Home » Stars! 2.6/7 » The Academy » Missile hit rate vs additional target tokens
| Re: Missile hit rate vs additional target tokens |
Tue, 14 March 2006 04:56   |
|
 m.a@stars |
 | Commander |
Messages: 2765
Registered: October 2004
Location: Third star to the left | |
|
| Kotk wrote on Mon, 13 March 2006 20:36 |
| LEit wrote on Mon, 13 March 2006 19:56 |
Where the 10 FFs in one stack? It sounds like it. And if so, it means that my theory is probably wrong too, although I suppose it is possible to get 9 misses in a row with Jihads. Could you repeat the test with 100 FFs (in one stack) and 100 chaff (in 100 stacks)?
|
Actually i cant reproduce the Staz outcome with my JRC4 unless these frigates are split out. 
Basically: 10 uncomputered Jihads against 1000 frigates in stack seem to love to kill none, 3, 7 or 10. I get only these discrete kills, nothing else. Seems that:
if none hit they kill none,
if 1 hits they kill 3,
If 2 hit they kill 7,
if 3 or more hit they kill 10.
Also ... the stack seems to get damage (when less than 10 is killed)!  So missing jihads still seem to aid the hitting ones to kill stack. 
|
So, if I read you correctly, your FFs are unstacked and firing against a 1000-chaff stack, right? And you're getting "extra" kills when one single missile misses but not when one hits?
That would mean there's some statistics of big numbers at work which my simple 2-chaff-stacking tests didn't show. 
It would also mean that the "one-missile-one-kill" rule has a devilish twin "one-miss-several-kills" rule that nobody has ever accounted for. 
So many Stars, so few Missiles!
In space no one can hear you scheme!  Report message to a moderator
|
|
|
|
| Re: Missile hit rate vs additional target tokens |
Tue, 14 March 2006 08:04 |
|
| Kotk | 
| Commander |
Messages: 1227
Registered: May 2003 | |
|
| m.a@stars wrote on Tue, 14 March 2006 11:56 |
So, if I read you correctly, your FFs are unstacked and firing against a 1000-chaff stack, right?
|
Actually i used stacked 10 heavily shielded and deflected nubians. Each had one jihad.
I fired at stack of 1000 frigate chaffs. Result:
0,3,7 or 10 chaffs destroyed and stack gets damage on case of losses of 3 and 7. When stack got more and more damaged during battle then also 4 and 8 started to die.
I fired also at 100 split out frigate chaffs. Result:
0,1,2,3 or 4 destroyed. On rare cases also 5-th destroyed. Other chaffs did not get damage.
Report message to a moderator
|
|
|
|
| Re: Missile hit rate vs additional target tokens |
Tue, 14 March 2006 10:22 |
|
| rowenstin |
 | Crewman 1st Class |
Messages: 38
Registered: December 2005 | |
|
| Kotk wrote on Tue, 14 March 2006 08:04 |
I fired at stack of 1000 frigate chaffs. Result:
0,3,7 or 10 chaffs destroyed and stack gets damage on case of losses of 3 and 7. When stack got more and more damaged during battle then also 4 and 8 started to die.
I fired also at 100 split out frigate chaffs. Result:
0,1,2,3 or 4 destroyed. On rare cases also 5-th destroyed. Other chaffs did not get damage.
|
Apparently, the unstacked chaff behaves as expected, but sounds like the program rounds up chaff very badly damaged up to 100% damaged (and thus destroyed). Why the numbers 3,7,4 and 8, I canīt figure.
[Mod edit: fixed quote]
[Updated on: Wed, 21 June 2006 05:18] by Moderator
Report message to a moderator
|
|
|
|
| Re: Missile hit rate vs additional target tokens |
Tue, 14 March 2006 10:35 |
|
| vonKreedon |
 | Lieutenant |
Messages: 610
Registered: March 2003
Location: Seattle, WA USA | |
|
| Kotk wrote on Tue, 14 March 2006 05:04 |
I fired at stack of 1000 frigate chaffs. Result:
0,3,7 or 10 chaffs destroyed and stack gets damage on case of losses of 3 and 7. When stack got more and more damaged during battle then also 4 and 8 started to die.
