Odds changed when different unit selected first

[nukkxx5058]

Hi, I have 2 units (A and B) eligible for an attack.
If I select A before B the odds are very different (2:1 vs 4:1) than if I select B before A.
Why that ?
Thanks

[Voker57]

Different seed for the random generator that calculates the odds. combat will go the same.

[nukkxx5058]

Wait wait wait, how that the odds have some randomness in it ??

[eddieballgame]

When selecting the units, I pay attention to the offensive & defensive modifier numbers.
I have watched many videos where the attacker will just click on a number of units regardless of the modifier effects & go with that using only the 'ODDS' number as the determining factor.
It appears to me (& I can not say for sure) but the order in which you select your units will impact those modifiers, some will even cause those modifiers to drop, based on their troop type, supply, etc. ...will have an impact on the result.
There is 'rng'; but regardless, I will always note the modifier numbers before finalizing my attack.
This could be tested via a saved file, but then that 'rng' comes into play & may not be worth the effort.

[nukkxx5058]

But why on earth would you want to add any sort of randomization to the odds calculation ? It makes no sense at all to me. I'd like if Vic could clarify this.

Yes, odds are a very unperfect indicator in SE and you'd better look at individual modifiers too. The main problem with odds, I think, is when the recon level is low. Then odds are meaningless. This makes sense to me because if you don't know what exaclty is in front of you, you cannot assess the odds of victory.

But why would you want to randomize the odds calculation ?!

[eddieballgame]

I agree, recon is extremely important.
In fact, one tactic (?) that can aid in recon is sending in 1 unit just to 'look around'.
It increases one's recon for a small (?) sacrifice.

[JeanleChauve]

The calculation of the odds is done by by comparing the MULTIPLICATION of the % of the offensive modifiers (your units) with their offensive power AND the MULTIPLICATION of the % of the defensive modifiers with their defensive power (enemies) and all this in the order of their appearance. This implies much greater potential variations than if they were added together and increases the importance of rounding.

If you swap the units, you see that some modifier values ​​are very slightly different, around 2 or 3%, but this can cause huge differences the more attacking and/or defending units there are and the modifiers due to the fact explained in the first paragraph.

On the other hand, where the most important difference may be, and it may be a bug, is when calculating the 2 leader skill modifiers: if you have a unit without OHQ and the other with OHQ.
Indeed, it seems that the game averages the values ​​of the bonuses granted by the 2 commanders instead of adding them. This means that a unit with a high level SHQ combined with a lower OHQ sees its 2 modifiers lower than those of a unit without an OHQ.

Le calcul des chances se fait par la comparaison entre la MULTIPLICATION des % des modificateurs offensifs (tes unités) avec leur puissance offensive ET la MULTIPLICATION des % des modificateurs défensifs avec leur puissance défensifs (ennemis) et tout cela dans l'ordre de leur apparition sur le tableau. Cela implique des variations potentielles bien plus importantes que si on les additionnait et augmente l'importance des arrondis.

Si tu intervertis les unités, tu vois que certaines valeurs des modificateurs sont très légèrement différentes, de l'ordre de 2 ou 3%, mais cela peut engendrer d'énormes différences plus il y a d'unités attaquantes et/ou défensives et de modificateurs dû au fait expliqué au premier paragraphe.

Par contre, où la différence la plus importante peut se situer, et c'est peut-être un bug, c'est lors du calcul des 2 leader skill modifiers : si tu as une unité sans OHQ et l'autre avec OHQ.
En effet, il semble que le jeu fait la moyenne des valeurs des bonus octroyés par les 2 commandants au lieu de les ajouter. Ce qui veux dire qu'une unité avec un SHQ de haut niveau combiné à un OHQ plus faible voit ses 2 modificateurs inférieurs à ceux d'une unité n'ayant pas d'OHQ.

[JeanleChauve]

Not only is reconnaissance extremely important, but unit integrity and morale are not taken into account when evaluating odds. And it is these 2 parameters (+ posture) that dictate retirement.

The recon in force posture card is the best if you use the eddieballgame tactic.

[Voker57]

But why on earth would you want to add any sort of randomization to the odds calculation ? It makes no sense at all to me. I'd like if Vic could clarify this.

Yes, odds are a very unperfect indicator in SE and you'd better look at individual modifiers too. The main problem with odds, I think, is when the recon level is low. Then odds are meaningless. This makes sense to me because if you don't know what exaclty is in front of you, you cannot assess the odds of victory.

But why would you want to randomize the odds calculation ?!

