Strange battles in 1.05RC1

[Ace]

I have observed strange battle results in CW2 while analyzing this thread.

This is battle run on 1.04 as posted in the thread:
[ATTACH]32050[/ATTACH]

And here are results after running the turn in 1.05RC1:
Test 1)
Everything seems ok here, after one round CSA retreats - same as 1.04.
[ATTACH]32051[/ATTACH]
Battlelog
[ATTACH]32052[/ATTACH]

Test 2)
Here it gets different. For some reason, CSA fails to retreat until routed (there is viable 100% MC retreat options). The resulting battle is much bloodier.
[ATTACH]32053[/ATTACH]
[ATTACH]32054[/ATTACH]

Test 3 in next post. This is when it got weird enough to post this thread.

[Ace]

And finally Test 3 - the one that deserves deeper analysis:

For some reason, CSA remained in defensive posture. But despite having plenty of available retreat region, USA insisted on pounding until routed. To have so fundamentally different results only based on few retreat rolls seems wrong. If I would be asked which patch is more preferable right now. Despite so many good things done in 1.05., 1.04. is definetly the "more stable patch" (stable in a sense there are no wierd battle results).
[ATTACH]32055[/ATTACH]
[ATTACH]32056[/ATTACH]

Save game:
[ATTACH]32057[/ATTACH]

P.S.

Don't get me wrong. I think the path game has ventured in 1.05. is the right one, but this issue has to be looked into. I don't want simply to revert to 1.04, I want to make 1.05. work as it should

[pgr]

Just to be clear.... you ran the same order sets 3 times in 1.05 and got these three results?

Just out of curiosity, if you ran the same turn 3 times in 1.04, would the battles always play out in the same way? (I'm just wondering if there is a "normal" variation in battle results, given all the different things that go into a retreat roll)

[Ace]

I really don't know what would happen if I reran it in 1.04, but battles were never so erratic.

[richfed]

Yes! I have been experiencing these absolute bloodbaths. PLEASE, Mr. Pocus --- RC2??????

[Ace]

Every save with a bloodbath and backup trn can help Pocus debug why are units not retreating.

[pgr]

Well the man did warn us that the ZOC would make things harder on the guy retreating....although in this case, it sounds like there were friendly 100% to go to.

[pgr]

Hey Ace, I took a look at the first battle log... (I'm on an I pad tonight, so I couldn't open the other two)

You are wrong about the result of the first battle test. EJ did not retreat. After round 0, he attempted to retreat and didn't roll retreat. He is routed after round one (jump to the very bottom of the battle log), and then you get a series of interesting entries. He routs, retreat roll is automatic, he then looks for retreat destinations, apparently there are no valid destinations (!), and here is the kicker message... "EJ managed a tactical retreat, but has no retreat possibilities in another region, Staying in battle in Defensive Posture. Wallace does not keep attacking because his stack fails its commitment roll at the end test 1.

I'm assuming the same basic thing happened in turn 2 and 3 (in three EJ goes defensive). For some reason, it didn't think there were any valid regions to go to.

[Barca]

Don't get me wrong. I think the path game has ventured in 1.05. is the right one, but this issue has to be looked into. I don't want simply to revert to 1.04, I want to make 1.05. work as it should

+1

The Holy Grail would be more logical retreats (back down your line of communications, essentially toward friendly depots) without bloodbaths and unrealistic encirclements.

[Captain_Orso]

Hey Ace, I took a look at the first battle log... (I'm on an I pad tonight, so I couldn't open the other two)

You are wrong about the result of the first battle test. EJ did not retreat. After round 0, he attempted to retreat and didn't roll retreat. He is routed after round one (jump to the very bottom of the battle log), and then you get a series of interesting entries. He routs, retreat roll is automatic, he then looks for retreat destinations, apparently there are no valid destinations (!), and here is the kicker message... "EJ managed a tactical retreat, but has no retreat possibilities in another region, Staying in battle in Defensive Posture. Wallace does not keep attacking because his stack fails its commitment roll at the end test 1.