I fired also at 100 split out frigate chaffs. Result:
0,1,2,3 or 4 destroyed. On rare cases also 5-th destroyed. Other chaffs did not get damage.
|
This sounds like what I would expect. You fire ten Jihads. The Jihads can destroy at most ten targets, but they are very inaccurate so they tend to hit considerably less than ten. When they hit any target they do more damage than it takes to kill the target and the excess blows through to the stack causing some damage to all non-hit ships in the stack. The only anomalies in this are that there is no blow through when all ten Jihads hit, and that the hit stats change depending on whether or not the targets are in one stack.
Report message to a moderator
|
|
|
|
| Re: Missile hit rate vs additional target tokens |
Tue, 14 March 2006 11:03 |
|
| LEit |
 | Lt. Commander |
Messages: 879
Registered: April 2003
Location: CT | |
|
It seems like the 1 missile 1 kill rule was tacked on after the main damage routine was written. And it is applied after damage is calculated, and is just:
if (# killed > # missiles)
# killed = # missiles
extra damage = 0
So when 1 out of 10 jihads hit, it will do 170 damage (double damage for unshielded), killing 3 FFs and doing 35 extra damage to the stack, 2 of 10 will do 340, killing 7, and doing 25 extra damage. If 3 or more hit, the check for 1 missile 1 kill will kick in and there will be 10 kills and no extra damage.
This is why when you shoot at chaff, you will almost always get # killed == # missiles.
Now, the next question is what happens when you shoot 20 missiles at a stack of 10 chaff, is it always 10 missiles left for the next stack, or could it be fewer if some miss the first stack. My impression is that it will always be 10 left for the next stack.
The other question I've had is, if a slot of missiles on one ship will spread to different stacks, given that the 10 jihads did spread to different stacks, I would guess that one slot on one ship would also spread. Test for this is easy, take a BB with high computing power, and fire at 20 split chaff, If it kills 3 (or 5 for a 20 missile BB) chaff, then they don't split, more then that and they do splits slots.
In any case, it appears that chaff is more effective when split, although the difference is going to be minor after BSCs are out, but against jihad cruisers with 4 computers (accuracy is about 67%), not splitting chaff means you need 50% more (150 chaff to stop 150 missiles when stacked vs 100 chaff needed to stop 150 missiles when split).
If people start doing this, then we'll have battle board overload happening.
- LEitReport message to a moderator
|
|
|
|
| Re: Missile hit rate vs additional target tokens |
Tue, 14 March 2006 12:30 |
|
| mazda |
| Lieutenant |
Messages: 655
Registered: April 2003
Location: Reading, UK | |
|
| Kotk wrote on Mon, 13 March 2006 19:36 |
Basically: 10 uncomputered Jihads against 1000 frigates in stack seem to love to kill none, 3, 7 or 10. I get only these discrete kills, nothing else. Seems that:
if none hit they kill none,
if 1 hits they kill 3,
If 2 hit they kill 7,
if 3 or more hit they kill 10.
Also ... the stack seems to get damage (when less than 10 is killed)! So missing jihads still seem to aid the hitting ones to kill stack.
|
Isn't that exactly what you would expect ?
Each Jihad that hits does 162 damage against unshielded FF.
If 3(or more) hit then damage is capped to exactly 450 (i.e. exactly 10 FF) due to the 1 missile 1 kill rule).
If only 1 missile hits then you get 3 kills with 27 damage left over, and if 2 missiles hit you get 7 kills with 9 damage left over.
Report message to a moderator
|
|
|
| | | | | |
| Re: Missile hit rate vs additional target tokens |
Thu, 16 March 2006 04:45 |
|
| mazda |
| Lieutenant |
Messages: 655
Registered: April 2003
Location: Reading, UK | |
|
Anyway here is an algorithm for Staz.
For each stack of missiles.
For each missile in the stack.