OHQ and SHQ rolls are random, literally every step of the combat is a roll of the dice, naturally, you can't similate all this without at least some randomness. Note OHQ and SHQ bonus readouts can change if you change the order.

[JeanleChauve]

But why on earth would you want to add any sort of randomization to the odds calculation ? It makes no sense at all to me. I'd like if Vic could clarify this.

Yes, odds are a very unperfect indicator in SE and you'd better look at individual modifiers too. The main problem with odds, I think, is when the recon level is low. Then odds are meaningless. This makes sense to me because if you don't know what exaclty is in front of you, you cannot assess the odds of victory.

But why would you want to randomize the odds calculation ?!

OHQ and SHQ rolls are random, literally every step of the combat is a roll of the dice, naturally, you can't similate all this without at least some randomness.

There are no randomness in the odds calculations, modifiers values are allways the same even when you close then reopen the attack screen.
Randomness appears only when the battle begins.

[Soar_Slitherine]

The developer had this to say on the topic:

Those odds are just an estimation guys. And it is baded on simulating 1 combat round of battle. Changing the order of the units could result in slightly different results in that simulation since a lot of random rolls are involved and something like the butterfly effect might occur.

Most likely, the game uses a different seed for the random number generation in the combat simulation when units are added in a different order. I've seen no evidence of any mechanic that would change actual combat outcomes based on the order units are added in, and the outcomes are based on the starting conditions of the combat plus RNG, not on the odds calculation.

[nukkxx5058]

The developer had this to say on the topic:

Those odds are just an estimation guys. And it is baded on simulating 1 combat round of battle. Changing the order of the units could result in slightly different results in that simulation since a lot of random rolls are involved and something like the butterfly effect might occur.

Most likely, the game uses a different seed for the random number generation in the combat simulation when units are added in a different order. I've seen no evidence of any mechanic that would change actual combat outcomes based on the order units are added in, and the outcomes are based on the starting conditions of the combat plus RNG, not on the odds calculation.

OK. It's weird but I understand. So the odds are based on simulating one combat round, not on calculation. So not surprising it leads to such oddites. However, I have seen many cases where we are much closer to the butterfly effect than to a "slightly different result".
I've seen things as big as 2:1 vs 6:1 (or even bigger differences) according to the order in which you add the units. In this case, better not look at the odds at all. It shouldn't be displayed in big like that ...
But if odds are not reliable, What whould we look at to estimate the probability of success of the attack ?

[nukkxx5058]

One way to counter that would be to simulate the odds not with one single combat round but let's say 100 rounds and take the average values. This is called Monte-Carlo simulation. If 100 rounds make the calculation too long, what about at least 15 or 20 and take the average ? But usually Monte Carlo simulations are made with 1000's of computations so 100 should already give a very correct result while not be too hard on the CPU on modern computers. It of course depends on the size of the computation loop. I'm sure the proper number of simulations can be easily determined by trying various values of the number of simulations.

[KingHalford]

I'm gonna pipe in once more and praise the game for what might initially look like a very odd design choice. I really dislike deterministic combat resolution and I feel that these odds calculations end up being much more realistic in their wild inaccuracy, much like in real warfare, you might have some idea of what is going on, but once those guns start firing, all plans go out of the window!

[nukkxx5058]

I'm gonna pipe in once more and praise the game for what might initially look like a very odd design choice. I really dislike deterministic combat resolution and I feel that these odds calculations end up being much more realistic in their wild inaccuracy, much like in real warfare, you might have some idea of what is going on, but once those guns start firing, all plans go out of the window!

You should not confuse randomness in combat resolution, which is absolutely normal, and randomness in odds calculation. Odds are just here to give you the probability of sucess of your attack. There is absolutely no reason to have any sort of randomness in the calculation of the odds. Because odds are just a probability. Why would you want that the probability calculation to have any sort of randomness in it ? It makes no sense at all. A 5:1 odds means there's a 5/6 probability of success of your attack. It doesn't mean that the result of the attack is deterministic.
In shadow empire, the fact that odds calculation have some uncertainty in it is because for some reason, it was too difficult for Vic to compute the exact number. But you can be sure that if he could have computed the exact probability figure, he would have had. And again, this would certainly not make the combat outcome deterministic.

[BlueTemplar]

This has been discussed before, and it's a shame that this bug still hasn't been fixed :
https://www.matrixgames.com/forums/view ... 7&t=370306
(Please don't start a new thread when one already exists ?)

