Damage Modifiers for Aliens


  • Please log in to reply
228 replies to this topic

#181 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 22 September 2006 - 03:58 AM

View PostMikeTheRed, on 21st September 2006, 10:51pm, said:

You could get real devious and edit the soldiers to 255 Melee Accuracy when you change Obdata to something low :D
That's a great idea! Though, 100% soldier melee accuracy is more than enough if you set the stun rod to 0%. :)

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#182 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 22 September 2006 - 07:43 PM

Hrm, I edited all my test soldiers to have 0% melee accuracy and ran a few limited trials. Basically, no effect. So then I edited obdata.dat [040] to 0 for the Stun Rod. Here again, no effect. In addition, the Stun Rod's accuracy in-game was still listed as 100%. I'm starting to think melee accuracy for the ol' prod is hard-wired in the executable somewhere. In any event, because my soldiers had 0% melee accuracy and still succeeded in nearly all their attacks, that would mean soldier melee accuracy is unused.  :D

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#183 MikeTheRed

MikeTheRed

    Sergeant

  • Members
  • PipPipPip
  • 95 posts
  • Location:Atlanta GA USA

Posted 22 September 2006 - 11:29 PM

Nods, looks like you nailed that shut, Z. Good work. Clearly soldier melee accuracy (SOLDIER.DAT bytes 50 and 60 added together and UNITREF.DAT offset 56) is not used by the game for the Stun Rod.

It is still semi-functional in the sense that it does give melee accuracy experience, but you've now shown this (both accuracy and gaining it through experience) to be entirely irrelevant/useless in a non-hacked game... there's only one MA weapon, and it doesn't use this accuracy stat.

It's also certain proof that OBDATA[40] is not used by the Stun Rod, either. Strictly speaking, it's not certain that OBDATA[40] is the melee accuracy field AFAIK. Still, we now know it isn't used by the Stun Rod - nor is there a 100 in any other OBDATA field for the Stun Rod. The simplest conclusion seems the best, just like you say... it's hardwired. Unless somebody ever finds otherwise.

So the only possible place melee accuracy might matter would be in a highly hacked game - if someone made other melee weapons, then tested the effect of these bytes. Even then, it still might not.

NKF has played with hacked melee weapons... I wonder if he'd have any insights.

Anyway, it's now clear that melee accuracy is entirely useless in a regular game of X-COM. No wonder they "hid" it.

Is it hidden in TFTD, too? I forget.

#184 NKF

NKF

    Mr. Badger in disguise

  • Site Staff
  • PipPipPipPipPip
  • 4,423 posts
  • Gender:Male
  • Location:In my mind

Posted 23 September 2006 - 01:18 AM

Any fiddling I've done with melee weapons hasn't really done anything major. All melee weapons use the same attack method as the stun rod (see TFTD's vibroblades).

I think it might have been mentioned earlier, but what of the melee accuracy when used by AI controlled units?

- NKF
NKF, narrow minded fuddy duddy who refuses to let go of the past and will not accept anything newer than 1979.

#185 MikeTheRed

MikeTheRed

    Sergeant

  • Members
  • PipPipPip
  • 95 posts
  • Location:Atlanta GA USA

Posted 23 September 2006 - 02:05 AM

Hiya NKF!

Right, I know aliens have been mentioned e.g. on the wiki. Another potential thing to play with.

Of course, when things get to the point of "irrelevant in regular games, and may or may not matter in highly hacked games", I guess it's up to whether some particular game hacker cares, shrug. There's plenty to work on in the regular game already.

Thanks again Z, great work. Good to see you, NKF!

#186 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 24 September 2006 - 02:58 AM

Aliens. Gotta check them to see if their integrated melee weapons are affected by either melee accuracy or melee weapon accuracy. I don't think the HTH aliens melee weapon accuracy has been found in the executable so melee accuracy will have to suffice for now.

It's very likely that the Stun Rod itself overrides any normal base chance to hit modifiers such as kneeling, wounds, and handedness. But this would be a nice thing to check with the aliens as their weapons may not be affected in the same manner. More testing is in order.  :D

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#187 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 09 November 2006 - 05:40 AM

View PostZombie, on 9th January 2005, 3:21am, said:

Are you absolutely positive that the Terror Units with turrets (Sectopod and Cyberdisc) have plasma weaponry? See, the shots do indeed look like plasma fire, and the UFOPaedia entries say these aliens are armed with a "powerful plasma beam". But the OSG (which is normally correct) says this about the Sectopod: "The weapon used by the Sectopod is a beam weapon, similar in function and power to terrestrial heavy-laser weapons". Now, if the Sectopod does use a laser-type weapon, it should do 1.5 times the listed damage against another Sectopod. If it only has a plasma-type weapon, it should do 0.8 times the listed damage against another Sectopod.