__Fire the missile and calc the damage.
__If total_damage could kill an entire enemy stack then
____if ships_in_enemy_stack > missiles fired at stack then
______do nothing
____end if
____if ships_in_enemy_stack = missiles fired at stack then
______kill ships and set total_damage back to 0
______(but with no excess damage to next stack ?)
______set missiles fired at stack to 0
____end if
____if ships_in_enemy_stack < missiles fired at stack then
______simply kill enemy stack
______total_damage = total_damage - stack (carry forward excess damage)
____end if
__end if
next missile (in stack)
at end of stack apply damage (might have to apply 1 missile 1 kill limit)
next stack of weapons
Comments ?
Errors ??
Omissions ???
[Updated on: Thu, 16 March 2006 05:41] Report message to a moderator
|
|
|
| | |
| Re: Missile hit rate vs additional target tokens |
Fri, 17 March 2006 04:13 |
|
| m.a@stars |
| Commander |
Messages: 2765
Registered: October 2004
Location: Third star to the left | |
|
| LEit wrote on Fri, 17 March 2006 01:53 |
I just had a base with 7 Battle computers, and 48 juggernaut missiles shoot at chaff in single stacks. It killed 48. Jugg accuracy with 7 comps is a bit over 83%, and odds of 48 hits in a row is .000148, which could happen with a huge amount of luck.
Unless it was luck, some how it carries over damage from one stack to the next. Also, splitting chaff does not seem to be an advantage.
|
Hhhmmm. If memory serves, unjammed chaff always gets a 100% (or very very close to 100%) dying ratio. It could be my bad memory, but I cannot remember a single case of at least one unjammed chaff surviving. 
This was indeed one of the things that made me type "accuracy" when I meant "deflection". Indeed, it looks like the "odds" with less-than-perfectly-accurate missiles aren't exactly as advertised when the target has zero jamming.
So many Stars, so few Missiles!
In space no one can hear you scheme! Report message to a moderator
|
|
|
|
| Re: Missile hit rate vs additional target tokens |
Fri, 17 March 2006 06:16 |
|
| mazda |
| Lieutenant |
Messages: 655
Registered: April 2003
Location: Reading, UK | |
|
| LEit wrote on Fri, 17 March 2006 00:53 |
I just had a base with 7 Battle computers, and 48 juggernaut missiles shoot at chaff in single stacks. It killed 48. Jugg accuracy with 7 comps is a bit over 83%, and odds of 48 hits in a row is .000148, which could happen with a huge amount of luck.
Unless it was luck, some how it carries over damage from one stack to the next. Also, splitting chaff does not seem to be an advantage.
|
What if, say, the first missile misses and the second missile hits ?
Could the second missile kill the first and second chaffs ?
Report message to a moderator
|
|
|
| | | |
| Re: Missile hit rate vs additional target tokens |
Sun, 19 March 2006 18:52 |
|
| Kotk |
| Commander |
Messages: 1227
Registered: May 2003 | |
|
| Sr.Seven wrote on Sun, 19 March 2006 23:50 |
Just a thought, but maybe you are approaching this the wrong way.
|
Just a thought, but maybe you understand wrongly what i am testing. 
Large stacks hitting large stacks ... read the header of thread maybe? We discuss how missiles hit lots of tokens, not how they hit large stacks.  | Quote: |
You are trying to deal with streaming damage using chaff (which was not originally envisioned as the usual target for Missiles), and extremely small numbers of them so that statistical oddities and rounding are becoming a problem.
|
I dont care what was envisioned originally, i am interested what the whole math is with split out tokens. Among other things i like to know what happens to chaff. Chaff is one of the most usual things that missiles actually hit, right???
My "extreme tiny" (your words) tests with chaff show that I have shot 10 missiles (with book accuracy 20%) ... 128 times (1280) and these managed to kill 475 chaffs. If it was 200 or 300 then it is fair to call it "statistic oddity". 475 is quite impossible with whatever statistic oddities.
As for bitpacking ... it is done to save game file sizes. Since accuracy is not stored in some turn files i see no reason to bitpack it. Abundant bitpacking makes exe actually longer not shorter. Jeffs seem to use 16 bit and 32 bit integers or floats in math, they dont look like doing math with bitpacked things.