Remember that the odds do NOT simulate actual combat rounds, but rather use a single aggregated combat value for each troop, which leaves out quite a lot of detail that happens in real combat.

EDIT: I'm relaxing my hardline position on this somewhat, see details in the original thread.

[BlueTemplar]

The calculation of the odds is done by by comparing the MULTIPLICATION of the % of the offensive modifiers (your units) with their offensive power AND the MULTIPLICATION of the % of the defensive modifiers with their defensive power (enemies) and all this in the order of their appearance. This implies much greater potential variations than if they were added together and increases the importance of rounding.

[...]

Le calcul des chances se fait par la comparaison entre la MULTIPLICATION des % des modificateurs offensifs (tes unités) avec leur puissance offensive ET la MULTIPLICATION des % des modificateurs défensifs avec leur puissance défensifs (ennemis) et tout cela dans l'ordre de leur apparition sur le tableau. Cela implique des variations potentielles bien plus importantes que si on les additionnait et augmente l'importance des arrondis.

[...]

From what I've seen, "offensive power" (and probably also "defensive power") is actually Combat (lowered for infantry) for the attacker (and probably also for the defender, but I didn't check).

[...]
Indeed, it seems that the game averages the values ​​of the bonuses granted by the 2 commanders instead of adding them. This means that a unit with a high level SHQ combined with a lower OHQ sees its 2 modifiers lower than those of a unit without an OHQ.

[...]
En effet, il semble que le jeu fait la moyenne des valeurs des bonus octroyés par les 2 commandants au lieu de les ajouter. Ce qui veux dire qu'une unité avec un SHQ de haut niveau combiné à un OHQ plus faible voit ses 2 modificateurs inférieurs à ceux d'une unité n'ayant pas d'OHQ.

Interesting theory, but I don't see how this can result in negative bonuses from Commander(s) ?

Not only is reconnaissance extremely important, but unit integrity and morale are not taken into account when evaluating odds. And it is these 2 parameters (+ posture) that dictate retirement.

The recon in force posture card is the best if you use the eddieballgame tactic.

I'm not sure what you mean by "integrity", but Readiness is clearly there : it's the four 3 green triangles icons.

[JeanleChauve]

[...]
Indeed, it seems that the game averages the values ​​of the bonuses granted by the 2 commanders instead of adding them. This means that a unit with a high level SHQ combined with a lower OHQ sees its 2 modifiers lower than those of a unit without an OHQ.

[...]
En effet, il semble que le jeu fait la moyenne des valeurs des bonus octroyés par les 2 commandants au lieu de les ajouter. Ce qui veux dire qu'une unité avec un SHQ de haut niveau combiné à un OHQ plus faible voit ses 2 modificateurs inférieurs à ceux d'une unité n'ayant pas d'OHQ.

Interesting theory, but I don't see how this can result in negative bonuses from Commander(s) ?

I made a quick video to demonstrate why I'm thinking that there is a bug concerning the 2 leader skill modifiers:
https://youtu.be/VpurdYFfsVg
But my english is very poor.
What I would said here is the the leader modifiers tooltips should be: Average High command bonus for Rover IV is 51% AND average OHQ bonus for Rover IV is 18 (total average bonus = 60% (51+18/2)).

Not only is reconnaissance extremely important, but unit integrity and morale are not taken into account when evaluating odds. And it is these 2 parameters (+ posture) that dictate retirement.

The recon in force posture card is the best if you use the eddieballgame tactic.

I'm not sure what you mean by "integrity", but Readiness is clearly there : it's the four 3 green triangles icons.

Integrity is at 100% if the unit doesn't need replacement.

[BlueTemplar]

Thanks, I'll check that later...

Merci, j'y jetterai un coup d'œil plus tard...


Ah, my bad, the Manual at least indeed calls that Integrity, I should have searched it first !
Well, the aggregated Combat value is obviously (?) linear with the number of troops.
(Not sure if the odds calculation takes into account the malus you get with (non-AI Sentinel ?) units when under 3 troops ?)

Oups, en effet, l'Integrité est bien indiquée dans le manuel, j'aurais dû faire une recherche avant !
Ah, mais la valeur Combat est bien sûr (?) linéaire avec le nombre de troupes.
(Je ne suis par contre pas certain que le calcul des chances prenne en compte le malus habituel des unites ayant moins de 3 troupes ?)

[solops]

The slight variance in odds prediction is logical and brilliant. With the variance in in Intel, timing and troop performance I cannot think of any commander ever who had exact knowledge of either the odds or the outcome. The current system is great.