| Pages: 1 [2] 3 4 5 6 7 :: one page |
| Author |
Thread Statistics | Show CCP posts - 0 post(s) |
Chainsaw Plankton
IDLE GUNS
IDLE EMPIRE
   |
Posted - 2009.01.16 21:05:00 -
[31]
that and it Might make sense to look at 1 racial and 2 off racial (or maybe 2 racial 1 off racial) on one target and 1 racial and one off racial on another target. as well that is probably what the falcon will have.
and yes 1 falcon can butt hurt 2 ships pretty good.
|
daisy dook
|
Posted - 2009.01.16 21:11:00 -
[32]
Nice objective numbers that give an indication of how effective ECCM is.
It is a shame you had to simplify your model and not use a real eve example of the jamming bird fitting one each racial.
|
chrisss0r
|
Posted - 2009.01.16 21:15:00 -
[33]
 Originally by: daisy dook
Nice objective numbers that give an indication of how effective ECCM is.
It is a shame you had to simplify your model and not use a real eve example of the jamming bird fitting one each racial.
i doubt it would change much since you gain in having the right racials fitted by chance and lose from not having the right ones. Of course it's not linear but i think the estimation is solid to work with unless someone brings up a more precise model.
Mine for sure is much more precise than the standard one :P
|
Etho Demerzel
Gallente
Holy Clan of the Cone
|
Posted - 2009.01.16 21:25:00 -
[34]
Originally by: chrisss0r
that calculation is not complete since it does not factor in the probability that one bs is jammed while the other is not. that is somewhat compensated by factoring relock time in by an estimate.
my estimate is that the break even for 2 bs vs 1 bs + 1 falcon is around 1 bs having 1 and 1 bs having 2 eccm fitted.
Now you lost me. 
I did account for the effect of locking time in both battleships.
As for missing the times one battleship will be jammed and the other won't, you are correct and I indeed mislooked it in my previous analysis (of the non ECCMed battleships).
In this analysis, on the other hand, I used the individual chances of one cycle as calculated by you, adjusted them to account for the impact of relocking times and used them as the average jam time. It is correct as far as I can tell, but if it is not and if you can point where it is wrong exactly and what are the correct numbers, please do it.
=====
"If a member of the EVE community finds he or she cannot accept our current level of transparency, we bid you good luck in finding a company that meets your needs." - CCP kieron... |
chrisss0r
|
Posted - 2009.01.16 21:28:00 -
[35]
Edited by: chrisss0r on 16/01/2009 21:29:57
46% is the base chance of having both targets jammed
yeah i saw you accounted for the locktime :) maybe misread my post? :)
giving the right number is quite tricky since you have to again factor in alot of stuff.
|
daisy dook
|
Posted - 2009.01.16 21:31:00 -
[36]
Originally by: chrisss0r
Edited by: chrisss0r on 16/01/2009 21:21:01
Originally by: daisy dook
Nice objective numbers that give an indication of how effective ECCM is.
It is a shame you had to simplify your model and not use a real eve example of the jamming bird fitting one each racial.
i doubt it would change much since you gain in having the right racials fitted by chance and lose from not having the right ones. Of course it's not linear but i think the estimation is solid to work with unless someone brings up a more precise model.
Mine for sure is much more precise than the standard one :P
i'm looking forward to see liliths java programm though.
That depends upon the assumption whether you will have the correct racial available for each of the two ships being jammed; it also assumes the Falcon/Rook pilot has 6 jammers.
Personally I always fly with a sensor booster (how else can I jam from suitable range) and mwd (how else do I have a chance when bubbled) which would leave 5 jammers.
|
chrisss0r
|
Posted - 2009.01.16 21:35:00 -
[37]
Originally by: daisy dook
Originally by: chrisss0r
Edited by: chrisss0r on 16/01/2009 21:21:01
Originally by: daisy dook
Nice objective numbers that give an indication of how effective ECCM is.
It is a shame you had to simplify your model and not use a real eve example of the jamming bird fitting one each racial.
i doubt it would change much since you gain in having the right racials fitted by chance and lose from not having the right ones. Of course it's not linear but i think the estimation is solid to work with unless someone brings up a more precise model.
Mine for sure is much more precise than the standard one :P
i'm looking forward to see liliths java programm though.
That depends upon the assumption whether you will have the correct racial available for each of the two ships being jammed; it also assumes the Falcon/Rook pilot has 6 jammers.
Personally I always fly with a sensor booster (how else can I jam from suitable range) and mwd (how else do I have a chance when bubbled) which would leave 5 jammers.
