacks were Corps, however the effect would be the same- the fresh divisions will tend to engage on rd 2, and if they are MTSGers, they will not have entrenchments. This needs to be explicitly tested in the future.

If no one MTSGs to support him on rd 2, ASJ withdraws southeast toward Bragg (he will withdraw before combat when set to B/O posture; I set him to B/R so there would be a battle). On the third round, all CSA stacks are equally damaged, and ASJ’s stack sometimes returns to battle. This continues, Grant trading hitswith the MTSGers with Grant usually withdrawing between rounds 4 and 6.

The worst losses for the CSA come on rounds when Bragg arrives alone and Grant targets him. ASJ’s stack does not participate (not targeted by Grant) leaving Bragg with one Division facing Grant’s 2 ish that fit in the frontage.

MTSGers also had a habit of targeting loose cavalry units, destroying them but wasting hits that could have gone to a Division instead. (Partway through I realized I needed to remove Corps command from Wallace so he would not MTSG in support of Grant with his disorganized stack, losing his wagons and cannon when the covering cav were targeted and killed).

Coming soon, analysis of two region defense methods.

It is important to note that a lone army HQ stack will never initiate combat by itself if in a region with another friendly unit.

I definitely encourage anyone to set up the Testbed in whatever configuration you are interested and post the results to this thread. I will analyze a couple of more setups myself, but there are many combinations and behaviors we can be studied using the [ATTACH]26939[/ATTACH] save. For that matter, the Testbed would probably benefit from a more organized set-up if anyone is interested in making a better Save to work from.

Ace,
If it is in a region with another friendly unit, it is not a lone Army stack .

The AACWiki says:

It is important to note that a lone army HQ stack will never initiate combat by itself.

In the combats I documented, Grant was in command of a lone Army stack and he initiated combat normally. I also did some tests splitting his force and synch-marching with a Corps under Wallace. Grant's Army stack targeted and engaged enemies in the first round, as did Wallace.

The AGEWiki says:

Army HQs will never attack by themselves and always fight in support of another stack, if any is present.

This appears to be what is happening when ASJ sits out the second round, although rather than "attack" and "fight" I think it really means "target an enemy stack during the stack selection phase."

Grant (the offensive stack) chooses one of the MTSGers. This would normally unlock other defensive stacks to be able to target an enemy stack that round, but the Army stack can only act in direct support. When it makes it's targeting decisionon the second round the MTSGers are not yet engaged, so the Army stack cannot act. On the third round an MTSG stack is re-engaging with Grant, so the Army stack can engage in support of it. The MTSGers are larger and thus more likely to fully engage with Grant's divisions, so in the battle screen ASJ's Divisions do not appear to be involved (his artillery are probably firing in support but he is not receiving return fire) although they are present on the battle screen.

I will run some tests and report on how a defensive Army behaves when there is a second stack in the region when Grant attacks. I did this some already but did not record the results. Based on what I saw, Grant chooses the larger stack to attack most times. I will run some tests with 3 Divs in Humboldt, one set where ASJ has the larger stack and another set where ASJ has the smaller stack and report back on what the Army does.

Than the manual does not say what it was intended.
It says lone army stack - it means empty stack with only Army general.

Excellent work ArmChairGeneral

An empty stack with only a 1* would also not engage in combat, so why would that behavior warrant a separate rule for the Army stack?

I am increasingly convinced that the important difference between Armies and other stacks is their behavior during the stack selection process at the beginning of each round of combat, but that it is complex and has not been summed up well in the two one-liners in the Wikis.

\varepsilon_i=y_i - \hat\beta_1 x_{i1} - \cdots - \hat\beta_p x_{ip}.

gives the spread on whether ACG's head explodes before 1.06.

I like the guy, so I'll take the over.

P. S.

Thanks for the awesome work, Comfy Guy.

Nice work. Interesting battles.

#2 and #3:



#2 Lone Army Stack
Some NM events. Pursuit hits 35-50.

Does fine, Army stays entrenched and fights every round.


[ATTACH]26992[/ATTACH]
[ATTACH]26993[/ATTACH]

#3 Two Corps Supporting Each Other and a Zero Army stack
No NM events. 35-50 Pursuit hits.

This situation corresponds to leader-only Army stacks, as well as to any formations where the Army stack is several regions from two Corps supporting each other (and so cannot MTSG).

[ATTACH]26994[/ATTACH]
[ATTACH]26995[/ATTACH]

#4 and #5:

# 4 Small Army, Large Corps
Some NM wins, Pursuit 40-55

Army only engages some rounds, Corps does most of the work


[ATTACH]26996[/ATTACH]
[ATTACH]26997[/ATTACH]

#5 Large Army, Small Corps
Some NM wins, Pursuit 40-55

Army engages most rounds, Corps engages most rounds.