It's probably true that the Sectopod is equipped with a plasma weapon, but I'd like to find out for sure. The damage modifiers should help to narrow the weapon down to one type (160 max damage if plasma, and 300 if laser). Since our Sectopods are modified with no armor and 200 for health, there shouldn't be any insta-kills with a plasma-type weapon maxing out at 160, but there should be plenty with a laser.  :(
*chuckles*  :)

Remember this folks? Well, I just so happened to be editing tank stats a while back and noticed that Sectopods have a damage type of 3 which is identical to the Tank/Laser Cannon. Laser it is, so it seems Mr Ellis was correct after all! (The ammo type of the Sectopod is the same as the Cyberdisc, but that just governs what the shot looks like comming out of the turret). I'll need to run a few tests to verify this, but Sectopods should be 150% susceptible to their own weapon! No wonder they cripple so easily. It also shows the best weapon against the Sectopods is Mind Control: Once you gain control of a Sectopod, you can have it shoot itself with that nasty auto-shot capability. Sure, Laser Tanks have 110 laser damage, but they lack an auto shot. :D

So on a related subject, today I was fooling around with tanks in general. Tanks and aliens have a few different damage types. From obdata.dat we know the following (please excuse the poor formatting, the code tags are inoperative again):

Damage Type		  Value
Armour Piercing		   0
Incendiary				  1
High Explosive			2
Laser						 3
Plasma					  4
Stun						  5
I noticed that Celatids have a damage type of 7, so that expands the table a little bit more. But what could 6 be? I thought it might possibly be melee, but wasn't sure. Since unarmored soldiers have a susceptibility rating of greater than 100% for only 2 damage types (Melee: 120% and Celatid Acid Spit: 160%), they would be perfect targets (6 couldn't be Acid Spit since that damage is #7). Off I went to the executable and did a little bit of editing: 0 armor for normal soldiers, a damage type of 6 for the Tank/Cannon with 167% snap shot accuracy to overcome the 60% tank firing accuracy. Since I decided my soldiers will have 100 health, I figured that a damage of 49 for the Tank/Cannon would be perfect: 49*2(max modifier)*100(susceptibility rating to everything) = 98 max damage to soldier. Since that is below 100 health, a soldier should aways survive. However, if melee is indeed damage type 6, then the max would be the following: 49*2(max modifier)*1.2(susceptibility to melee) = 117. Therefore, soldiers should die 17% of the time with this scenario. Off I went to the executable and edited damage to 49.

Now I went to a mission to check my handiwork. Yup, soldiers had 0 armor, and the tank/cannon had the 100% snap shot accuracy. Great, time to test. It didn't take long and *aargh!* one of my guys died! That proves it, 6 is melee. Sweet, you learn something new everyday.  >:]

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#188 NKF

NKF

    Mr. Badger in disguise

  • Site Staff
  • PipPipPipPipPip
  • 4,423 posts
  • Gender:Male
  • Location:In my mind

Posted 09 November 2006 - 06:43 AM

Zombie, there was a question asked on the wiki that got me thinking. Melee stun damage. I know projectile stun damage has been tested extensively, but does the same ring true for the melee application of stun damage? I think this has been discussed, but has not found its way on the wiki damage section.

Wait a minute, do we have a separate section for explaining the ins and out of stun damage? Would it be duplicating too much work to have separate sections that describe all the intricacies of the different damage types?

- NKF
NKF, narrow minded fuddy duddy who refuses to let go of the past and will not accept anything newer than 1979.

#189 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 10 November 2006 - 02:05 AM

Got a link to that question NKF? I searched around for it, but it escapes me.

Not sure what you mean by Melee Stun damage though. Do you mean changing the damage of the Stun Rod in obdata.dat so that it pumps out melee damage instead of stun? Or do you mean the little bit of stun which gets applied to health when someone is attacked by a melee alien. It's probably neither of these, so please enlighten me.

