Combat damage

[FrEDa]

Originally posted by mikel

If we combine the 2 sets of data then 0 damage occurred in (163 + 157) / (200 +200) = 80.0% of the hours.

If combat is resolved on an hourly basis then 1 GD blocked 80% of the shots (not 67.7%).

My questions are:

Is combat resolved hourly or is a daily resolution "allocated" to each hour?

If combat is resolved hourly is it worth doing anymore tedious tests with Attack = 1 and GD = 1?

I think we can assume around 80%. We can only get a better estimate by checking a huge number of round 1 (since STR reducation possibly influences the P(block)) battles.

Another good thing is that we can now eliminate the P(hit) (did-we-hit?) we've been talking about (perhaps it was introduced when you presented the data with round1 and round 0 mixed up). Unless, of course, there's something like that but it's restricted to when a unit has GD (unlikely since Paradox didn't design units to have 0 GD).

Regarding hourly vs. day, I don't think they have gone through the trouble of first simulating a whole day and then applying the results on an hourly basis. It wouldn't make any sense. It would also lead to strange results when 'retreating' in the midst, or at least force the code to be more complex as it must signal to all units that a unit has retreated and re-calculate a partial day and then apply again.

Right now I can't think of any reason we might want to do more single 1/1 tests. No, wait I have two reasons we might want to :

1. To check how STR reduction works (using a variety of STR values from start). It would
a) Confirm that A is always >= 1
b) Confirm the rounding scheme of HoI.

2. Calculate a very precise value for P(blocked). We have 80%, which should be a good estimate, but one that might be effected by STR reductions. Actual value might be higher due to STR reduction, or it might not if STR reduction only effects the value of GD.


But these things can rest.. as you should

/F

 

[unmerged(13914)]

I tried running my complex battle with blocking at 100 % (assumed X=80 + 20 bonus from Soviets to Germans).

As expected, the full-strength German units suffered no hits apart from one mysterious "nick" similar to the ones Mikel observed earlier. The German units that were eff = 1 % got smashed up in the usual way.

Then I tried X + 19, same outcome. One problem is that the expected number of hits per round is not very high -- 1 % get through of about 135 shots per hour, so it takes a number of hours to be sure.

But with X + 15, I immediately saw damage to the full-strength units. So I think the value has to be less than 85, although I couldn't rule out 81 at this point.

 

[FrEDa]

Originally posted by Math Guy
I tried running my complex battle with blocking at 100 % (assumed X=80 + 20 bonus from Soviets to Germans).

As expected, the full-strength German units suffered no hits apart from one mysterious "nick" similar to the ones Mikel observed earlier. The German units that were eff = 1 % got smashed up in the usual way.

Then I tried X + 19, same outcome. One problem is that the expected number of hits per round is not very high -- 1 % get through of about 135 shots per hour, so it takes a number of hours to be sure.

But with X + 15, I immediately saw damage to the full-strength units. So I think the value has to be less than 85, although I couldn't rule out 81 at this point.

Good, now we're closing in on it. But remember it's still possible that STR influences the P(block) value.

/F

 

[unmerged(13914)]

Yes, as usual it's quite difficult to prove a negative . . . to demonstrate that STR does NOT affect P(hit).

Still, a lot of progress in the past few days.

EDIT: By the way, I agree with you that the only possible combat scheme is strictly hourly. We're seeing new results calculated in real time each hour. One very good way to prove that is to save, then re-run the same battle. The results at any given hour are different each time.

 

[unmerged(14603)]

Originally posted by FrEDa
...Right now I can't think of any reason we might want to do more single 1/1 tests. No, wait I have two reasons we might want to :

1. To check how STR reduction works (using a variety of STR values from start). It would
a) Confirm that A is always >= 1
b) Confirm the rounding scheme of HoI.

