Slitherine

I have a few points and questions about combat statistics:

1) I think the variability of a combat outcome is much too high. In fact it may vary between no effect and the complete destruction of an intact enemy unit. Leaving the question of reality aside, this makes it nearly impossible to gain victories without too much exploiting the save/load function. In my opinion, things should be more predictable with some rare exceptions. A variability of +/- 1 with regard to the precalculated values would be appropiate. The current situation is just rolling a dice.

2. I wonder if the combat statistics are modified with the difficulty level. And if so, if these modifications are shown in the precalculation values. Btw, from the discussions in this forum I understand that the only modifications are AI performance (more "intelligent" and more aggressive at higher levels) and slower aquiring of experience at higher levels. Have I missed something?

3.) Having played nearly everything on different levels (from colonel upward) to me it appears that on FM level the combat outcome is in more than 90% "worse" than the precalculated values. This again indicates that either battle statistics change with difficulty level and/or this is not reflected in the precalculation values.

Any reply appreciated.

1) There's always the "chess" cheat. (Turns off random factor)

2) and 3) As far as i'm aware, diff has no DIRECT effect on combat results.
Only number of units and experience gain is affected.
No seperate AI for diff levels.

Many of us agree that the RNG is out of whack.

It needs to be toned down by 10% or remove the out-liers.

Try MP and you will see it is completely out of whack.

It does not balance out in a 25 to 45 turn game.

There are more negatives than positive outcomes.

It needs to be toned down by 10% or remove the out-liers.
There are more negatives than positive outcomes.

Seconded !

And yes there are more negatives than positive outcomes which kind of is annoying - to avoid some reloading actions I really just make 0:x or 1:xxx combats until absolutely necessary as the outcome is at least better predictable in such cases.

Anyway 10% toning down randomness at least ! And especially remove "out-liers" like prediction 0:5 - outcome 9:1

I know most people dont remember the good roll and tend to remember the bad ones since they can lead to unit loss. I think in a campain it should balance out since you have lots of battles, but it would be nice to have some option to see how is the RNG working (predicted vs actual battle results).

If you see lots of bad rolls in MP does that mean the other player is getting mostly favorable scores? I dont see anyone complaining about this on the forum

But removing the outliers might be a good idea.

I disagree with all these complains about combat results being too unpredictable. Yes, most of the results are not the predicted ones but they are close. There are also some large variations but they work against me as well as in my favor. And I support the idea of largely wild results as long as they are not the majority of outcomes because they add a pinch of realism to the simulation. Many planned actions on paper went wrong in reality due to a multitude of reasons, ranging from underestimation of enemy to simply a chain of unlucky events.

I give you an excerpt relevant for this discussion:

"[...] But these veneered images conceal the actual nature and ultimate determinants of the conflict. The first, quite simply, was wealth. The incalculable economic might of the United States sustained the Allied war effort to a level which ultimately overwhelmed the resources of Germany and Japan.

The second was chaos. This is not to deny the staggering feats of co-ordination and co-operation achieved during the conflict, or to belittle the extraordinary endurance of its participants. It is a measure of those achievements that they were made in the face of the destructive dynamics of the conflict. That millions all over the world, civilan as well as military, survived the widespread and ruthless chaos - the dislocation of societies, the devastation of economies, the brutal demands of warfare itself - is little short of miraculous, though it was done at incalculable cost. The confusions and fortuities of war, as well as its destructiveness, formed a great part of its momentum. Mistakes and miscalculations often played as great a part as strategy and tactics. Hindsight should not allow us to diminish the ruthless chaos, which presided over the action on all levels during WW2." - The MacMillan Dictionary of the Second World War

So, in face of a disastrous result in spite of favorable prediction, bite the bullet and move on. With strategy and perseverance you will ultimately prevail.

In the interests of not re-inventing the wheel, this topic has been debated at great length a number of times, the most length can be found here: viewtopic.php?t=28199&highlight=rdm

The upshot was that the RDM (Random Die-roll Modifier) would not be modified.

This may have been mentioned in the other post El_C mentioned (I haven't read all 6 pages yet ), but it struck me that a reasonable way to deal with the RDM which didn't got to the chess extreme was to implement an averaged system like in CEaW has in the GS mod. To quote from the manual:

Combat Results - Averaged Calculations Possible

Combat results for all battles except strategic bombardment can be based on 1 realization of the battle (i.e., 1 roll) or the average between 2 to 4 battles ( or rolls).

This number of battles (or rolls) is controlled by the variable in genera,txt:
NUMBER_OF_BATTLE_ROLLS

This means you will rarely get spectacular or terrible results.

The default for now, subject to more play testing is 3. Additionally, modified the attacker’s predicted losses so that it should be a more accurate predictor than before.

If I could be bothered I'd do the sums on what it means for the 0-3 roll in PzC (e.g., what is the combinatorial probability of 3 zeros?), but in general, it will tend to produce a Gaussian distribution around the expected ratios (according the the central limit theorem), and fewer extreme events (thought they remain possible).

Edit: Read all 6 pages, it wasn't mentioned there.

Combat Results - Averaged Calculations Possible

Combat results for all battles except strategic bombardment can be based on 1 realization of the battle (i.e., 1 roll) or the average between 2 to 4 battles ( or rolls).

This number of battles (or rolls) is controlled by the variable in genera,txt:
NUMBER_OF_BATTLE_ROLLS

This means you will rarely get spectacular or terrible results.

The default for now, subject to more play testing is 3. Additionally, modified the attacker’s predicted losses so that it should be a more accurate predictor than before.

THIS would be VERY much better than what it is now.

The total number of averagings might be even lower perhaps even 2 averaging would suffice. But I would start out with 3, too.