| Pages: 1 2 3 [4] 5 6 7 :: one page |
|
|
| Author |
Topic |
Blazde
    |
Posted - 2004.08.01 12:31:00 -
[91]
Edited by: Blazde on 01/08/2004 12:32:40
Nice work, very helpful for those of us trying to understand how turrets work  But I'm not at all convinced about this ship - dart board analogy.
At these kind of distances you're basically aiming at the ship itself, (not trying to score hits directly on any particular critical component). So you're probably aiming at the centre of the ship, but you're actually hitting it all over. Up close, you'll hit the centre more, far away you'll hit towards the outside more (or mis entirely). But why should hitting the centre do more damage? Well if the ship is spherical (and you're using projectile weapons) the shot might 'glance off' the edges, doing less kinetic damage perhaps. But most ships aren't spherical, and this modelling wouldn't (as well) for energy/hybrids.
I've always viewed a ship as being a mess of different components, and you might hit any one of them at random. On one shot you might hit a particularly well armoured bit of cosmetic hull. On another you might slip between two armour plates and rupture a fuel line. So then the distribution comes down to the distrubtion of critical components on the ship, and who can say what that is, it probably varies from ship to ship. And does any of this even apply when you're still breaking through the shields? How do shields work? Does anyone know...
In short, I don't think you can say a uniform distribution is wrong anymore than I can say a normal distribution is wrong.
Your second point about the range drop off is more convincing, a sudden anomaly at optimal range, then dropping off but still always scoring wrecking hits at exactly the same amount of damage really is just plain wrong. But not everything can (or should be) 100% realistic. There's gameplay/server load/coding considerations to take into account as well. The way capacitors work is a good example, noone would ever run out of cap if they worked like real life ones. And don't forget we're talking about technology many millenia in the future, you can't know what is really realistic cos it isn't real  .
__________________
4S Corporation Kill List |
Blazde
|
Posted - 2004.08.01 12:31:00 -
[92]
Edited by: Blazde on 01/08/2004 12:32:40
Nice work, very helpful for those of us trying to understand how turrets work But I'm not at all convinced about this ship - dart board analogy.
At these kind of distances you're basically aiming at the ship itself, (not trying to score hits directly on any particular critical component). So you're probably aiming at the centre of the ship, but you're actually hitting it all over. Up close, you'll hit the centre more, far away you'll hit towards the outside more (or mis entirely). But why should hitting the centre do more damage? Well if the ship is spherical (and you're using projectile weapons) the shot might 'glance off' the edges, doing less kinetic damage perhaps. But most ships aren't spherical, and this modelling wouldn't (as well) for energy/hybrids.
I've always viewed a ship as being a mess of different components, and you might hit any one of them at random. On one shot you might hit a particularly well armoured bit of cosmetic hull. On another you might slip between two armour plates and rupture a fuel line. So then the distribution comes down to the distrubtion of critical components on the ship, and who can say what that is, it probably varies from ship to ship. And does any of this even apply when you're still breaking through the shields? How do shields work? Does anyone know...
In short, I don't think you can say a uniform distribution is wrong anymore than I can say a normal distribution is wrong.
Your second point about the range drop off is more convincing, a sudden anomaly at optimal range, then dropping off but still always scoring wrecking hits at exactly the same amount of damage really is just plain wrong. But not everything can (or should be) 100% realistic. There's gameplay/server load/coding considerations to take into account as well. The way capacitors work is a good example, noone would ever run out of cap if they worked like real life ones. And don't forget we're talking about technology many millenia in the future, you can't know what is really realistic cos it isn't real .
__________________
4S Corporation Kill List |
Blazde
|
Posted - 2004.08.01 13:20:00 -
[93]
On a different note, we've been told the wrecking hits happen 1 in 100 shots. I guess it's possible they happen exactly once every 100 shots (exactly 100 shots apart), rather than just on average every 100 shots. In which case the counter might get reset after docking/downtime/changing modules which might explain the 1.33% value.
If it is random, the probability of getting >= 30 wrecking hits, when you expect 22.5 with a sample size of 2250, is 7.3% (Using this tool: http://home.clara.net/sisa/binomial.htm). So not all that unlikely.
And the reason they're often called triple damage hits is that they do triple the nominal damage of the weapon. Which is also the average done on non-wrecking hits. We've been told in some blog somewhere non-wrecking hits do 50% - 150% of the quoted weapon damage, which is confirmed by your data. Which makes wrecking hits 300%, hence the name triple. It's not wrong, it's just a different scale, you could just as easily call a wrecking hit twice maximum damage or six times minimum damage.
__________________
4S Corporation Kill List |
Blazde
|
Posted - 2004.08.01 13:20:00 -
[94]
On a different note, we've been told the wrecking hits happen 1 in 100 shots. I guess it's possible they happen exactly once every 100 shots (exactly 100 shots apart), rather than just on average every 100 shots. In which case the counter might get reset after docking/downtime/changing modules which might explain the 1.33% value.