There is a Stun damage page in the wiki. We also have sections on smoke, fire, explosives and melee. They all describe damage in some way. Do you want to make a page for just damage types? We can do that. Though we already have a damage page. :D

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#190 NKF

NKF

    Mr. Badger in disguise

  • Site Staff
  • PipPipPipPipPip
  • 4,423 posts
  • Gender:Male
  • Location:In my mind

Posted 10 November 2006 - 04:25 AM

A question was asked on the stun rod talk page is what got me wondering.

I was referring stun damage dealt by a stun rod. I mean, does it follow the same rules as normal non-explosive damage (* 2, * stun multiplier, -armour)? I'm sure it does, but I wasn't able to get reference through the damage section. This was what got the brain cells crawling.

Also, perhaps not a separate page for different damage types, but there should be somewhere where you can quickly access information on the various damage types. The individual damage type pages may want to be as self contained as possible. I suppose a few links on the damage page would probably be enough.

- NKF
NKF, narrow minded fuddy duddy who refuses to let go of the past and will not accept anything newer than 1979.

#191 MikeTheRed

MikeTheRed

    Sergeant

  • Members
  • PipPipPip
  • 95 posts
  • Location:Atlanta GA USA

Posted 10 November 2006 - 05:01 AM

NKF, I think I've seen more info on the Stun Rod around... but if I'd been asked who would know the most, it would have been you :D

I vote for keeping info on various Damage types on the Damage page. Perhaps with their own TOC sections, ok, but many are very straightforward and don't need a separate page. How much info can it be? Whenever you break things out into entirely separate pages, it's a lot more work keeping links straight, repeating info, etc.

Melee may warrant its own page. It's so different from all the rest. Then again, making this info be a section(s) on the Stun Rod page might keep confusion (organization) to a minimum. If one has so many references back and forth when making "theory" based pages, that the overall text is lessened by making one practical page... you see what I mean. I say less is better - as long as it makes sense. My two cents.

Talk about organization... panic/berserk/morale/rank/psi attack/etc. ... now that's one to think about.

#192 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 10 November 2006 - 06:25 AM

View PostNKF, on 9th November 2006, 10:25pm, said:

A question was asked on the stun rod talk page is what got me wondering.

I was referring stun damage dealt by a stun rod. I mean, does it follow the same rules as normal non-explosive damage (* 2, * stun multiplier, -armour)? I'm sure it does, but I wasn't able to get reference through the damage section. This was what got the brain cells crawling.
Ah, okay. Gotcha. Yes indeed, stun follows the normal non-explosive, non-incendiary damage types. The results of my trials are a few pages back in this topic: Stun Rod vs. Sectopod. Haven't tested it with a unit having a susceptibily to stun of anything other than 100% yet, but that is easily fixed. :)

Perhaps we should have a "Overview of damage types" page. That way, it pulls all the stuff together into one package, like the Overview of aliens" page does. Can't really go into too much detail with every ammo type on a single unified page otherwise it balloons way out of proportion (the Explosions page is literally huge, adding that info to a page with the other damage types would dilute the content). My vote is to have separate shorter pages for each damage types plus an overview page. :D

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#193 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 31 December 2006 - 10:39 PM

Hi folks! Remember this post a while back?

View PostZombie, on 19th March 2005, 9:57pm, said:

I gathered some values for the amount of stun added to a Quarterpod with HE-50 ammo. The maximum stun value I reached after 200 trials is 15. If you remember, HE-25 ammo managed half that number, or 7. So there definately is a link between stun and the maximum/normal damage from HE ammo. If you double the ammo strength, you double the stun inflicted.

I'm thinking that the maximum damage possible (including the damage modifier) is used for calculating the  maximum amount of stun.
For HE-25, max damage is 29. 7 points of stun / 29 max damage = roughly 1/4 or 25%.
For HE-50, max damage is 60. 15 points of stun / 60 max damage = 1/4 or 25%.

To put it another way, the maximum amount of stun added is equal to 1/4 of the maximum amount of damage inflicted by HE ammo. Since the minimum for the amount of stun added is 0, the stun range is 1/4 max damage + 1.

I'll have to try this with the Quarterpod standing on the ground to verify this, but I am almost positive that my determination is correct.