[ATTACH]26998[/ATTACH]
[ATTACH]26999[/ATTACH]

# 6:

# 6 Army in Rear-Support

Many NM events. One of them scored 100 Rifles and 600 Prisoners, and another gained a promotion for Hardee.
Pursuit damage was 50-61(!) the highest of all configurations.

[ATTACH]27000[/ATTACH]
[ATTACH]27001[/ATTACH]

Final Analysis:

[table="width: 700, class: grid, align: left"]
[tr]
[td]Map# [/td]
[td]AVG Hits Landed[/td]
[td] AVG Hits Taken[/td]
[td]Difference[/td]
[td]Apparent Variance[/td]
[/tr]
[tr]
[td]1[/td]
[td]271 [/td]
[td]262[/td]
[td]9[/td]
[td]Medium[/td]
[/tr]
[tr]
[td]2[/td]
[td]242[/td]
[td]121[/td]
[td]121[/td]
[td]High[/td]
[/tr]
[tr]
[td]3[/td]
[td]241[/td]
[td]149[/td]
[td]92[/td]
[td]High[/td]
[/tr]
[tr]
[td]4[/td]
[td]283[/td]
[td]177[/td]
[td]61[/td]
[td]Low[/td]
[/tr]
[tr]
[td]5[/td]
[td]242[/td]
[td]151[/td]
[td]91[/td]
[td]High[/td]
[/tr]
[tr]
[td]6[/td]
[td]284[/td]
[td]197[/td]
[td]87[/td]
[td]Low[/td]
[/tr]
[/table]

Winner: Army (with half Cav Division) in Support
Prize: HQ Unit (+15 Cohesion Recovery, +2 CPs, +1 Exp per element per turn)
More hits = elements destroyed = NM
Low variance = consistent performance

2nd Place: Lone Army Stack
Prize: Balloon (+1 CP, Partially Lifts Fog of War (5 Detection), Looks Cool)
An Army without other stacks in region performs well on Defense, large hit differential

Last Place: 3 region MTSG coverage with Army as primary Defender
Booby Prize: Runner Card (Allows you to trade 1 NM for 8 WS)
Few hits infilcted, Many hits taken, sometimes loses battle. Don't spend those 8 WS all in one place.

Side Bet:
Granite Stater Wins:
Zero Army stack (#3) moderately effective; Army with Division in Support (#6) Very effective!
[ATTACH]27003[/ATTACH]

Grand Prize:
moni kerr for correctly diagnosing MTSG Corps behavior and inspiring this experiment:
3 Floyd's Dreadnaught, Indiana's Finest and most sought after beer
[ATTACH]27004[/ATTACH]

Final Analysis:

[table="width: 700, class: grid, align: left"]
[tr]
[td]Map# [/td]
[td]AVG Hits Landed[/td]
[td] AVG Hits Taken[/td]
[td]Difference[/td]
[td]Apparent Variability*[/td]
[/tr]
[tr]
[td]1[/td]
[td]271 [/td]
[td]262[/td]
[td]9[/td]
[td]Medium[/td]
[/tr]
[tr]
[td]2[/td]
[td]242[/td]
[td]121[/td]
[td]121[/td]
[td]High[/td]
[/tr]
[tr]
[td]3[/td]
[td]241[/td]
[td]149[/td]
[td]92[/td]
[td]High[/td]
[/tr]
[tr]
[td]4[/td]
[td]283[/td]
[td]177[/td]
[td]61[/td]
[td]Low[/td]
[/tr]
[tr]
[td]5[/td]
[td]242[/td]
[td]151[/td]
[td]91[/td]
[td]High[/td]
[/tr]
[tr]
[td]6[/td]
[td]284[/td]
[td]197[/td]
[td]87[/td]
[td]Low[/td]
[/tr]
[/table]
* I didn't actually calculate the variances, just eyeballed how close the results were to the average.


Winner: Map #6 Army (with half Cav Division) in Support
Prize: HQ Unit (+15 Cohesion Recovery, +2 CPs, +1 Exp per element per turn)
More hits = elements destroyed = NM
Low variance = consistent performance

2nd Place: Map #2 Lone Army Stack
Prize: Balloon (+1 CP, Partially Lifts Fog of War (5 Detection), Looks Cool)
An Army without other stacks in region performs well on Defense, large hit differential

Last Place: Map #1 3 region MTSG coverage with Army as primary Defender
Booby Prize: Runner Regional Decision (Allows you to trade 1 NM for 8 WS, yuck)
Good amount of hits inflicted, but many hits taken, sometimes loses battle. Don't spend those 8 WS all in one place.

Side Bet:
Zero Army stack (#3) moderately effective; Army with Division in Support (#6) Very effective!
Granite Stater Wins:
[ATTACH]27005[/ATTACH]

Grand Prize:
moni kerr for correctly diagnosing MTSG Corps behavior and inspiring this experiment:
3 Floyd's Dreadnaught, Indiana's most sought after beer
[ATTACH]27006[/ATTACH]