If it is random, the probability of getting >= 30 wrecking hits, when you expect 22.5 with a sample size of 2250, is 7.3% (Using this tool: http://home.clara.net/sisa/binomial.htm). So not all that unlikely.
And the reason they're often called triple damage hits is that they do triple the nominal damage of the weapon. Which is also the average done on non-wrecking hits. We've been told in some blog somewhere non-wrecking hits do 50% - 150% of the quoted weapon damage, which is confirmed by your data. Which makes wrecking hits 300%, hence the name triple. It's not wrong, it's just a different scale, you could just as easily call a wrecking hit twice maximum damage or six times minimum damage.
__________________
4S Corporation Kill List |
Sewell
|
Posted - 2004.08.01 14:56:00 -
[95]
Very nice work. CCP - please hire TornSoul.
|
Sewell
|
Posted - 2004.08.01 14:56:00 -
[96]
Very nice work. CCP - please hire TornSoul.
|
Khyle
|
Posted - 2004.08.01 17:24:00 -
[97]
I hope you two settle the matter on your own, but id like to lend my support to Estarriol. His comments are spot on.
Furthermore it was not only unfair and hurting the discussion to bring the blowhead link into discussion, its totally misplaces also. Mathematics is just the education to have for this discussion.
My personal opionion regarding the distribution is, if you fire 100 shots during an engagement, its like rolling 100 dice (evenly distributed 6-sided f.e.). Your damage you do during this encounter will resemble a gaussian bell shape, and that is all that counts (total damage during the encounter will win or lose the fight, among other things, not a single shot).
You could of course also make a single shot have some gaussian distribution, but then you get a distribution of total damage done which largely overemphasizes the mean value, beeing very steep and having low mean deviation.
Apart from that, why bother. Evenly distributed single shots get you perfectly acceptable distribution of damage over time, or per encounter, or whatever.
|
Khyle
|
Posted - 2004.08.01 17:24:00 -
[98]
I hope you two settle the matter on your own, but id like to lend my support to Estarriol. His comments are spot on.
Furthermore it was not only unfair and hurting the discussion to bring the blowhead link into discussion, its totally misplaces also. Mathematics is just the education to have for this discussion.
My personal opionion regarding the distribution is, if you fire 100 shots during an engagement, its like rolling 100 dice (evenly distributed 6-sided f.e.). Your damage you do during this encounter will resemble a gaussian bell shape, and that is all that counts (total damage during the encounter will win or lose the fight, among other things, not a single shot).
You could of course also make a single shot have some gaussian distribution, but then you get a distribution of total damage done which largely overemphasizes the mean value, beeing very steep and having low mean deviation.
Apart from that, why bother. Evenly distributed single shots get you perfectly acceptable distribution of damage over time, or per encounter, or whatever.
|
DeFood
|
Posted - 2004.08.01 21:24:00 -
[99]
Edited by: DeFood on 01/08/2004 21:29:49
I *think* what Estoril is saying is he is an idiot. There is no gaussian distribution of hits here, unless Esotrils fantasy world has redefined gausian to look like a flat line.
The point ultimately is that TornSoul expects, not unreasonably, that "Optimal" and "Falloff" model aiming accuracy. Basically its impossible to point a gun straight - so you expect to have some sort of cone, projected out the front of the gun, with the shots landing with a gaussian type distribution in the cone.
Now, if we instead considered this a naval game, then the accuracy curves might make sense:
Imagine a large ship with a gun on it. Its probably "accurate" to within some *coff* reasoable accuracy within some "optimal" range. Some target, within this range has a reasonably high & constant chance of being hit. This is while the gun is relativly flat.
At some point however, instead of pointing the gun relativly AT the target, to get the range the gun has to be pointed more UP. Perhaps even the explosive charge is increased, I dont know. But cross wind comes to play a larger part, and small inaccuracies in the gun multiply more strongly. there is therefore a range in which the accuracy decreases sharply. At the end of the range, if the guns barrel is raised any higher, the shots will start to land closer again. This gives the gun an absolute cutoff at the falloff.
This system seems to be what TomB originally modelled.
|
DeFood
|
Posted - 2004.08.01 21:24:00 -
[100]
Edited by: DeFood on 01/08/2004 21:29:49
I *think* what Estoril is saying is he is an idiot. There is no gaussian distribution of hits here, unless Esotrils fantasy world has redefined gausian to look like a flat line.
The point ultimately is that TornSoul expects, not unreasonably, that "Optimal" and "Falloff" model aiming accuracy. Basically its impossible to point a gun straight - so you expect to have some sort of cone, projected out the front of the gun, with the shots landing with a gaussian type distribution in the cone.
Now, if we instead considered this a naval game, then the accuracy curves might make sense:
Imagine a large ship with a gun on it. Its probably "accurate" to within some *coff* reasoable accuracy within some "optimal" range. Some target, within this range has a reasonably high & constant chance of being hit. This is while the gun is relativly flat.