- Zombie
The theory behind stun added by damage is certainly not new but unfortunately I never really verified it. This past week I decided to hand-create a new testing scenario to use. It took a fair amount of work but is well worth it. Basically, I enticed a bunch of Snakemen to pay a visit to my base - albeit a heavily modified version. All the walls were removed from the 3 connected Hangars to make one gigantic testing "room". Then I modified the Heavy Cannon to have 130% accuracy for both Snap and Aimed shots, removed the clip restriction (converting it to a Poor-man's Laser Weapon) and changed the TU's to shoot to 2 (great for checking the number of shots taken). After Mind Controlling all 10 of the Snakemen which spawned (modifying all of them to have the same health and armor), I had them drop their weapons and get into one long firing line. I figured that a nice thing to test would be an Armor Piercing round since Snakemen are 100% susceptible, and made the HC-AP round to dish out a nice and even 100 damage points (100 listed, max of 200). Then it was just a matter of loading up BB's logger and shooting at each of the 10 aliens.

Theoretically, the amount of stun applied should be between 0-50 points. So far, after 3000 hits, the max stun I noticed was 49. So what is happening? Well, recall from farther back that as stun increases, the probability of the higher stun numbers happening is reduced.

Stun		0	   1	   2	   3	   4	   5	   6	   7
Percent   19.38   18.03   19.96   16.59   11.96   7.43	5.01	1.64
So it's 1.64% for 7 stun points with a range of 8 and max damage of 29. Now increase the tail-end of the curve out to 50 and you can see how rare numbers sitting at the end really are (a stun of 48 happened 3 times which equates to 0.1%). Just because 50 stun points didn't show up with with 3000 trials doesn't mean that it cannot happen - it is just rare. More trials will need to be conducted to find the answer.

That being said, I am confident that my theory of stun is correct - simply because the numbers can be calculated directly. The other day I worked out the equations to determine stun and they all hold true:

First is the initial calcualtion for the max damage a damage type can inflict. Let's take a look at the two common damage types: 1) high explosives and 2) either Armor Piercing, Laser or Plasma. Listed damage (LD) is the number given in-game while Max Damage (MD) depends on damage type:
Max Damage (MD)= INT(LD * 3 / 2) High Explosive
Max Damage (MD)= INT(LD * 2) Other Damage Types

Second comes the Damage Modifier (DM) calculation to determine the Adjusted Max Damage (AMD) applied to the unit in question:
Adjusted Max Damage (AMD) = INT(MD * DM / 100) All damage types

Finally is the calculation for the amount of stun added. The min is always 0 while the max is as follows:
Max Stun (MS) = INT(AMD / 4) All damage types

Taking all these equations into consideration, let's do a couple quick calculations. First is the Quarterpod (quarter of a Sectopod with a DM of 80% for HE) against HE-25 ammo.
LD = 25, DM = 80
MD = INT(25 * 3 / 2)
   = INT(75 / 2)
   = INT(37.5)
MD = 37

AMD = INT(37 * 80 / 100)
	= INT(2960 / 100)
	= INT(29.6)
AMD = 29

MS = INT(29 / 4)
   = INT(7.25)
MS = 7
Yup, that checks off. Now let's do a calculation for a Snakeman (DM of 100% for AP) against AP-100 ammo.
LD = 100, DM = 100
MD = INT(100 * 2)
   = INT(200)
MD = 200

AMD = INT(200 * 100 / 100)
	= INT(20000 / 100)
	= INT(200)
AMD = 200

MS = INT(200 / 4)
   = INT(50)
MS = 50
So far, this checks off too. :)

Looking at those low probabilities, I think I'm going to take a whack at automating the data gathering aspect for stun generation with high AMD damages and low MS percents. With less than a 0.1% probability of a max stun of 50 happening over 3000 trials, I'm going to need roughly 10 times the number of data points I currently have which equates to 30000. That's far too many to tackle alone, so automating seems to be warranted.

I guess this proves that age-old fact that shooting at an alien usually imparts some stun (91% of the time for my AP-100 ammo and 81% for the HE-25 example). (Of course, this neglects armor. Technically, anytime you are able to lower an alien's health you also add some stun. I should also mention that some aliens cannot be stunned either so their internal modifier overrides these calculations). If the alien's health is kinda low from a previous hit and you impart little to no damage on your next shot, you might be able to knock it unconscious if the stun roll is high enough. This would indicate to use a weapon which deals the bare amount of damage to constantly peck away at an alien until you get a favorable roll on the last stun determination. (You could also use a high powered weapon to initially wound the alien to allow for less shots, but you will only get one stun roll whereas the low powered shots allow many more stun rolls). :)

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#194 MikeTheRed

MikeTheRed

    Sergeant

  • Members
  • PipPipPip
  • 95 posts
  • Location:Atlanta GA USA

Posted 01 January 2007 - 06:21 PM

Great stuff, Z! It sounds like a ton of work. If you can get AutoHotKey to fire at successive aliens before saving the game, that would really save time. (Or did you do that already?) BB's system (logger and macro) would have been immensely useful for the psi testing. :)

When doing reaction training with a pistol, aliens become unconscious fairly often. Just like you say. To be entirely clear - the stun roll is separate from the damage roll?