Sure. The model is obviously very simple but i don't feel very urged to put the time for more complex calcs into it.
CCPs economist should have the empirical abilities to deliver a somewhat complete model. And he on top of that get's paid. Maybe they should give me a free faction bs or so
|
Camilo Cienfuegos
Earned In Blood
|
Posted - 2009.01.16 21:36:00 -
[38]
Quote:
Maybe they should give me a free faction bs or so
A free ECCM module, surely... 
|
Derek Boyd
|
Posted - 2009.01.16 21:54:00 -
[39]
I am by no means an expert on this topic. But (1-0.5)^2 would imho only be the probability of jamming both ships if the two events (succesful jam for ship a or b) are independent. I don't think they are as f.e. a sucessfull jam with the first jammer on ship 1 will leave 5 jammers for ship 2 and as such alter the probability of ship 2 getting jammed. Thus the multiplication rule for independent events cannot be applied as far as i understand.
I think you need to do the correct math first otherwise your simulation is meaningless. But maybe i just suck at maths. :) |
Camilo Cienfuegos
Earned In Blood
|
Posted - 2009.01.16 21:55:00 -
[40]
The simulation is based on 3 jammers each time per ship, not however many jammers are currently free. |
chrisss0r
|
Posted - 2009.01.16 21:56:00 -
[41]
Edited by: chrisss0r on 16/01/2009 21:56:35
Originally by: Derek Boyd
I am by no means an expert on this topic. But (1-0.5)^2 would imho only be the probability of jamming both ships if the two events (succesful jam for ship a or b) are independent. I don't think they are as f.e. a sucessfull jam with the first jammer on ship 1 will leave 5 jammers for ship 2 and as such alter the probability of ship 2 getting jammed. Thus the multiplication rule for independent events cannot be applied as far as i understand.
I think you need to do the correct math first otherwise your simulation is meaningless. But maybe i just suck at maths. :)
What if i told you that was EXACTLY what i did?
Originally by: Camilo Cienfuegos
The simulation is based on 3 jammers each time per ship, not however many jammers are currently free.
vice versa. the simulation is based on how many jammers are free while the static calculus is based on 3 jammers fix on each ship |
Camilo Cienfuegos
Earned In Blood
|
Posted - 2009.01.16 22:00:00 -
[42]
Originally by: chrisss0r
vice versa. the simulation is based on how many jammers are free while the static calculus is based on 3 jammers fix on each ship
Your code is fixed to 6 jammer over two ships just as it finds them:
Quote:
# first flip. 0 is heads, 1 means tails
c1 = self.flip()
c2 = self.flip()
# loop for remaining throws
amount_throws = 0
while amount_throws < 4:
if c1 == 0:
c1 = self.flip()
amount_throws += 1
if c2 == 0:
c2 = self.flip()
amount_throws += 1
|
chrisss0r
|
Posted - 2009.01.16 22:02:00 -
[43]
Edited by: chrisss0r on 16/01/2009 22:03:45
Originally by: Camilo Cienfuegos
Originally by: chrisss0r
vice versa. the simulation is based on how many jammers are free while the static calculus is based on 3 jammers fix on each ship
Your code is fixed to 6 jammer over two ships just as it finds them:
nope
Look at the code again. if c1 is already heads c2 will continue flipping till it's heads aswell or all jammers are used up
The initial flip uses up 2 jammers.
4 are free and distributed to the coins beeing flipped. if both coins are tails, both coins get flipped gain. if not only one get's flipped again and again.
3 jammers each is the simple calc giving the lower numbers since not accounting that jamchance on ship 1 oand 2 are not independant |
Camilo Cienfuegos
Earned In Blood
|
Posted - 2009.01.16 22:04:00 -
[44]
Quote:
4 are free and distributed to the coins beeing flipped. if both coins are tails, both coins get flipped gain. if not only one get's flipped again and again
One word:
d'oh!  |
Derek Boyd
|
Posted - 2009.01.16 22:19:00 -
[45]
Originally by: chrisss0r
What if i told you that was EXACTLY what i did?
I would probably read your post again and realize that i can only blame me being tired. 
|
Lilith Velkor
Minmatar
DEATH'S LEGION
Red Box.
|
Posted - 2009.01.16 22:44:00 -
[46]
Edited by: Lilith Velkor on 16/01/2009 22:49:42
So here are some actual numbers regarding locktime experiments. Granted the scenarios I provide arent the most realistic but you'll be able to check any other you like, see below.