At some point however, instead of pointing the gun relativly AT the target, to get the range the gun has to be pointed more UP. Perhaps even the explosive charge is increased, I dont know. But cross wind comes to play a larger part, and small inaccuracies in the gun multiply more strongly. there is therefore a range in which the accuracy decreases sharply. At the end of the range, if the guns barrel is raised any higher, the shots will start to land closer again. This gives the gun an absolute cutoff at the falloff.
This system seems to be what TomB originally modelled.
|
|
TornSoul
|
Posted - 2004.08.02 02:04:00 -
[101]
@Blazde
 Quote:
I've always viewed a ship as being a mess of different components, and you might hit any one of them at random. On one shot you might hit a particularly well armoured bit of cosmetic hull. On another you might slip between two armour plates and rupture a fuel line.
Excacly.
But it would be fair to assume that the places where you can 'slip between two armor plates' is (significally) smaller than those where you hit a 'particularly well armoured bit' (else I'd not want to fly that ship...).
Ie. the chance of hitting a critical (high damage inflicted) spot is smaller than hitting the well armoured (for less damage) bit.
It doesnt matter where on the ship it is - it's the overall chance that matters.
I'm in no way implying that you have to hit the center of the ship to achieve high damage.
The analogy to the dart board is used because, the chance to hit the center (high damage) is smaller than hitting one of the outer rings (less damage) - thus 'emulating' this difference in chance of getting high damage vs. low damage shots
Quote:
So then the distribution comes down to the distrubtion of critical components on the ship, and who can say what that is, it probably varies from ship to ship.
Which boils down to having varying thicknesses of the rings on the dartboard (noone says they have to be equally thick - although this is ofc usually implicitly assumed) - and in that way adjusting the probability of getting a specific amount of damage (in effect adjusting the Mean value - But this is stretching the analogy to its limit....)
This you cant do with uniform distribution, because its just that - Uniform.
With uniform distribution you are basically saying that there is an equal amount of 'particularly well armoured bits' and places where you can 'slip between two armor plates'
And thats what I feel is wrong.
(I'd have that engineer shot on the spot... or better yet, have him fly the damn ship himself....)
As has been pointed out - Your total damage over time is not affected either way, but thats not really my beef here at all - and is irrelevant for the case I'm trying to make.
On another note - I'm not suggesting CCP immediatly drops everything and fixes this (theres far more important matters to attend to atm) - The current system *does* work afterall
I'd just like to improve on it is all - and this is my proposal to do just that.
----
@Khyle
I found my link justified after I was just beeing told (in not so many words) I did not understand even basic probabilities.
If you make comments like that, you have to expect to get something back your own way.
Quote:
You could of course also make a single shot have some gaussian distribution, but then you get a distribution of total damage done which largely overemphasizes the mean value, beeing very steep and having low mean deviation.
So are you saying you prefer total damage over time to vary alot (more)?
I for one prefer to know what I can expect from my guns..... Therefore a small SD for the total damage over time suits me just fine.
It's in some sense the same argument for the distribution of the single shots - I'd like to know what I can expect from my guns.
 BIG Lottery
[u |
TornSoul
|
Posted - 2004.08.02 02:04:00 -
[102]
@Blazde
Quote:
I've always viewed a ship as being a mess of different components, and you might hit any one of them at random. On one shot you might hit a particularly well armoured bit of cosmetic hull. On another you might slip between two armour plates and rupture a fuel line.
Excacly.
But it would be fair to assume that the places where you can 'slip between two armor plates' is (significally) smaller than those where you hit a 'particularly well armoured bit' (else I'd not want to fly that ship...).
Ie. the chance of hitting a critical (high damage inflicted) spot is smaller than hitting the well armoured (for less damage) bit.
It doesnt matter where on the ship it is - it's the overall chance that matters.
I'm in no way implying that you have to hit the center of the ship to achieve high damage.
The analogy to the dart board is used because, the chance to hit the center (high damage) is smaller than hitting one of the outer rings (less damage) - thus 'emulating' this difference in chance of getting high damage vs. low damage shots
Quote:
So then the distribution comes down to the distrubtion of critical components on the ship, and who can say what that is, it probably varies from ship to ship.
Which boils down to having varying thicknesses of the rings on the dartboard (noone says they have to be equally thick - although this is ofc usually implicitly assumed) - and in that way adjusting the probability of getting a specific amount of damage (in effect adjusting the Mean value - But this is stretching the analogy to its limit....)
This you cant do with uniform distribution, because its just that - Uniform.
With uniform distribution you are basically saying that there is an equal amount of 'particularly well armoured bits' and places where you can 'slip between two armor plates'
And thats what I feel is wrong.
(I'd have that engineer shot on the spot... or better yet, have him fly the damn ship himself....)