I'm assuming the same basic thing happened in turn 2 and 3 (in three EJ goes defensive). For some reason, it didn't think there were any valid regions to go to.

*ummmm* what?

Does the log say way Davidson, TN is disregarded? This would be the most logical location in which to retreat.

[pgr]

Ok, so I took a deeper look at Ace's battle logs, and edited them down to the relevant sections concerning retreat decision making.

Edit: Captain_O. It's almost as if Davidson, TN is simply unevaluated. See below.

Ace Test 1:
At the end of round 0, EJ tries to retreat, but is unsuccessful.
At the end of round 1, EJ is routed, the retreat is automatic... and then there is this interesting entry.

17:14:08 (Reporting) Side: 1000002 Confederate States of America is routing!
17:14:08 (Reporting) Group 1006326 E. Johnson' Corps is routed, retreat is automatic
17:14:08 (Reporting) E. Johnson' Corps succeeded in retreating
17:14:08 (Reporting) E. Johnson' Corps will take 46 hits while retreating (though no hits can be done on round 0)
17:14:08 (Reporting) Retreat Path Inspecting 2327 Dover Bend
17:14:08 (Reporting) Can't move into, discarded
17:14:08 (Reporting) Retreat Path Inspecting 2316 Middle Cumberland River
17:14:08 (Reporting) Can't move into, discarded
17:14:08 (Reporting) Retreat Path Inspecting 580 Sumner, TN
17:14:08 (Reporting) Retreat Path Inspecting 556 Warren, KY
17:14:08 (Reporting) Retreat Path Inspecting 547 Muhlenburg, KY
17:14:08 (Reporting) Retreat Path Inspecting 543 Trigg, KY
17:14:08 (Reporting) Retreat Path Inspecting 2313 Donelson Narrows
17:14:08 (Reporting) Can't move into, discarded
17:14:08 (Reporting) Retreat Path Inspecting 578 Stewart, TN
17:14:08 (Reporting) Retreat Path Inspecting 596 Davidson, TN
17:14:08 (Reporting) Retreat Path Inspecting 625 Humphreys, TN
17:14:08 (Reporting) E. Johnson' Corps managed a tactical retreat, but has no retreat possibilities in another region, staying in battle in Defensive posture
17:14:08 (Reporting) Commit Chance 1005411 L. Wallace' Corps 55 %, Rolled: 64 Not committed

The retreat path Inspections seem to be not happening (except to the river regions, who are discarded.) So E.J. stays in place. Wallace fails to commit, so the battle ends.

Test Battle 2
Largely same as above, except at round 0 E.J. seems to succeed in retreating. The "retreat path inspecting lines" are exactly the same as in Test 1 Round 1, and again the conclusion is "E. Johnson' Corps managed a tactical retreat, but has no retreat possibilities in another region, staying in battle in Defensive posture."

At the end of round 1, there is a different retreat log:

17:18:30 (Reporting) Battle in 579 Montgomery, TN Day: 1 Round: 2
17:18:30 (Reporting) 1005411 L. Wallace' Corps new target is 1006326 E. Johnson' Corps
17:18:30 (Reporting) Commit Chance 1008163 9. CSA Detachment 30 %, Rolled: 82 Not commited
17:18:30 (Reporting) Checking combat stance for United States of America in region 579 Montgomery, TN
17:18:30 (Reporting) Auto retreat triggered, TotalHits remaining: 719 Avg Cohesion%: 72 Base AutoRet%: 20 Hits taken: 203
17:18:30 (Reporting) Auto retreat: Hits received altered by the CiC's ROE retreat will 1003424 Lewis L. Wallace 80
17:18:30 (Reporting) L. Wallace' Corps There is no enemy on offensive, no need to retreat outside region.
17:18:30 (Reporting) Checking combat stance for Confederate States of America in region 579 Montgomery, TN
17:18:30 (Reporting) Auto retreat triggered, TotalHits remaining: 813 Avg Cohesion%: 48 Base AutoRet%: 20 Hits taken: 211
17:18:30 (Reporting) Auto retreat: Hits received altered by the CiC's ROE retreat will 1007249 Edward Johnson 100
17:18:30 (Reporting) 9. CSA Detachment There is no enemy on offensive, no need to retreat outside region.
17:18:30 (Reporting) E. Johnson' Corps There is no enemy on offensive, no need to retreat outside region.
17:18:30 (Reporting) Group E. Johnson' Corps has failed to retreat
17:18:30 (Reporting) 1005411 L. Wallace' Corps new target is 1006326 E. Johnson' Corps
17:18:30 (Reporting) Commit Chance 1008163 9. CSA Detachment 15 %, Rolled: 37 Not commited