2. Calculate a very precise value for P(blocked). We have 80%, which should be a good estimate, but one that might be effected by STR reductions. Actual value might be higher due to STR reduction, or it might not if STR reduction only effects the value of GD./F

We definitely need to understand the rounding scheme. When the Math Guy looked into leader modifiers he found rounding to be a major issue.

If you suggest the tests I will set them up but...

Originally posted by FrEDa
...But these things can rest.. as you should /F

I'm going to take your advice!

 

[unmerged(14603)]

Originally posted by Math Guy
EDIT: By the way, I agree with you that the only possible combat scheme is strictly hourly. We're seeing new results calculated in real time each hour. One very good way to prove that is to save, then re-run the same battle. The results at any given hour are different each time.

Now why didn't I think of that?

I agree we are close to understanding land assualt combat. However rounding is one issue we have to resolve.

Will be interested in any further results you get, especially with X + 16, X +17 and X + 18.

 

[FrEDa]

Originally posted by Math Guy
Yes, as usual it's quite difficult to prove a negative . . . to demonstrate that STR does NOT affect P(hit).

Still, a lot of progress in the past few days.

EDIT: By the way, I agree with you that the only possible combat scheme is strictly hourly. We're seeing new results calculated in real time each hour. One very good way to prove that is to save, then re-run the same battle. The results at any given hour are different each time.

I've been playing EU2 1.07 this weekend (first time in I don't know how long), but I think I'll do some test tomorrow with low STR units and see what we get. If there isn't any effect for low STR units then it's very likely there isn't one at all..

Yes, progress have been exponential But then again the whole combat system turned out to be a lot less complex than I first imagined it. In a week or so we should be able to produce combat-tables that show how much damage unit A will do against unit B, and that will be very helpful for modders.

/F

 

[unmerged(13914)]

Mikel, looking back I see that you did tests at X + 16 and X + 18 and got significant damage, and you did tests at X + 21 and got only 99 % "nicks," so there's not much left to do.

X + 17 wouldn't prove anything if X + 16 and X + 18 both showed damage; I did X + 20 and only got nicks, so it's down to whether it's X + 19 or X + 20.

I'll try to do some additional tests but I may have to do some actual work-work instead of HOI-work . . .

 

[unmerged(14603)]

Originally posted by Math Guy
Mikel, looking back I see that you did tests at X + 16 and X + 18 and got significant damage, and you did tests at X + 21 and got only 99 % "nicks," so there's not much left to do.

X + 17 wouldn't prove anything if X + 16 and X + 18 both showed damage; I did X + 20 and only got nicks, so it's down to whether it's X + 19 or X + 20.

I'll try to do some additional tests but I may have to do some actual work-work instead of HOI-work . . .

Couldn't resist so I set up some tests.

Inf vs Inf SA = 900 GD = 900 so I could see damage easily.
Varied gdeff with results as follows:

gdeff = +80 Avg Blocks = 99.0%
gdeff = +25 Avg Blocks = 99.0%
gdeff = +20 Avg Blocks = 98.9%
gdeff = +19 Avg Blocks = 99.0% 99.0 - 19 = 80.0
gdeff = +18 Avg Blocks = 98.0% 98.0 - 18 = 80.0
gdeff = +17 Avg Blocks = 97.1% 97.1 - 17 = 80.1
gdeff = +16 Avg Blocks = 96.2% 96.2 - 16 = 80.2
gdeff = -20 Avg Blocks = 60.4% 60.4 + 20 = 80.4

The standard number looks like 80% blocks with 99% blocks the max allowed. Makes sense when you think that effectivity is never allowed to go below 1%.

 

[FrEDa]

Hmm, looks good. Any news otherwise lads?

 

[unmerged(13914)]

Yes, here's a BIG footnote

I've found something interesting while working through the Russo-Finnish Winter War. I haven't posted the results yet, but in a big battle in Sortavala province I observed some really brutal attrition penalties imposed during combat.