As has been pointed out - Your total damage over time is not affected either way, but thats not really my beef here at all - and is irrelevant for the case I'm trying to make.
On another note - I'm not suggesting CCP immediatly drops everything and fixes this (theres far more important matters to attend to atm) - The current system *does* work afterall
I'd just like to improve on it is all - and this is my proposal to do just that.
----
@Khyle
I found my link justified after I was just beeing told (in not so many words) I did not understand even basic probabilities.
If you make comments like that, you have to expect to get something back your own way.
Quote:
You could of course also make a single shot have some gaussian distribution, but then you get a distribution of total damage done which largely overemphasizes the mean value, beeing very steep and having low mean deviation.
So are you saying you prefer total damage over time to vary alot (more)?
I for one prefer to know what I can expect from my guns..... Therefore a small SD for the total damage over time suits me just fine.
It's in some sense the same argument for the distribution of the single shots - I'd like to know what I can expect from my guns.
BIG Lottery
[u |
Selim
|
Posted - 2004.08.02 02:06:00 -
[103]
Edited by: Selim on 02/08/2004 02:09:57
The best thing for CCP to do right now would to be give TomB a 'vacation' and hire someone who tries to add content instead of taking away options. Every patch for the past few months has been for taking away options instead of adding content. And even though he is trying to balance ships, he doesnt know a damn thing about what he's doing. Guns are screwed up, cruisers are useless, cruiser guns are useless, most battleship guns are, too, and projectiles were hurt the most. Only missiles are good now. People use frigate neutron blasters on cruisers now, because they are better damage over time than even the best cruiser guns.
I dont think they are even listening to this thread...
|
Selim
|
Posted - 2004.08.02 02:06:00 -
[104]
Edited by: Selim on 02/08/2004 02:09:57
The best thing for CCP to do right now would to be give TomB a 'vacation' and hire someone who tries to add content instead of taking away options. Every patch for the past few months has been for taking away options instead of adding content. And even though he is trying to balance ships, he doesnt know a damn thing about what he's doing. Guns are screwed up, cruisers are useless, cruiser guns are useless, most battleship guns are, too, and projectiles were hurt the most. Only missiles are good now. People use frigate neutron blasters on cruisers now, because they are better damage over time than even the best cruiser guns.
I dont think they are even listening to this thread...
|
dalman
|
Posted - 2004.08.02 14:44:00 -
[105]
Originally by: Selim
Edited by: Selim on 02/08/2004 02:09:57
The best thing for CCP to do right now would to be give TomB a 'vacation' and hire someone who tries to add content instead of taking away options. Every patch for the past few months has been for taking away options instead of adding content. And even though he is trying to balance ships, he doesnt know a damn thing about what he's doing. Guns are screwed up, cruisers are useless, cruiser guns are useless, most battleship guns are, too, and projectiles were hurt the most. Only missiles are good now. People use frigate neutron blasters on cruisers now, because they are better damage over time than even the best cruiser guns.
I dont think they are even listening to this thread...
Sigh, Selim, stop it.
Plz tell me ONE battleship gun which is useless. Because there is none. Unlike prior to the patches, where there were some useless guns.
And no, small guns does NOT perform better on cruisers than medium guns, except against frigates if you don't have a webbifier fitted (but without one you'll not hit a frigate anyway).
And CCP are working on cruisers. (like the 2:nd bonus for example).
M.I.A. since 2004-07-30 |
dalman
|
Posted - 2004.08.02 14:44:00 -
[106]
Originally by: Selim
Edited by: Selim on 02/08/2004 02:09:57
The best thing for CCP to do right now would to be give TomB a 'vacation' and hire someone who tries to add content instead of taking away options. Every patch for the past few months has been for taking away options instead of adding content. And even though he is trying to balance ships, he doesnt know a damn thing about what he's doing. Guns are screwed up, cruisers are useless, cruiser guns are useless, most battleship guns are, too, and projectiles were hurt the most. Only missiles are good now. People use frigate neutron blasters on cruisers now, because they are better damage over time than even the best cruiser guns.
I dont think they are even listening to this thread...
Sigh, Selim, stop it.
Plz tell me ONE battleship gun which is useless. Because there is none. Unlike prior to the patches, where there were some useless guns.
And no, small guns does NOT perform better on cruisers than medium guns, except against frigates if you don't have a webbifier fitted (but without one you'll not hit a frigate anyway).
And CCP are working on cruisers. (like the 2:nd bonus for example).
M.I.A. since 2004-07-30 |
Estarriol
|
Posted - 2004.08.02 15:17:00 -
[107]
Originally by: TornSoul
@Khyle
I found my link justified after I was just beeing told (in not so many words) I did not understand even basic probabilities.
If you make comments like that, you have to expect to get something back your own way.
If you start a discussion by tossing out a bunch of statistics, you can't then go back and complain that someone didn't explain the basic terms and rules of probability clearly enough for you.