Jam model is as follows:
0) the chance to jam formula (jamstr/sensorstr)=jamchange is considered accurate
1) target locks have to be acquired
2) jammer knows about target lock capability immediately
3) as much jammers are applied as needed to take one ship out, then next ship *
4) fulltime lockdown is jamming goal **
5) ecm strength is from t2 racial+multispec on a lvl5 eft falcon with 3x SDA2
* I know this isnt perfect, but see scenarios below, not an issue, will be fixed
** not perfect either, there can be extra time wasted for target by giving 1-2 seconds delay before hitting cyle, can be fixed too if you insist
Software used to generate data [this'll be a link soon, how do I put it on eve-files please?]
So here goes, 120 seconds simulation time, 5 runs each (numbers are time locked in seconds, rounded to full seconds). Edit: this is the time the target ships would have had a lock established:
5x gravimetric / 5x caldari ship (22 sensor / 5 locktime), all @10k samples
0 5 23 61 99
0 5 24 61 100
0 5 23 61 100
0 5 24 62 100
0 5 24 62 100
5x multispec / 5x caldari ship (22 sensor / 5 locktime)
5 28 66 99 112
5 28 66 99 113
5 28 66 99 112
5 28 66 99 113
5 28 67 99 113
5x gravimetric / 5x caldari ship (44 sensor / 5 locktime)
13 49 89 109 114
13 49 89 109 114
13 49 89 109 114
13 49 89 109 114
13 49 88 109 114
5x multispec / 5x caldari ship (44 sensor / 5 locktime)
29 76 105 113 114
29 76 105 113 114
29 76 105 113 114
29 76 105 113 114
29 76 105 113 114
5x gravimetric / 5x caldari ship (66 sensor / 10 locktime)
22 67 98 108 109
22 67 99 108 109
22 67 99 108 109
22 67 99 108 109
22 67 99 108 109
5x multispec / 5x caldari ship (66 sensor / 10 locktime)
38 86 106 109 109
38 86 106 109 109
38 86 106 109 109
38 86 106 109 109
38 86 106 109 109
Dont know what that does tell me tbh, realistically 5 of the proper racial is a bit extreme to encounter, so considered a worst-case prolly.
|
Zhilia Mann
|
Posted - 2009.01.16 22:50:00 -
[47]
Originally by: Lilith Velkor
5) ecm strength is from t2 racial+multispec on a lvl5 eft falcon with 3x SDA2
Very minor nitpick, but that's only 2x SDA2s, right?
Otherwise, yes, this is getting to be interesting.
|
Lilith Velkor
Minmatar
DEATH'S LEGION
Red Box.
|
Posted - 2009.01.16 22:53:00 -
[48]
Originally by: Zhilia Mann
Originally by: Lilith Velkor
5) ecm strength is from t2 racial+multispec on a lvl5 eft falcon with 3x SDA2
Very minor nitpick, but that's only 2x SDA2s, right?
Otherwise, yes, this is getting to be interesting.
My falcon alt uses 3x SDA2 in lows, so I was thinking that'd do
|
chrisss0r
|
Posted - 2009.01.16 22:54:00 -
[49]
i don't quite get what these numbers try to tell me? ;D
|
Lilith Velkor
Minmatar
DEATH'S LEGION
Red Box.
|
Posted - 2009.01.16 23:04:00 -
[50]
Originally by: chrisss0r
i don't quite get what these numbers try to tell me? ;D
Heh, my wording is poor
It is basically an estimate how many seconds from 120 seconds a ship in a fleet would have a target locked if a certain number of jammers is applied to the fleet.
cycling of the jammers for best performance is applied, as well as time to relock and the initial lock at start of period.
averaged over 10000 runs, repeated 5 times each.
format
ship1 ship2 ship3 ship4 ship5
|
chrisss0r
|
Posted - 2009.01.16 23:11:00 -
[51]
Edited by: chrisss0r on 16/01/2009 23:12:48
Ahhhh alot clearer now.
thx and very nice work.
hm how about displaying it in percent of 120 seconds. would make the firepower calcs alot easyer
|
Karentaki
Gallente
Oberon Incorporated
Morsus Mihi
|
Posted - 2009.01.16 23:15:00 -
[52]
The probability of jamming x ships given y jammers and z probability per jammer is actually a binomial distribution:
P(X >= x)~B(y,z)
Essentially this comes down to:
1 - [sum r=0 to x-1](yCr (1-z)^(y-r) z^r)
for x=2, y=6, and z=0.5 this equation gives 0.8906 = 89.06%
That is almost exactly what the model predicted, and also surprisingly high compared to what I expected.
