Forums » General
I've noticed that lot's of VO players have become real physics-whizzes while trying to puzzle out the specifics of this game. I am taking another aspect of the game and setting it to mathematics: PvP victory chances.
However, unlike the physics equations, the following prediction equations are not fact at all. They are simply weighted statistics and probability that come together to form what could be called a fairly accurate set of equations. They are completely estimates, and it is impossible to take into consideration certain aspects such as skill level and shot spray accurately, so they will not always be true. They are simply guideposts from which I plan to make more extensive predictive equations. So once again, these are not going to be even close to exact.
Now, for the equations. I worked these out by taking a simple equation and merely ‘spread it out,’ that is, making each part of it more complex. I'll try to explain the derivations of these to the best of my ability. But it’s just basic algebra and logic, so it shouldn’t be hard. In my equations, I had to use numerous variable to account for the numerous variables that affect a battle. Here's the list of variables I may use (NOTE: Some of these variables are simply parts of each other, and two or more separate variables may combine to form others. You'll see):
A Accuracy: Hits/# of shots
HP Ship Armor
SD Shot Damage
M Ship Maneuvaribility
BC Battery Capacity
BR Battery Recharge
SE Shot Energy
SV Shot Velocity
ST Shot Delay
W Ship Weight
SW Shot Weight
N Number of Hits
S Number of Shots
SES Shot Energy Per Second
SS Number of shots per second
TS Seconds shooting
D Total Damage Inflicted
DS Damage Per Second
L Ship Length
DF Distance Factor
D Distance
PS Pilot Skill (The pilot skill is found with another set of equations I derived, but for the most part, it goes from 1-100)
When the variable names are Upper Case they mean according to your ship or you, and when they are lower case, it means according to your opponent's ship or them. Ok, here it goes:
The most basic equation for winning a battle would be:
0=hp-(SD*N)
(I know the parentheses aren't needed, but it makes it easier to read) This equation is extremely basic- when the opponent's original ship armor minus the number of hits times the hit damage equals 0, they are dead. Basic, simple.
Notice that this is all assuming that you hit them with every shot you take. This is as if they are standing still and you are just shooting at them directly. I will continue with this and will factor in their movement near the end. But first I needed to work out how the equation will predict if you win or not.
I decided that the best way to do this would be for the equation to represent how much time it would take to kill your opponent. That way you could compare the equation from your opponent to you and you to your opponent, and the one with less time would be the one that would win. So in order to make this time-related, I factored in time in the form of shot delay.
0=hp-[(SD*TS)/ST]
This simply says that the amount of damage one inflicts over time is equal to the Shot Damage times the Time Shooting over the Shot Delay. Let's plug in some numbers:
You're shot damage is, say 100 per shot, and you shoot for x seconds (as that is what we're trying to figure out-how long it takes), you're shot delay is .1 seconds, and you're opponent's ship armor is 5,000. That means:
0=5000-(100x/.1)
0=5000-1000x
1000x=5000
x=5 seconds.
It would, at that rate, take 5 seconds to kill your opponent with hitting every shot.
(NOTE: At this point, some of you are probably thinking 'but i'm not constantly shooting in a real battle.' True, but you are shooting for the most part when you have a chance at hitting them. As long as you shoot whenever you have a chance to hit them, this equation still works for time. Basically, for 'Time Shooting' you're just not constantly shooting during that time.)
Now to factor in accuracy. Accuracy (A) is simply a percentage: your number of shots over your number of hits. Since accuracy is a percentage of a whole, it will still go in the numerator of our fraction, but decrease the fraction as it is supposed to. So the basic revised equation would be:
0=hp-[(A*SD*TS)/ST]
All I did was multiply in the accuracy. Let's take this equation for a test ride now. Let's use the same example as last time, and start with the same equation. Let's also say you hit 25% of your shots. The following equations would be:
0=5000-[(100)(x)(.25)]/.1
0=5000-(25x/.1)
0=5000-250x
250x=5000
x=20
It would now take 20 seconds to kill your target. (NOTE: This would have been a lot easier-since you would only hit 1/4 of the time, we could have just multiplied the 5 seconds by four to get 20, but I wanted to prove that the equation still worked)
The equation is still obvious, following a basic train of thought. More importantly, all the variables I have used so far are concrete variables-they can all be found objectively. I haven't yet used a 'skill variable' or something that would change. Also, your shots are intelligently based. This fact may be taken for granted, however, some later equations, it is necessary to regard shots as if they were randomly generated.
I would once again like to acknowledge that I have so far left out many aspects of fighting. However, some of the aspects that one may think I have left out actually are inherently included in, such as the ‘seconds shooting’ aspect that I discussed earlier.
In this next equation, I spread apart accuracy to include multiple aspects. I will spread ‘accuracy’ to include not only your number of hits over your total shots, but also how well you’re enemy’s ship can dodge. In this next equation, I’ll simply change accuracy to # of hits over the # of shots taken.
0=hp-(N*SD*TS)/(ST*S)
Simple, again. And still using all concrete variables, though accuracy is a tough one to find. Now I’ll break off this train of thought for my next equation. In the following equation, the shot percentage (accuracy) is as though your shots were randomly generated all over your screen. In this equation, I take into account the size of the ship you are fighting, the distance you are from that ship, and your total screen size.