The craziest thing about this is that you are trying to "reverse-engineer" the turret probabilities in the game by using a relatively small number of hits and then proclaiming with certainty what you think the distribution is (by inspection). All when CCP obviously has the tables/formulas in hand, and TomB already posted graphs weeks or months ago showing the turret hit densities vs tracking.
Estarriol
P. S.: I mentioned my educational background to let you know that using Google search linking to http://www.mathworld.com wasn't going to be necessary to having a discussion. I'm still waiting for your response on technical grounds, either here or in private discussion.
 |
Estarriol
|
Posted - 2004.08.02 15:17:00 -
[108]
Originally by: TornSoul
@Khyle
I found my link justified after I was just beeing told (in not so many words) I did not understand even basic probabilities.
If you make comments like that, you have to expect to get something back your own way.
If you start a discussion by tossing out a bunch of statistics, you can't then go back and complain that someone didn't explain the basic terms and rules of probability clearly enough for you.
The craziest thing about this is that you are trying to "reverse-engineer" the turret probabilities in the game by using a relatively small number of hits and then proclaiming with certainty what you think the distribution is (by inspection). All when CCP obviously has the tables/formulas in hand, and TomB already posted graphs weeks or months ago showing the turret hit densities vs tracking.
Estarriol
P. S.: I mentioned my educational background to let you know that using Google search linking to http://www.mathworld.com wasn't going to be necessary to having a discussion. I'm still waiting for your response on technical grounds, either here or in private discussion.
|
TornSoul
|
Posted - 2004.08.02 16:20:00 -
[109]
Edited by: TornSoul on 02/08/2004 16:22:18
@Estarriol
I dont know why you keep trying to 'pick a fight'...
I *do* have the intention of looking you up on IRC to have a discussion off these forums - But time is such a precious commodity...
But you once again do what you did in you first post...
Quote:
you can't then go back and complain that someone didn't explain the basic terms and rules of probability clearly enough for you.
I did no such thing - Becuase I dont need those explained thank you very much (again you make assumptions about my level of comprehension of statistics - I suggest you proove me wrong in my statements instead!)
What I 'complained' about, was that you put forth mathematical fact that touched on a 'subject' I had not commented on at all.
Becuase of this, I read your comment 'in context' and did not understand the point of the fact presented (indeed you didnt really make one relevant to the case I was presenting) - The mathematical fact itself I understand perfectly.
And you do this once more now
Quote:
TomB already posted graphs weeks or months ago showing the turret hit densities vs tracking.
Where have I discussed tracking? Nowhere - So once again you bring up a fact that has nothing to do with the case im presenting.
Quote:
using a relatively small number of hits and then proclaiming with certainty what you think the distribution is
Two things:
1) Baring an incredible statistical 'fluke' the amount of data I collected is more than sufficient to lend credence to my 'theory'.
2) "proclaiming with certainty' and 'what you think' = oxymoron.
Quote:
linking to http://www.mathworld.com wasn't going to be necessary to having a discussion
That was the easy way of presenting the subject (instead of repeating it all in my post) to wich I asked if that was what you meant or not.
Besides - It might give give others a chance to maybe learn a thing or two if they had the interest.
Why you continue to try and question my ability to understand statistics, but have yet to discredit mathematically any of the points I've presented, is beyond me....
Could we please concentrate on the case at hand instead...
BIG Lottery
[u |
TornSoul
|
Posted - 2004.08.02 16:20:00 -
[110]
Edited by: TornSoul on 02/08/2004 16:22:18
@Estarriol
I dont know why you keep trying to 'pick a fight'...
I *do* have the intention of looking you up on IRC to have a discussion off these forums - But time is such a precious commodity...
But you once again do what you did in you first post...
Quote:
you can't then go back and complain that someone didn't explain the basic terms and rules of probability clearly enough for you.
I did no such thing - Becuase I dont need those explained thank you very much (again you make assumptions about my level of comprehension of statistics - I suggest you proove me wrong in my statements instead!)
What I 'complained' about, was that you put forth mathematical fact that touched on a 'subject' I had not commented on at all.
Becuase of this, I read your comment 'in context' and did not understand the point of the fact presented (indeed you didnt really make one relevant to the case I was presenting) - The mathematical fact itself I understand perfectly.
And you do this once more now
Quote:
TomB already posted graphs weeks or months ago showing the turret hit densities vs tracking.
Where have I discussed tracking? Nowhere - So once again you bring up a fact that has nothing to do with the case im presenting.
Quote:
using a relatively small number of hits and then proclaiming with certainty what you think the distribution is
Two things:
1) Baring an incredible statistical 'fluke' the amount of data I collected is more than sufficient to lend credence to my 'theory'.
2) "proclaiming with certainty' and 'what you think' = oxymoron.
Quote:
linking to http://www.mathworld.com wasn't going to be necessary to having a discussion
That was the easy way of presenting the subject (instead of repeating it all in my post) to wich I asked if that was what you meant or not.