[Updated on: Sun, 19 March 2006 19:39] Report message to a moderator
|
|
|
| | |
| Re: Missile hit rate vs additional target tokens |
Mon, 20 March 2006 05:46 |
|
| mazda |
| Lieutenant |
Messages: 655
Registered: April 2003
Location: Reading, UK | |
|
| LEit wrote on Fri, 17 March 2006 00:53 |
Unless it was luck, some how it carries over damage from one stack to the next. Also, splitting chaff does not seem to be an advantage.
|
To set your mind at rest (and as Kotk also points out) splitting chaff is an advantage.
I used 3 Jihads with 2 BCs (darn close to 50% accuracy).
First I fired at large stacks of chaff.
You get 0 kills or 3 kills (as expected).
You get 0 kills 1/8 of the time (also as expected).
So no problems there.
| Kotk wrote on Sun, 19 March 2006 20:22 |
[Accuracy seems to be better than expected. Jihads accuracy seems to be somewhere in 37%-41% against frigate chaff (instead of 20%).
So something strange goes on and nothing too easy to explain. Doing statistics is not too fun there because cant split out lot of ships. Also split out stuff tends to make battles slow to watch.
|
I also fired the 3 jihads at split up chaff.
First thing that is obvious is that you get less kills when firing at split chaff (but more than what pure odds dictates).
I got 127 kills in 216 missiles (72 rounds).
It was the distibution that baffled (and still does) me.
0 hits - 10, 1 hit - 19, 2 hits - 21, 3 hits 22. (10/19/21/22)
Now if each missile hit was an entirely random event of probability 0.5 and only killed 1 chaff then you'd expect something like 9/27/27/9.
However, the suggestion that the base chances of a hit are increased are not borne out by this. The number of times I got 0 hits is more or less exactly what you would expect for firing 3 missiles with 48% or 49% accuracy.
If you increase the odds of a hit slightly (to 60% say) then you get a shape like 5/20/30/16, where the chances of 0 hits rapidly declines.
It's almost as if that _some_ missiles that hit kill 3 chaff instead of only 1 (or as LEit says, some damage is carried over - I'd say that _sometimes_ some damage is carried over).
I'm going to simplify the test further.
2 Jihads with 1 nexi.
There are only 4 outcomes.
both missiles miss (16%)
missile 1 hits, 2 misses (24%)
missile 2 hits, 1 misses (24%)
both hit (36%)
Hopefully the outcome ratios 0/1/2 will match some sum of those probabilities.
If not then we have a situation where 2 apparently exactly similar events give different results, and I don't like the sound of that.
Report message to a moderator
|
|
|
|
| Re: Missile hit rate vs additional target tokens |
Tue, 08 August 2006 09:10 |
|
| JasonC |
 | Petty Officer 2nd Class
Stars! V.I.P |
Messages: 58
Registered: July 2006
Location: Arlington, MA | |
|
The large FF token example just shows that the "1 missile, 1 kill" maximum rule, refers to all firing missiles, not to all missiles that hit.
The token has 45000 dp.
When 1 jihad hits, the damage is 170. Full ship kills are Floor(170/45) = 3. 35 dp are left to apply to the rest of the token.
When 2 jihads hit, the damage is 340. Full ship kills are Floor(340/45) = 7. 25 dp are left to apply to the rest of the token.
When 3 jihads hit, the damage is 510. Full ship kills are Min ( missiles fire, Floor(510/45) = Min(10, 11) = 10.
The same will be true for any higher number of hits.
There is no damage from missiles that miss. But the one missile one kill rule does not apply to hits only, but to shots. You might say this "feature" is a sign that the code applies the "min(number fired, x)" at the wrong step in the procedure.
Note this means more accurate torpedos don't eat through large stacks of chaff any faster than less accurate capital ship missiles.
Note that accuracy is still real, it determines the number of hits and therefore the base damage, ensures that it happens in integer increments, allows zero hits and thus zero damage etc. It is only the "one missile, one kill" rule that is wacky here. It is one missile = one kill, *not* one ***hit*** = one kill.
Report message to a moderator
|
|
|
| |
Goto Forum:
Current Time: Sat May 04 13:25:19 EDT 2024
|
Members
Search
Help
Register
Login
Home
We know that they used whole byte for missile accuracy values in technology browser. However now you say that these are for looks. For real ones they use another bitpacked mystery table with 3 bits per accuracy. So ... 12 bits per accuracy. 