No enemy on offensive, no need to retreat outside region. No retreat happens, yet the forces continue to engage. The battle ends after round 5 (at the end of the day)

What is interesting, is that later in the log E.J. retreats on day 13 from Humpharies TN:

17:18:30 (Reporting) Checking combat stance for Confederate States of America in region 625 Humphreys, TN
17:18:30 (Reporting) Side: 1000002 Confederate States of America is routing!
17:18:30 (Reporting) Group 1006326 E. Johnson' Corps is routed, retreat is automatic
17:18:30 (Reporting) E. Johnson' Corps succeeded in retreating
17:18:30 (Reporting) E. Johnson' Corps will take 25 hits while retreating (though no hits can be done on round 0)
17:18:30 (Reporting) Retreat Path Inspecting 594 Perry, TN
17:18:30 (Reporting) Initial value from MC: 100
17:18:30 (Reporting) Value after move cost: 100
17:18:30 (Reporting) Value after structures ownership: 100
17:18:30 (Reporting) Value after enemy presence: 100
17:18:30 (Reporting) Value after friendly presence: 100
17:18:30 (Reporting) Value after bottleneckness / Passage hub : 160
17:18:30 (Reporting) Not coming from this region, no change
17:18:30 (Reporting) ---> New best region: 594 Perry, TN Value: 160
17:18:30 (Reporting) Retreat Path Inspecting 2319 Waverly Run
17:18:30 (Reporting) Can't move into, discarded
17:18:30 (Reporting) Retreat Path Inspecting 578 Stewart, TN
17:18:30 (Reporting) Initial value from MC: 0
17:18:30 (Reporting) Value after move cost: 0
17:18:30 (Reporting) Value after structures ownership: 0
17:18:30 (Reporting) Value after enemy presence: 0
17:18:30 (Reporting) Value after friendly presence: 0
17:18:30 (Reporting) Value after bottleneckness / Passage hub : 0
17:18:30 (Reporting) Not coming from this region, no change
17:18:30 (Reporting) Retreat Path Inspecting 2327 Dover Bend
17:18:30 (Reporting) Can't move into, discarded
17:18:30 (Reporting) Retreat Path Inspecting 2316 Middle Cumberland River
17:18:30 (Reporting) Can't move into, discarded
17:18:30 (Reporting) Retreat Path Inspecting 596 Davidson, TN
17:18:30 (Reporting) Initial value from MC: 100
17:18:30 (Reporting) Value after move cost: 100
17:18:30 (Reporting) Value after structures ownership: 220
17:18:30 (Reporting) Value after enemy presence: 220
17:18:30 (Reporting) Value after friendly presence: 788
17:18:30 (Reporting) Value after bottleneckness / Passage hub : 1291
17:18:30 (Reporting) Not coming from this region, no change
17:18:30 (Reporting) ---> New best region: 596 Davidson, TN Value: 1291
17:18:30 (Reporting) Retreat Path Inspecting 595 Hickman, TN
17:18:30 (Reporting) Initial value from MC: 100
17:18:30 (Reporting) Value after move cost: 100
17:18:30 (Reporting) Value after structures ownership: 100
17:18:30 (Reporting) Value after enemy presence: 100
17:18:30 (Reporting) Value after friendly presence: 112
17:18:30 (Reporting) Value after bottleneckness / Passage hub : 198
17:18:30 (Reporting) Not coming from this region, no change
17:18:30 (Reporting) Retreat Path Inspecting 577 Henry, TN
17:18:30 (Reporting) Initial value from MC: 46
17:18:30 (Reporting) Value after move cost: 46
17:18:30 (Reporting) Value after structures ownership: 51
17:18:30 (Reporting) Value after enemy presence: 51
17:18:30 (Reporting) Value after friendly presence: 63
17:18:30 (Reporting) Value after bottleneckness / Passage hub : 131
17:18:30 (Reporting) Not coming from this region, no change
17:18:30 (Reporting) Retreat Path Inspecting 579 Montgomery, TN
17:18:30 (Reporting) Initial value from MC: 0
17:18:30 (Reporting) Value after move cost: 0
17:18:30 (Reporting) Value after structures ownership: 0
17:18:30 (Reporting) Value after enemy presence: -292
17:18:30 (Reporting) Value after friendly presence: -284
17:18:30 (Reporting) Value after bottleneckness / Passage hub : -284
17:18:30 (Reporting) Coming from this region but negative value, no change
17:18:30 (Reporting) Retreat Path Inspecting 592 Decatur, TN
17:18:30 (Reporting) Initial value from MC: 100
17:18:30 (Reporting) Value after move cost: 100
17:18:30 (Reporting) Value after structures ownership: 105
17:18:30 (Reporting) Value after enemy presence: 105
17:18:30 (Reporting) Value after friendly presence: 105
17:18:30 (Reporting) Value after bottleneckness / Passage hub : 179
17:18:30 (Reporting) Not coming from this region, no change
17:18:30 (Reporting) Region picked : 596 Davidson, TN
17:18:30 (Reporting) E. Johnson' Corps is retreating toward Davidson, TN
17:18:30 (Reporting) Group E. Johnson' Corps has retreated