Besides - It might give give others a chance to maybe learn a thing or two if they had the interest.
Why you continue to try and question my ability to understand statistics, but have yet to discredit mathematically any of the points I've presented, is beyond me....
Could we please concentrate on the case at hand instead...
BIG Lottery
[u |
|
JoCool
|
Posted - 2004.08.02 17:10:00 -
[111]
Edited by: JoCool on 02/08/2004 17:18:20
Very good work TornSoul, I would have never thought that the game uses uniform distribution of damage - that's like playing one armed bandit with one wheel - while it should be like playing one with suspension to avoid 'Jackpot' with 3 wheels (you have less chances to get 3 rights than 2 or 1 and less chances to get 3 Jackpots than 3 Apples), like you've pointed out. If you don't know what one armed bandit is I'm sorry, I just know the german term and I'm not familiar with its british/american name.
I hope the Devs implement a system better and less random than the current. This way we'll actually be able to have "reliable" weapons we can count on depending on the circumstances and not luck.
Thanks for working it out
Addendum: Looking back at it, mine is a rather silly example, but I would not simply copy yours. Well at least some more people could understand what you showed us now, probably
|
JoCool
|
Posted - 2004.08.02 17:10:00 -
[112]
Edited by: JoCool on 02/08/2004 17:18:20
Very good work TornSoul, I would have never thought that the game uses uniform distribution of damage - that's like playing one armed bandit with one wheel - while it should be like playing one with suspension to avoid 'Jackpot' with 3 wheels (you have less chances to get 3 rights than 2 or 1 and less chances to get 3 Jackpots than 3 Apples), like you've pointed out. If you don't know what one armed bandit is I'm sorry, I just know the german term and I'm not familiar with its british/american name.
I hope the Devs implement a system better and less random than the current. This way we'll actually be able to have "reliable" weapons we can count on depending on the circumstances and not luck.
Thanks for working it out
Addendum: Looking back at it, mine is a rather silly example, but I would not simply copy yours. Well at least some more people could understand what you showed us now, probably
|
DeFood
|
Posted - 2004.08.02 19:48:00 -
[113]
Flat distribution is what you typically expect in computer games. Most random number generators are designed to have as flat a distribution as possible. Exactly flat if its a cryptographic class prng.
The point being that the simplest way to roll for damage is to call the random number generator once.
Perhaps its lazieness, but its compunded by the fact that it doesnt matter. Gausian or flat, the average damage will be the same. In any combat thats not resolved in just a few rounds, there is no perceptible difference.
|
DeFood
|
Posted - 2004.08.02 19:48:00 -
[114]
Flat distribution is what you typically expect in computer games. Most random number generators are designed to have as flat a distribution as possible. Exactly flat if its a cryptographic class prng.
The point being that the simplest way to roll for damage is to call the random number generator once.
Perhaps its lazieness, but its compunded by the fact that it doesnt matter. Gausian or flat, the average damage will be the same. In any combat thats not resolved in just a few rounds, there is no perceptible difference.
|
TornSoul
|
Posted - 2004.08.02 20:15:00 -
[115]
Ah - Right to the point/core of it. I like it
Flat distribution is what you typically expect in computer games
Yes - But most games dont have the (large) spread in damage that we have with (espcially the large) guns in EVE.
The point being that the simplest way to roll for damage is to call the random number generator once.
Absolutly.
Perhaps its lazieness, but its compunded by the fact that it doesnt matter. Gausian or flat, the average damage will be the same.
Correct - As pointed out a couple of times already.
In any combat thats not resolved in just a few rounds, there is no perceptible difference.
Now this is where I disagree - and why I've bothered with this post in the first place.
It has *always* bugged me that your damage from your guns seems to be 'all over the place'.
That this is noticable at all is ofc because of the large spread on the possible damage - say 125 to 375 (a range of 250!!!).
In most other games (that I've played anyways) - The spread is far far less, and thus a uniform distribution there really doesnt matter.
But in EVE - This is different.
This is ofc my subjective view, but to me its *very perceptible*, and it's always bugged me....
BIG Lottery
[u |
TornSoul
|
Posted - 2004.08.02 20:15:00 -
[116]
Ah - Right to the point/core of it. I like it
Flat distribution is what you typically expect in computer games
Yes - But most games dont have the (large) spread in damage that we have with (espcially the large) guns in EVE.
The point being that the simplest way to roll for damage is to call the random number generator once.
Absolutly.
Perhaps its lazieness, but its compunded by the fact that it doesnt matter. Gausian or flat, the average damage will be the same.
Correct - As pointed out a couple of times already.
In any combat thats not resolved in just a few rounds, there is no perceptible difference.
Now this is where I disagree - and why I've bothered with this post in the first place.
It has *always* bugged me that your damage from your guns seems to be 'all over the place'.