Any thoughts on why it tails off, instead of being a random roll within that 25% damage range?

One thing you might also look into... you're using BB's Unitref logger, right? I don't think anyone's figured out the equation for Fatal Wounds. Maybe that data is also in front of you?

Good to see ya! And Happy New Year - Mike

#195 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 02 January 2007 - 02:22 AM

Hiya Mike, nice to hear from you again! I did some more work on this overnight and today and I think I finally have the stun issue figured out. So let's dive right in, it might explain some things. :)

Ok, after some fooling around with AutoHotKey for about an hour last night, I turned it loose on my 10 snakeman testing group for a few trials. After a little more editing of the .ahk script, I was able to optimize it with hardly any "sleep" time between the 10 shots (I think it's down to 15 or 10 right now and it still works fine) and I lowered the initial sleep time to something like 30 and the last sleep to 1500-2000. Those values were plenty long for my machine. Once optimized for maximum performance, I let the script run for about an hour and a half while I celebrated the new year. :)

When I came back, the script was still running and I had about 7500 values. So I stopped gathering data and did a quick number crunching. Bingo! There was the stun value of 50!

This was not the end of my stun research though. After sleeping on it overnight, I decided to run a brand new trial but this time instead of looking at only the stun, I would do a comparison between damage inflicted and stun. But I needed numbers, so I started the script while I ran some errands. When I returned, the script had stopped for some reason but I at least had 850 trials in (8500 shots). I got the script going again, ate lunch and returned. Yup, still going and now I had 11000 shots (1100 trials). That should be enough for a quickie-analysis.

Indeed, 50 stun was in there again. But now I took the damage values and placed the stun directly next to it for a side-by-side comparison in Excel. I sorted by damage (low-to-high) and took a look at the stun values for each damage number. What I saw was a little confusing at first, but after running a frequency distribution of the stun numbers in each damage level it started to make sense.

Basically, the game rolls the damage that a shot will do, applies the damage modifier and then applies this to armor. If damage is greater than armor it penetrates and does damage to health. In the mean time, it looks at the damage dealt and figures out the stun it needs to apply. It simply divides the damage by 4 and then truncates the result: Max Stun (MS) = INT(AMD / 4) to get the max, then does a another random roll between 0 and max stun to find the total stun.

For instance, if the AP-100 round did 0 damage, the stun would always be 0. The same holds true for damages of 1, 2 and 3. However, if the damage is 4, the stun is either 0 or 1. This holds up to 7 damage. When you get to 8, the stun is either 0, 1 or 2. For 12-15 the stun is 0, 1, 2 or 3. For 100 damage, stun is between 0 and 25. For 200 damage, stun is between 0 and 50. Notice that all the damages have the possibility to do 0 stun, so this is why it shows up so often. It's just that as the damage increases the possibility of hitting max stun is rather low since there is only one damage level which will allow the the stun roll to happen between 0 and 1/4 the max.

For my hypothetical AP-100 shell, it can dish out a max of 200 damage points for a total number of damage/stun combinations of 5151. Zero stun can happen 201 times out of that so the anticipated probability is 3.90%. One stun can happen 200 times out of 5151 for a anticipated probability of 3.88%... and so on and so forth. Once you get to 200 damage (the max), the stun value of 50 can happen in only one way out of 5151 for an anticipated probability of 0.02%! (See why this is so rare when the damage is high)? Out of 11,000 separate shots one would therefore expect 50 to show up twice. I believe I saw it 3 times so that fits well with this situation.

I think that about covers stun for now. I'm going to run a few trials with HE shells to see what happens when there is always a minimum damage, but the theory should still hold.

As for fatal wounds, recall that aliens cannot get them so I'll need to shoot my own men for this scenario (well, I could MC those snakmen but that would be no fun). All in the name of science, eh? I can be really ruthless sometimes. Will work on this next. :)

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#196 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 04 January 2007 - 02:13 AM

Ok, I'm back with some more news.