Quote:
EVE is like a sandbox with landmines. Deal with it.
|
Etho Demerzel
Gallente
Holy Clan of the Cone
|
Posted - 2009.01.16 23:56:00 -
[53]
Originally by: Karentaki
The probability of jamming x ships given y jammers and z probability per jammer is actually a binomial distribution:
P(X >= x)~B(y,z)
Essentially this comes down to:
1 - [sum r=0 to x-1](yCr (1-z)^(y-r) z^r)
for x=2, y=6, and z=0.5 this equation gives 0.8906 = 89.06%
That is almost exactly what the model predicted, and also surprisingly high compared to what I expected.
The higher the probability is, the less difference it makes to know the result of the previous jammers, meaning the closer the binomial model approaches to chrisss0r's model.
|
Number 86
Eat Ship and Die
|
Posted - 2009.01.17 00:03:00 -
[54]
Do you guys think activations in varied intervals would change effective jam durations?
ex.- 6 jammers each with 15 sec activation duration, but rather than activating them all simultaneously space the activations out every 3 seconds. So you would end up with a chance to jam every 3 seconds rather than 6 chances every 15 seconds. |
chrisss0r
|
Posted - 2009.01.17 00:04:00 -
[55]
Edited by: chrisss0r on 17/01/2009 00:06:46
Edited by: chrisss0r on 17/01/2009 00:04:25
probability per jammer is actually a binomial distribution:
P(X >= x)~B(y,z)
Essentially this comes down to:
1 - [sum r=0 to x-1](yCr (1-z)^(y-r) z^r)
for x=2, y=6, and z=0.5 this equation gives 0.8906 = 89.06%
That is almost exactly what the model predicted, and also surprisingly high compared to what I expected.
Not exactly.
The chance you calculated by the binomial formula is correct but you use it wrong. the formula gives you the probabilty to have 2 jammers working out of 8. It usually is correct if it does not matter which ones are the working ones but in case of jammers it matters.
The reason why it solves for the right probability is that our staging aproach kinda sorts the jammers so that there can't be 2 working ones on the same ship.
It's a coincidence of wrong aproach, right result here. Usually the formula yoused would be wrong, together with the staging game it get's right
Though very very nice since i would not have dreamed of the binomical-distribution solving for an exact probability here
Also this coincidence suddenly disapears if we let the module get a tad more complex. like relock time or player reaction time to apply the second and third stage of jammers. Bayes is the broader approach and the correct one |
chrisss0r
|
Posted - 2009.01.17 00:07:00 -
[56]
Originally by: Number 86
Do you guys think activations in varied intervals would change effective jam durations?
ex.- 6 jammers each with 15 sec activation duration, but rather than activating them all simultaneously space the activations out every 3 seconds. So you would end up with a chance to jam every 3 seconds rather than 6 chances every 15 seconds.
it sure would but that's a very *****y thing to calc |
Etho Demerzel
Gallente
Holy Clan of the Cone
|
Posted - 2009.01.17 00:09:00 -
[57]
Originally by: Number 86
Do you guys think activations in varied intervals would change effective jam durations?
ex.- 6 jammers each with 15 sec activation duration, but rather than activating them all simultaneously space the activations out every 3 seconds. So you would end up with a chance to jam every 3 seconds rather than 6 chances every 15 seconds.
That may help you if your target's lock time is big and your chances to jam if you apply all modules at once are not very high. Otherwise I would stick with the OP method. |
Lilith Velkor
Minmatar
DEATH'S LEGION
Red Box.
|
Posted - 2009.01.17 09:01:00 -
[58]
Edited by: Lilith Velkor on 17/01/2009 09:02:28
Overlapping of leftover jammers does indeed increase the average time jammed, but it seems that a good number of excess jammers as well as a good base chance per jammer is needed to make the effect strong.