That this is noticable at all is ofc because of the large spread on the possible damage - say 125 to 375 (a range of 250!!!).
In most other games (that I've played anyways) - The spread is far far less, and thus a uniform distribution there really doesnt matter.
But in EVE - This is different.
This is ofc my subjective view, but to me its *very perceptible*, and it's always bugged me....
BIG Lottery
[u |
Estarriol
|
Posted - 2004.08.02 21:22:00 -
[117]
Originally by: TornSoul
I dont know why you keep trying to 'pick a fight'...
I love it how the 'net makes everyone an expert, and anyone else who questions the expert is 'picking a fight.'
Quote:
Why you continue to try and question my ability to understand statistics, but have yet to discredit mathematically any of the points I've presented, is beyond me....
The disagreements I have are very straightforward. As I stated from the beginning, the total damage over time will approach a normal distribution centered around the 'mean' hit even when the individual hits are uniformly distributed (or virtually anything else). What you are asking CCP to do is further complicate the calculation in order to have something that fits your 'feelings' on the subject but has very little effect on the end result (damage) considering the number of shots fired in a combat between equal classes of ships.
If you review your replies in order, you seemed say in turn that:
1. The damage over time is the mean value * number of hits, not a distribution (actually it is a distribution, since there will some variance in the total damage).
2. Summing two dice results in a triangular distribution, not a Gaussian (it is a discrete distribution, not a triangular, and was used as an example for what happens when you begin to sum uniformly distributed events).
3. The the sum of multiple dice is a uniform sum distribution instead of a normal (in fact the uniform sum approaches a Gaussian over time)
4. That you understand the rules of probability and that a degree in maths is irrelevant to a mathematical discussion (fine).
5. You don't care if the result after the battle is Gaussian or not because you want your guns to do some certain thing, shot after shot after shot, or close to it (I can't argue with your feelings, fine).
6. That the plots of damage vs. turret tracking posted by TomB are irrelevant to a discussion about damage distributions (duh).
Your dartboard example was also not factually correct. If you weight a concentric dartboard by scores with the bulleye at the top point value and with decreasing values, there is no guarantee that every lower score will be more likely than every higher score. The scattering of shots may look Gaussian, but the expectation value of the score will depend greatly on the accuracy of the throws (the mean variance from center, in this case). It's quite possible for example that on such a dartboard, numbered from 1-10, that the average score will be above 9 for an excellent thrower (low variance).
Besides which, a real dartboard doesn't work that way. The bullseye is worth the most, but individual 'segments' on the dartboard have widely varying values which are decoupled from their radial distance from the center. This might in fact be a better model for ship damage, in that a ship may have various critical systems scattered about on its structure. There is no telling whether the 'center' of the ship will do more damage on average than hitting its edges, for example. As for what 'realistic' distribution of damage should be used to model this imaginary ship is anybody's guess.
And most of all, you do not seem to understand that there is nothing 'privileged' about the Gaussian distribution. It arises naturally from processes which involve repeated, identical events. There is no reason in the first place to assume that any individual natural event must follow a Gaussian distribution.
Estarriol
P. S.: Please call off your lapdogs. That is, those who have nothing to add except 'I have no understanding of the details of this discussion, but i think TornSoul is a swell guy who does great work so he must be right.'
|
Estarriol
|
Posted - 2004.08.02 21:22:00 -
[118]
Originally by: TornSoul
I dont know why you keep trying to 'pick a fight'...
I love it how the 'net makes everyone an expert, and anyone else who questions the expert is 'picking a fight.'
Quote:
Why you continue to try and question my ability to understand statistics, but have yet to discredit mathematically any of the points I've presented, is beyond me....
The disagreements I have are very straightforward. As I stated from the beginning, the total damage over time will approach a normal distribution centered around the 'mean' hit even when the individual hits are uniformly distributed (or virtually anything else). What you are asking CCP to do is further complicate the calculation in order to have something that fits your 'feelings' on the subject but has very little effect on the end result (damage) considering the number of shots fired in a combat between equal classes of ships.
If you review your replies in order, you seemed say in turn that:
1. The damage over time is the mean value * number of hits, not a distribution (actually it is a distribution, since there will some variance in the total damage).
2. Summing two dice results in a triangular distribution, not a Gaussian (it is a discrete distribution, not a triangular, and was used as an example for what happens when you begin to sum uniformly distributed events).
3. The the sum of multiple dice is a uniform sum distribution instead of a normal (in fact the uniform sum approaches a Gaussian over time)
4. That you understand the rules of probability and that a degree in maths is irrelevant to a mathematical discussion (fine).
5. You don't care if the result after the battle is Gaussian or not because you want your guns to do some certain thing, shot after shot after shot, or close to it (I can't argue with your feelings, fine).
6. That the plots of damage vs. turret tracking posted by TomB are irrelevant to a discussion about damage distributions (duh).