Firstly, I'm unsure if my analysis on the probabilities for the AP-100 stun numbers are correct from my previous post. I need to look at the results a little closer to see if anything else is going on in there besides a simple probability. Give me a couple days as there are still a lot of numbers to crunch. Looking back, I realize that I should never do tests (automated or manual) with weapon strengths that high as you need to crunch so may more numbers to get to the root of the theory. Oh well, you live, you learn. :)

Secondly, I decided to take the easy way out for testing fatal wounds. All those Snakemen were sitting there so pretty, so I just had to MC all 10 of them and restart the automated trials. Here again, I was using the AP-100 shells for testing Fatal Wounds (FW), so I ended up collecting 11000+ values to gain any type of understanding. (My previous trials at the Xcomufo.com forums were using a Heavy Laser with 85 listed damage (170 max) so it also suffered from lack of data).

So what did I detemine with those values? Looking at FW and damage side-by-side, and then sorting by damage showed that between 0 and 10 damage points yielded a fatal wound range of 0 to 3. Any damages higher than 10 and there were only between 1 and 3 fatal wounds. Closer inspection showed that I needed more data within the 0-10 range so I am currently running a modified test with AP-5 ammo (max of 10). Preliminary results (5550 values) indicate that 0 damage always = 0 FW (duh!) while a damage of 1 approximately = 0 FW 90% of the time with the remainder of FW between 1 and 3. A damage of 2 yielded a 80-85% chance of having 0 FW, a 70-75% chance of 0 FW with dam=3, etc etc etc all the way up to 10 damage which showed an 8.85% chance of 0 FW with 1-3 FW being the primary range 91% of the time (or a 30% chance of either a 1,2 or 3). I'm not sure of the excat percentages involved within the 0-10 damage range, but the tests I'm running now will hopefully shed some light on the subject. 15000-20000 numbers should nail those FW percentages down more, right? :)

Like I mentioned before, for damages greater than 10, the number of fatal wounds was always between 1 and 3. The probabilities suggest an even breakdown (33% chance) of hitting any single value. I haven't looked too close at where FW were found on the body, but it seems like a random chance of any body part. With a 161% final firing accuracy (which includes the kneeling modifier) and situated only 3 tiles away I doubt drift plays any kind of role of where a wound will happen. It's just random.

Anyhow, I'll keep you updated on my progress. Better results should be on their way soon! :)

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#197 MikeTheRed

MikeTheRed

    Sergeant

  • Members
  • PipPipPip
  • 95 posts
  • Location:Atlanta GA USA

Posted 04 January 2007 - 08:05 AM

Hi Z, more great stuff... I can hear your elation at getting 10 soldiers at a time automated. For me it was like when I found I could modify terrain to see explosive damage, at least in a crude way... and then BombBloke blew that away with his numerical tiles, which was even more elating, lol.

Sometime you will have to post the script you made. I may need something like it some day. Also of course your findings should make it onto the wiki. Maybe the Discussion tab for the Stun page would be a good place for the .ahk? Just a thought.

FWIW I have generated tons of firing accuracy data as per that other thread here, but have not had time to analyze it yet and can't get to it real soon. :) If anybody wants the raw data, just ask.

I guess we all suspected X-COM was doing something simple under the hood with Stun, and voila, it was. One question though: You have been testing on targets with 0 armor, right? I see your AMD equation re: stun but... are we sure it doesn't occur, regardless of armor? If someone with Kevlar IRL is shot, they are "stunned" (knocked down a little while, etc.).

Now that I think about it relative to your results, I see that in Reaction Training, there isn't an increased likelihood for the many small-damage pistol shots to stun, per se. The reason those targets are stunned more often is actually due to the fact that the pistol is less likely to kill them with "the final" shot; it's more likely to fall at the point where stun kicks in. So, even though a lot of small shots causes targets to be knocked out more often, it's not because small shots cause more stun than usual, it's just that low damage is less likely to make an outright kill, and more likely to land the target in the stun zone instead, versus heavier weapons.

That kind of data summary - to show correlations between health damage and stun - is real easy in MS Access. I'm telling ya, you should try it. You're a very data oriented person. GMTA. :)

I bet that Fatal Wounds is doing a range-bounding slider. Smells like the Experience one... it slid up to where 11 experience points caused 2-6 skill increases, and that's it... getting more than 11 didn't matter. But FW is also dependent on health damage, which clouds the formula. You probably already have enough data for someone to make a guess, but you're fixing to get tons more, so knock it out of the park for us. :)

#198 MikeTheRed

MikeTheRed

    Sergeant

  • Members
  • PipPipPip
  • 95 posts
  • Location:Atlanta GA USA

Posted 07 January 2007 - 12:50 AM

Hey Z,

If you have the data (and/or, for future trials, if not),

Please consider capturing what armor facings are hit, in your face-to-face scenario. Or whatever your scenario is. Is it always the Front or ... ?