Download link to the simulator, currently hardcoded to handle 10ships/15jammers maximum. A lot of ad popup **** on the site, but eve-files didnt work for some reason 
--------------------------
4 ships, one each race, sensor strength 28.4, locktime 7 seconds, simulation starts without target lock, 2000 samples per run, 120 seconds interval, numbers are seconds with established target lock on average, jammer strength are as before t2 multi/racial on eft lvl5 falcon with 3x SDA2 and no rigs
--------------------------
2x each racial, 2x multi, no overlap
6.9725 9.818 11.3775 12.894
7.091 9.2345 11.262 13.526
7.1435 9.4645 11.9685 13.363
6.802 9.5215 11.4725 13.1815
7.108 9.648 11.47 13.305
2x each racial, 2x multi, with overlap
5.3465 7.6725 8.9085 9.906
5.3285 7.4955 8.9485 10.5085
5.3585 7.1185 8.815 9.9455
5.315 7.2395 8.566 10.313
5.4835 7.206 8.6655 9.86
3x each racial, 3x multi, no overlap
1.515 1.8865 2.3945 3.0575
1.554 2.131 2.557 2.733
1.444 1.962 2.539 3.1415
1.61 1.9315 2.4625 3.006
1.578 2.092 2.5085 2.9225
3x each racial, 3x multi, with overlap
0.941 1.2345 1.37 1.6185
0.7345 1.1355 1.393 1.5675
0.855 1.145 1.2085 1.5575
0.7785 1.1995 1.3675 1.562
0.6975 1.0935 1.324 1.45
---------------------------------
For a eccm comparision, look at our 4 ships with 56.8 sensor strength each:
2x each racial, 2x multi, no overlap
34.41 43.2 48.056 51.466
34.2845 44.3015 49.243 52.1555
33.458 43.502 49.817 51.7925
33.8695 43.974 49.2765 52.0495
33.6685 44.014 48.662 51.984
2x each racial, 2x multi, with overlap
32.65 42.6455 48.336 51.413
32.468 43.589 48.6995 50.499
33.4025 42.909 48.2585 52.708
33.113 42.5515 47.9135 51.4705
33.395 42.3785 48.3135 51.2365
3x each racial, 3x multi, no overlap
17.7435 23.492 29.142 31.7475
17.3845 23.6805 27.454 31.4895
18.35 25.0235 29.2575 31.4715
17.4745 24.207 28.184 32.0995
18.119 23.923 26.932 32.9165
3x each racial, 3x multi, with overlap
16.4625 22.3495 27.1235 29.4475
16.248 21.8745 26.069 29.265
16.006 22.2285 26.504 30.1775
16.001 21.5625 26.385 29.231
16.0005 22.4685 26.1365 29.7915
|
chrisss0r
|
Posted - 2009.01.17 11:16:00 -
[59]
Edited by: chrisss0r on 17/01/2009 11:17:27
Very nice little programm.
Well our little experiment by now brought up the following conclusions:
1. and most important:
Bayesian probability calc is not something fancy i made up to look smart while it has no impact on jamming
2.
Ecm on a few targets works alot more effective than people usually believe. For 2 targets and 6 jammers it's 88-89% instead of ra 76%
3. This does nothing to eccm. While the base chances of getting jammed are higher than commonly believed, this has no impact on how ECCM works.
2 ships with eccm fitted have half the base probability of beeing locked down, which results in permajamming probabilities falling rapidly over a few cycles (44% to be permajammed through 6 cycles vs 0.09% with eccm fitted (both targets permajammed))
Conclusion:
Falcons in small gang situations are alot stronger than people think, while eccm also works alot better than people think.
What to make out of it? I don't know but i hope the accurate numbers will help base the discussion on fact's closer to the eve reality
|
Lilith Velkor
Minmatar
DEATH'S LEGION
Red Box.
|
Posted - 2009.01.17 12:37:00 -
[60]
Edited by: Lilith Velkor on 17/01/2009 12:48:27
Edited by: Lilith Velkor on 17/01/2009 12:45:30
I have to admit the impact is bigger than I did estimate.
I'd like to point out the importance of eccm on BC hulls and up again, basically if you are 1on1 against a typical falcon fit it makes the difference of being able to shoot at all or not.
For a gang situation, 8 BS vs 3 falcons does look pretty bad without eccm, but with one eccm per BS we can get enough time to shoot 3 falcons down realistically.
Important aspect here is the hardening of key vessels, i.e. you want your snipers be able to have as much unjammed time as possible, so extra eccm + remote eccm should pay off very well on such ships, as well as maybe on a counter-jamming scorpion, you get the idea.
One final note about the 'permajamming', consider the typical falcon fit with 2x caldari / 1x each other racial which is very common, and we might have caldari players in general having more permajamming issues as other races, and caldari population being biggest could lead to a certain bias in perception regarding permajamming issues. |
| |
|
| Pages: 1 [2] 3 4 5 6 7 :: one page |
| First page | Previous page | Next page | Last page |