This quote illustrates a log that is showing a full Retreat Path Inspection...that for some reason was not happening in Montgomery, TN.

Test 3 log:
Similar to the others, except that after round 0, the CSA seems to have the upper hand in power and does not try to retreat. Wallace seems to stick in the battle, even after routing, because of the "There is no enemy on offensive, no need to retreat outside region." message. But of course, the engagement keeps happening. It should be noted that there is a battle in Fayetteville AR, which has a fully delineated Retreat Path Inspection.

Conclusion:
There seem to be two odd things going on here. First, the retreat path inspection calculation didn't seem to fully fire for EJ in Montgomery, TN (but did fire in Humphreys, TN, and Fayetteville AR.) I wonder if the fact that Montgomery was 100% US controlled, and a river separated EJ from 100% CSA controlled zones had something to do with it. (Pocus did beef up the "bottleneckness" importance, and mentioned getting stuck surrounded by water would be a bad deal).

Second, the stacks are thinking that there are no offensive enemy stacks present, so there is no need to leave the region. I don't know if this is because the stances change, for some reason, after the retreat calculation is carried out...

Anyway, those are my thoughts... hope it helps. If you want to look at the edited logs, be my guest.

[Ace]

There are 2 things Pocus changed. One is retreat rules, the other is if the enemy is not in offensive posture, we will not retreat from the region. The problem with number 2 is that after Wallace concluded he wants to disengage, he remained in offensive posture regardless (he should switch to passive if he decided to disengage).

[pgr]

Of course, if Wallace was in Offensive, wouldn't EJ want to retreat? Both stacks wind up making the same conclusion. Is there a chance that the stacks are passave, and then they go offensive again? (Say because EJ is has 0%MC and wants to fight his way to 5?)

Or is it that combat ends only in 3 ways, a successful retreat, a successful rout, or the day ends. (You would think that if they both go defensive, the fighting should stop...)