It's recently become obvious in a thread on XCOMUFO (see post 122 ff. near the end) that my wiki Kill Modeling page might be ok for Experience Training, but it's not good enough for regular combat. For one thing, it focussed on low power weapons. For another thing, it assumed you're trying to keep aliens alive, and therefore deliberately spread out which armor facings would be hit.

It's not applicable to real battle... but I could extend it to that. Yet, how does real battle work?

I did some testing a while back, but I never followed it all the way through, and the notes are lost. I seem to recall that side armor can be hit a small percent of the time, even when facing an enemy "front to front". If you can assist with any numbers here, I can work them into my kill model when I generate new numbers more suitable for real battle.

Thanking you in advance!

   MTR

#199 Zombie

Zombie

    Mr. Grognard of X-COM

  • Admin
  • PipPipPipPipPipPip
  • 5,529 posts
  • Gender:Male
  • Location:Wisconsin, USA

Posted 07 January 2007 - 08:44 PM

View PostMikeTheRed, on 4th January 2007, 2:05am, said:

Sometime you will have to post the script you made. I may need something like it some day. Also of course your findings should make it onto the wiki. Maybe the Discussion tab for the Stun page would be a good place for the .ahk? Just a thought.
Well, my script is just a slightly modified version of Bomb Bloke's which can be found here. Posting up just the .ahk script is meaningless unless there is a saved game to go along with it as everyones testing scenario will be set up differently. In addition, some "optimizations" I made to the script might not work with everyones machine so these should be changed accordingly (I'm thinking about the sleep lines in particular).

View PostMikeTheRed, on 4th January 2007, 2:05am, said:

I guess we all suspected X-COM was doing something simple under the hood with Stun, and voila, it was. One question though: You have been testing on targets with 0 armor, right? I see your AMD equation re: stun but... are we sure it doesn't occur, regardless of armor? If someone with Kevlar IRL is shot, they are "stunned" (knocked down a little while, etc.).

Now that I think about it relative to your results, I see that in Reaction Training, there isn't an increased likelihood for the many small-damage pistol shots to stun, per se. The reason those targets are stunned more often is actually due to the fact that the pistol is less likely to kill them with "the final" shot; it's more likely to fall at the point where stun kicks in. So, even though a lot of small shots causes targets to be knocked out more often, it's not because small shots cause more stun than usual, it's just that low damage is less likely to make an outright kill, and more likely to land the target in the stun zone instead, versus heavier weapons.
Bingo, you described it perfectly! Yes, all my tests were with 0 armor, but I'm pretty sure that armored units still get stun from weapons. Damage modifiers should play a role here too. Haven't tested with anything other than MC'd Snakemen either, so that may or may not play a role with X-COM units wearing X-COM manufactured armor.

Ok, back to fatal wounds. My computer seemed to enjoy gathering damage/fatal wound values so I let it continue until it reached 15000 values (well 15010, but close enough). After that, it was my turn to crunch some numbers. First of all, here is the frequency distribution of fatal wounds with a simple AP-5 shell.

Frequencies
		Fatals
Damage	 0	  1	  2	  3	 Total
   0	 1355	  0	  0	  0	 1355
   1	 1272	 37	 40	 47	 1396
   2	 1107	 80	 78	 90	 1355
   3	  957	117	105	133	 1312
   4	  881	167	188	159	 1395
   5	  748	205	210	189	 1352
   6	  674	254	237	240	 1405
   7	  480	269	306	296	 1351
   8	  384	306	355	331	 1376
   9	  241	386	375	366	 1368
  10	  123	421	404	397	 1345
									  15010
As you can see, the frequency of 0 fatal wounds decrease steadily as damage increases. But why this focus on 0 fatal wounds, you may be asking? Well, since damages > 10 never produce 0 fatals and since the frequency is evenly distributed between 1 and 3, it makes no sense to concentrate efforts on those values. However, for damages <= 10, the 0 fatal wound scenario is the linchpin which determines the probability of 1-3 fatals happening (if this makes any sense), so it pays to take a closer look.

Obviously, simple frequencies do not show much except for a general understanding of the situation. To investigate further, I broke the distributions down into percents of the total for each damage.

