retreat problem in 1.040?

[HeroFromBavaria]

Hi everybody,

while playing the standard Total War campaign with version 1.040, we've observed a retreat situation we don't quite understand. A huge german force (40+ units) attacks Kharkov held by an almost equal large stack of Soviet troops. The retreat matrix shows Kursk as the most likely region, closely followed by Rostov (21+d21 vs. 17+d17). When a Soviet retreat is occurs, all the remaining troops retreat to the same area. As I understand the readme for 1.040 and WanderingHeads description in this forum, the region is checked for each unit individually, so even if some defenders end up in Rostov, the larger stack would retreat to Kursk. But it's always all of them either in Kursk or Rostov (luckily not Sevastopol [:)]), which is not likely at all (for Rostov).

Another observation was that repeating this attack from a save game always lead to a retreat to the same area just as if the combat resolution was resolved anew but the retreat roll was already determined (we observed this behaviour in the discontinued version of WaW where OP-Fire rolls would remain the same when reloading the game). By the way, we're playing with the simple supply rules (avoiding to much micro management in PBEM) so no influence due to supply units alters the retreat values.

Have I missed a rule or is this behaviour a bug?

[Lebatron]

If the defenders are forced out of a region they always go to one place, not several. The patches should not have changed this basic premise. Retreats were improved via the patches but not this basic aspect of design.  

[Lucky1]

Actually, I had the understanding that there was to be measure of randomization introduced into retreats. Not sure whether there were additional changes in this regard, but the following thread refers:

tm.asp?m=1992229

[Lebatron]

Yes I remember now, but as Hero was saying, retreating stacks, according to his observations, go to one place. This has also been my experience, although I may have missed a few times when a few defending units might have went to another adjacent region. I sometimes don't watch that carefully to catch this. But anyway, even though the randomness is there in theory, in practice it appears much less often than the odds would seem to indicate. Brian did mention that briefly in the thread you linked too. 

[HeroFromBavaria]

Thanks for Your replies. So it looks like the possible scattering of units during a retreat was intended but isn't implemented correctly in 1.040. In the Readme for 1.040 the feature (???) sections contains the following explanations:

Retreat metric in region popup: the metric computed to determine which possible region will be used for retreat is now shown explicitly to the player in the region popup. The retreat metric is randomly determined as x+die(x), where x is determined from population, factories, and rail in the candidate region and adjacent. The random value is evaluted per unit, so units might now retreat to more than one region.

Here, supply units aren't mentioned compared to WanderingHead's post concerning 1.030.

Maybe this minor flaw / inconsistency gets fixed with a future patch, if there will be any...