Your dartboard example was also not factually correct. If you weight a concentric dartboard by scores with the bulleye at the top point value and with decreasing values, there is no guarantee that every lower score will be more likely than every higher score. The scattering of shots may look Gaussian, but the expectation value of the score will depend greatly on the accuracy of the throws (the mean variance from center, in this case). It's quite possible for example that on such a dartboard, numbered from 1-10, that the average score will be above 9 for an excellent thrower (low variance).
Besides which, a real dartboard doesn't work that way. The bullseye is worth the most, but individual 'segments' on the dartboard have widely varying values which are decoupled from their radial distance from the center. This might in fact be a better model for ship damage, in that a ship may have various critical systems scattered about on its structure. There is no telling whether the 'center' of the ship will do more damage on average than hitting its edges, for example. As for what 'realistic' distribution of damage should be used to model this imaginary ship is anybody's guess.
And most of all, you do not seem to understand that there is nothing 'privileged' about the Gaussian distribution. It arises naturally from processes which involve repeated, identical events. There is no reason in the first place to assume that any individual natural event must follow a Gaussian distribution.
Estarriol
P. S.: Please call off your lapdogs. That is, those who have nothing to add except 'I have no understanding of the details of this discussion, but i think TornSoul is a swell guy who does great work so he must be right.'
|
DeFood
|
Posted - 2004.08.02 21:56:00 -
[119]
Edited by: DeFood on 02/08/2004 21:57:55
Originally by: Estarriol
The disagreements I have are very straightforward. As I stated from the beginning, the total damage over time will approach a normal distribution centered around the 'mean' hit even when the individual hits are uniformly distributed (or virtually anything else).
Total rubbish. Unless youre using some different meaning to distribution that I (and everyone else I know) use.
According to dictionary.com
distribution A set of numbers and their frequency of occurrence collected from measurements over a statistical population.
The set of numbers we are dealing with are damage amounts. The statistical population is accumulated by shooting something repeatedly.
Now, lets look at what you say: "As I stated from the beginning, the total damage over time will approach a normal distribution centered around the 'mean' hit"
Ok. Im lost. What does that gobeldy **** mean? The "total damage over time" I understand. Its going to approach a single scalar quantity. The average hit in fact.
On any paritcular test the total damage will tend to...
OMG.
really?
Why in the hell are you even arguing that? I mean so what? Thats like saying 1+1=2. Its so damned obvious its not worth bringing up and its certainly pointless.
Given a large number of ships, all which perform a stastically significant number of shots at a target, yes, I'd expect their total damage over time to fall into a normal distribution around the mean.
However that has very little to do with TornSouls argument that the actual distribution of individual damage hits. TornSoul, and your opponents in general, are arguing that the damage distribution, NOT the total damage over time, is, well, flat.
I dont particularly think it matters because, gaussian distribution or flat, the total damage over time, as you so rightly point out, will in either case approach the mean damage, forming a normal distribution.
There. I'm sorry for thinking youre an idiot. I just wish you'd bothered to explain your point just a little more clearly (or perhaps why you thought it important). |
DeFood
|
Posted - 2004.08.02 21:56:00 -
[120]
Edited by: DeFood on 02/08/2004 21:57:55
Originally by: Estarriol
The disagreements I have are very straightforward. As I stated from the beginning, the total damage over time will approach a normal distribution centered around the 'mean' hit even when the individual hits are uniformly distributed (or virtually anything else).
Total rubbish. Unless youre using some different meaning to distribution that I (and everyone else I know) use.
According to dictionary.com
distribution A set of numbers and their frequency of occurrence collected from measurements over a statistical population.
The set of numbers we are dealing with are damage amounts. The statistical population is accumulated by shooting something repeatedly.
Now, lets look at what you say: "As I stated from the beginning, the total damage over time will approach a normal distribution centered around the 'mean' hit"
Ok. Im lost. What does that gobeldy **** mean? The "total damage over time" I understand. Its going to approach a single scalar quantity. The average hit in fact.
On any paritcular test the total damage will tend to...
OMG.
really?
Why in the hell are you even arguing that? I mean so what? Thats like saying 1+1=2. Its so damned obvious its not worth bringing up and its certainly pointless.
Given a large number of ships, all which perform a stastically significant number of shots at a target, yes, I'd expect their total damage over time to fall into a normal distribution around the mean.
However that has very little to do with TornSouls argument that the actual distribution of individual damage hits. TornSoul, and your opponents in general, are arguing that the damage distribution, NOT the total damage over time, is, well, flat.
I dont particularly think it matters because, gaussian distribution or flat, the total damage over time, as you so rightly point out, will in either case approach the mean damage, forming a normal distribution.
There. I'm sorry for thinking youre an idiot. I just wish you'd bothered to explain your point just a little more clearly (or perhaps why you thought it important). |
|
| Pages: 1 2 3 [4] 5 6 7 :: one page |
| First page | Previous page | Next page | Last page |