Percentages
		Fatals
Damage	 0		 1		 2		 3
   0	100.00%	 0.00%	 0.00%	 0.00%
   1	 91.12%	 2.65%	 2.87%	 3.37%
   2	 81.70%	 5.90%	 5.76%	 6.64%
   3	 72.94%	 8.92%	 8.00%	10.14%
   4	 63.15%	11.97%	13.48%	11.40%
   5	 55.33%	15.16%	15.53%	13.98%
   6	 47.97%	18.08%	16.87%	17.08%
   7	 35.53%	19.91%	22.65%	21.91%
   8	 27.91%	22.24%	25.80%	24.06%
   9	 17.62%	28.22%	27.41%	26.75%
  10	  9.14%	31.30%	30.04%	29.52%
Do you see it? The answer is there in the form of a simple multiplication table. (The pattern is 9, 18, 27, 36, 45, 54, 63, 72, 81 and 90 if that makes things any clearer). I saw this straight off after running the 15000 numbers through Excel, but failed to recognize any pattern with the preliminary 5550 numbers as the percentages worked out to 100%, 90,63%, 82.20%, 72.25%, 64.57%, 57.20%, 47.61%, 36.36%, 24.38%, 16.27% and 8.89% which is why I needed to gather more data points. After this very lucky break, everything else fell into place as the anticipated probabilities could now be computed from the theoretical 0 fatal wound column.

Theoretical Percentages
		Fatals
Damage	 0	 Ant 1,2,3   Act 1,2,3   Ant 1-3   Act 1-3
   0	 100%	  0.00%	   0.00%		0%	  0.00%
   1	  90%	  3.33%	   2.96%	   10%	  8.88%
   2	  81%	  6.33%	   6.10%	   19%	 18.30%
   3	  72%	  9.33%	   9.02%	   28%	 27.06%
   4	  63%	 12.33%	  12.28%	   37%	 36.85%
   5	  54%	 15.33%	  14.89%	   46%	 44.67%
   6	  45%	 18.33%	  17.34%	   55%	 52.03%
   7	  36%	 21.33%	  21.49%	   64%	 64.47%
   8	  27%	 24.33%	  24.03%	   73%	 72.09%
   9	  18%	 27.33%	  27.46%	   82%	 82.38%
  10	   9%	 30.33%	  30.29%	   91%	 90.86%
1,2,3 means the probability of 1, 2 or 3 fatal wounds, while 1-3 means the combination of all fatal wounds. Everything seems to match quite nicely, so I'm fairly certain that the theory is sound. I suppose the link between Damage and the probability of a fatal wound happening could be described with the following equation:

0 Fatal wound probability = (11 - Damage) * 9

Of course, this doesn't apply to the 0 damage scenario as the result would be 99% instead of 100%. It also doesn't apply when damages are > 10, but the negative percentage could be treated as a 0 for the calculation in which case everything is fine and dandy. After the probability determination if 0 fatal wounds can happen for a damage level, then the game would go about actually calculating the number of fatals. (Grab a random number between 0 and 100 inclusive. If it falls within the 0 fatal probability, assign 0 as fatals; otherwise, grab another random number beween 1 and 3 (again, inclusive) and apply that as the number of fatals). Simple stuff really, but it is unknown whether the game actually uses this format or something else entirely. Still, it's not too bad as we at least can now model fatal wound generation. Cool, eh? Nevermind, it's cool to me at least. :)

- Zombie

My X-COM Patch Kit For UFO Defense | Emergency XCOM Meeting spoof on YouTube




JellyfishGreen said:

Zombie: Empirical data's your only man, when formulating a research plan.
A soldier's death is never in vain if it makes the formula more plain.
A few dozen make a better case for refining that third decimal place.
They call me Zombie because I don't sleep, as I slowly struggle to climb this heap,
of corpses, data points, and trials, but from the top - I'll see for miles!

#200 Danial

Danial

    Resident Muton

  • Fan Fiction
  • PipPipPipPip
  • 271 posts

Posted 08 January 2007 - 08:02 AM

I agree... That is very cool :-)

Nice work as always, Zombie. Kudos.

So with what you're saying about numbers greater than 10, does that the next line would be:

Damage	0	 Ant 1,2,3   Ant 1-3
 >10	   0%	  33.33%	  100%

Just checking to see if I've gotten the grasp of it.

P.S. Have you posted this table and equation on the wiki yet?




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users