Each day of a 3-day battle the Soviet (invading) side received a penalty of almost exactly 10 % per unit, at 1:00 GMT. So their total losses were 30 % plus whatever my Finnish defenders cost them. These couldn't have been critical hits because I only had two defending divisions, versus 13 Soviet attackers, and even between them my divisions didn't have 13 SA to fire per hour.

I believe these hits are due to low infrastructure, but they are MUCH bigger than the low-infra penalties received when your units are just sitting in the province. Sortavala has 20 infra and this normally results (as I observed in a series of trials) in 2 % attrition per day for friendly forces not native to the province, e.g. Germans in the 1941 scenario.

I find it interesting that the penalty is 10 % instead of 2 % . . . Have to do some experiments with different infra values.

I will be posting the details of the battle in a while, kind of short of time at the moment though.

See my Winter War Diaries AAR as well.

EDIT: I see I just got a promotion for my 500th post!

 

[FrEDa]

Ah, interesting. Keep us posted m8..

 

[unmerged(14603)]

I ran some tests with 1.05c that give us more insight into combat damage.

Land assault:

Armor and mech always defend as "hard" (HA is used to attack them). All other units defend as "soft" (SA is used to attack them).

Each SA/HA point that is matched by 1 GD point of the opponent has a 20% chance of hitting. When there are more SA/HA points than GD the excess SA/HA points have a 100% chance of hitting. Excess GD points have no impact.

The gound_def_eff modifier acts in a perverse way (it affects the opponent's defense). This means a +5 would cause the opponent to block 80+5=85% of all shots that are matched by a GD point.

Soft attacking soft or hard: strength damage/hour/hit = either (.075,.150,.225) with the average = .150

Hard atacking hard: strength damage/hour/hit = either (.075,.150,.225) with the average = .150

Hard atacking soft: strength damage/hour/hit = either (.075,.150) with the average = .1125

There are no "critical" hits in land assault combat. (Blitz combat was not tested).

Strength and organization level have no impact on damage inflicted.

Example: Inf vs Inf (soft vs soft) SA = 10, GD = 8. Expected average strength damage per hour = ((10-8)*.15) + (8*.2*.15) = .54


Naval (ship versus ship) combat:

Subs are always attacked by sub_attack points. All others are attacked by naval_attack (also called "sea attack") points. Naval_defense (also called "sea defense") points are used for defense.

The attack and defense points are matched in the same way as land assault combat.

Sub_attack: strength damage/hour/hit = either (.75,1.50,2.25) with the average = 1.50. There is a 5% chance that the unit will suffer "critical" damage instead. Critical damage = either (7.5,15.0,22.5) with the average = 15.0

Naval_attack: strength damage/hour/hit = either (.25,.50,.75) with the average = .50. There is a 5% chance that the unit will suffer "critical" damage instead. Critical damage = either (2.5,5.0,7.5) with the average = 5.0

Strength and organization level have no impact on damage inflicted.

Example: Sub (NA=10,ND=8) vs DD (SubA=10,ND=8. Expected average strength damage per hour to the sub = (((10-8)*1.5)+ (8*.2*1.5))*.95 + (((10-8)*15)+ (8*.2*15))*.05 = 7.83. Expected average strength damage per hour to the DD = (((10-8)*0.5)+ (8*.2*0.5))*.95 + (((10-8)*5)+ (8*.2*5))*.05 = 2.61


Air-to-air combat:

Air_attack points are used to attack and air_defense points are used to defend.

The attack and defense points are matched in the same way as land assault combat except that each attack point that is matched by 1 defense point of the opponent has a 34% chance of hitting.

Air_attack: strength damage/hour/hit = either (1,2,3) with the average = 2.

There are no "critical" hits in air-to-air combat.

Strength and organization level have no impact on damage inflicted.

Example: Fighter vs Fighter AirA = 10, AirD = 8. Expected average strength damage per hour = ((10-8)*2) + (8*.2*2) = 7.2