[Captain_Orso]

Wallace went to DP after the battle--I think this is standard if loss of troops and/or cohesion is too high--so that would explain that.

If you look at EJ's status before the turn is executed, he's in Passive Posture with the Evade Combat SO, unless you've changed that Ace. After the battle he's in DP through, so I guess that is caused by the engine saying that since Wallace had gone to DP he should too.

What I found weird EJ going to OP or DP from PP. I can't remember ever having that happen to one of my stacks. Whether it has happened to my enemy's stacks... well you don't really get to see that, so maybe.

This whole thing is just going to take some getting used to.

[FelixZ]

Sure do wish, Pocus would weigh in on this thread and advise what is happening.

[ArmChairGeneral]

Could someone please cut and paste the new retreat algorithm for reference? I have not switched over yet, since it is not clear to me that the new rules are actually working properly. (BTW, except for occasional ping-ponging, I thought the 1.04 rules were fine once you knew what to expect.)

I am happy to adapt to new rules, but would like to know exactly what they are, and that they are working like they are supposed to. It seems that encirclement (though in a slightly different form) is back? Haven't wiped out a Corps sized stack since AACW; can't say that I missed it that much.

[FelixZ]

Could someone please cut and paste the new retreat algorithm for reference? I have not switched over yet, since it is not clear to me that the new rules are actually working properly. (BTW, except for occasional ping-ponging, I thought the 1.04 rules were fine once you knew what to expect.)

I am happy to adapt to new rules, but would like to know exactly what they are, and that they are working like they are supposed to. It seems that encirclement (though in a slightly different form) is back? Haven't wiped out a Corps sized stack since AACW; can't say that I missed it that much.

+1 for both of your paragraphs

[pgr]

Wallace went to DP after the battle--I think this is standard if loss of troops and/or cohesion is too high--so that would explain that.

If you look at EJ's status before the turn is executed, he's in Passive Posture with the Evade Combat SO, unless you've changed that Ace. After the battle he's in DP through, so I guess that is caused by the engine saying that since Wallace had gone to DP he should too.

What I found weird EJ going to OP or DP from PP. I can't remember ever having that happen to one of my stacks. Whether it has happened to my enemy's stacks... well you don't really get to see that, so maybe.

This whole thing is just going to take some getting used to.

EJ starts in passive (and attempting to cross to Humphreys, TN) , and Wallace starts in offensive posture. I loaded the test file myself in 1.05. And results vary. Sometimes there is no battle. Sometimes the battle goes two rounds, and sometimes it goes 6 and one of the two stacks ground down pretty good.

ArmChair, I'm not sure Pocus has put up the new calculations yet...it is a work in progress.

To my knowledge, Pocus did a few things:

First he fixed a bug that had the "return to" coefficient kicking in beyond the immediate turn (in a way movement history over multiple turns was being remembered...which was leading to the ping-pongs) By the way, I don't think this is causing what we are seeing here. Pocus published a little quick fix before publishing the bigger 1.05RC1, that basically just changed this aspect of the retreat code. I played around with it for a few weeks before 1.05RC1 was put out, and there wasn't any of this erratic behavior.

On top of this, Pocus re-activated ZOC calculations when evaluating retreat region destinations. This I suppose was to make retreating forces less "slippery," but less arbitrary than the min 5%MC rule (since ZOC and evasion can influenced by the player with force mixture). This would make encirclements relatively easier (which I'm not sure is a bad thing), but you would still have to surround the target region completely. This isn't the case with the example above. There are valid places to go to... but for some reason they weren't being evaluated (according to the logs).

Finally, apparently Pocus added a check of for an enemy in offensive posture in the region. If there isn't one, a stack won't retreat out of the province. The logs show this check is keeping stacks in the same region...which would explain why the battles are going on 6 rounds instead of 1 or 2.


So there is clearly something going on with this RC.... but it is a test patch at this point... I'm sure by the time we get to RC5 or 6 things will run smoother

(And for the time being, it's probably best to stick with 1.04... unless you feel like being experimental! arty