<edit>woah! the pi symbols came out really weird, so the little o's with the squigglies over them are actually pi symbols.<edit>
0=hp-(2*L*SD*TS)/(4š*D2*ST)
This equation may look really fancy with the pi and everything, but it’s actually not. This is just another way to factor in accuracy, without having to know in advance how many shots you are going to take and how accurate you will be. The things that were added in included:
4*Pi*D^2- I added in a pi and a 4 and a distance squared. Imagine your ship in space. It is surrounded by a 3-dimensional ‘celestial sphere,’ on which is everything you can see and shoot at is placed. You can only see about 1/8 of this sphere at a time, however, for the purpose of this equation, it doesn’t matter: the shots are randomly generated anyway. The distance to your enemy ship (D) would be the radius of this sphere. And the equation to find the surface area of you ‘celestial sphere’ is 4*pi*r^2. Which is 4*pi*D^2.
2*L- L is the length of your ship. Length indicates linear distance, but in order to find how much space a ship takes up, you need a 2-dimensional surface area figure. In order to find this, I had to find about an average for all ships between their length and surface area. This was difficult, seeing as though a valk has about the same length as a cent, but a much larger surface area. Assuming that a ship is at about a 45 degree angle towards you (which is probably about what it is most of the time) I estimated the average surface area to be the twice the length of the ship in square meters. This is the first estimation I have made so far.
So now we’ve got 2/4 so the equation simplifies to:
0=hp-(L*SD*TS)/(2*š*D2*ST)
Ok, I realize that an equation that assumes that randomly generated shots will be taken is not very useful, but this is just a building block for the next large equation that takes into account player skill. For this equation, I’ll just state it then prove it.
0=hp-(10*L*SD*TS*PS)/(š*D2*ST*ps)
This equation is basically the same as the last one, except it factors in player skill and takes away the concept of randomly generated shots. The player skill factor is, I approximated the same whether the player is shooting or dodging. Basically, I assumed that if a player is extremely accurate, they must be extremely good at dodging too. I know this is not necessarily true, but for the most part it will be about correct. Next, to get rid of the ‘randomly generated shot’ factor, I multiplied the top equation by 40. It may seem like a lot, but here is my reasoning: with a randomly generated shot, it assumes you shoot anywhere in your ‘celestial sphere.’ Now, I assumed that a player will only shoot at least when their opponent is inside they’re field of view, which is 1/8 as small an area as the whole sphere. Then I decided that within one’s field of view, if it were divided into 5 equal sized boxes with the enemy in the middle, a player would shoot into 2 of those boxes. This increases the factor by 2/5, making the numerator of the final equation divided by 2/40. Multiply by the reciprocal and you get 20 times the original numerator. However, there was a 2 in the denominator, so I simplified and got a final multiple of 10. Now the shots are intelligently based.
This is what I have worked out so far. I realize that I have neglected to factor in many, many things such as mass, thrust, multiple weapons, battery usage, etc., but I plan on working out some more equations to factor these in.
My goal of this little project was not to create an entirely accurate predictor of battles (that would be impossible), but to simply better understand the way fighting works, and what affects what. Through this, I have learned many things that help me in terms of weapon and ship choices, including that low delay on weapons is important, and that spraying with energy weapons isn’t very effective.
Thanks for reading all this, and please reply! If you have found some flaws in my algebra, or in my reasoning, please tell me. Also, if you have any suggestions as to other things to factor in, post ‘em. I hope to come up with more stuff.
However, unlike the physics equations, the following prediction equations are not fact at all. They are simply weighted statistics and probability that come together to form what could be called a fairly accurate set of equations. They are completely estimates, and it is impossible to take into consideration certain aspects such as skill level and shot spray accurately, so they will not always be true. They are simply guideposts from which I plan to make more extensive predictive equations. So once again, these are not going to be even close to exact.
Now, for the equations. I worked these out by taking a simple equation and merely ‘spread it out,’ that is, making each part of it more complex. I'll try to explain the derivations of these to the best of my ability. But it’s just basic algebra and logic, so it shouldn’t be hard. In my equations, I had to use numerous variable to account for the numerous variables that affect a battle. Here's the list of variables I may use (NOTE: Some of these variables are simply parts of each other, and two or more separate variables may combine to form others. You'll see):
A Accuracy: Hits/# of shots
HP Ship Armor
SD Shot Damage
M Ship Maneuvaribility
BC Battery Capacity
BR Battery Recharge
SE Shot Energy
SV Shot Velocity
ST Shot Delay
W Ship Weight
SW Shot Weight
N Number of Hits
S Number of Shots
SES Shot Energy Per Second
SS Number of shots per second
TS Seconds shooting
D Total Damage Inflicted
DS Damage Per Second
L Ship Length
DF Distance Factor
D Distance
PS Pilot Skill (The pilot skill is found with another set of equations I derived, but for the most part, it goes from 1-100)
When the variable names are Upper Case they mean according to your ship or you, and when they are lower case, it means according to your opponent's ship or them. Ok, here it goes:
The most basic equation for winning a battle would be:
0=hp-(SD*N)
(I know the parentheses aren't needed, but it makes it easier to read) This equation is extremely basic- when the opponent's original ship armor minus the number of hits times the hit damage equals 0, they are dead. Basic, simple.
Notice that this is all assuming that you hit them with every shot you take. This is as if they are standing still and you are just shooting at them directly. I will continue with this and will factor in their movement near the end. But first I needed to work out how the equation will predict if you win or not.
I decided that the best way to do this would be for the equation to represent how much time it would take to kill your opponent. That way you could compare the equation from your opponent to you and you to your opponent, and the one with less time would be the one that would win. So in order to make this time-related, I factored in time in the form of shot delay.
0=hp-[(SD*TS)/ST]
This simply says that the amount of damage one inflicts over time is equal to the Shot Damage times the Time Shooting over the Shot Delay. Let's plug in some numbers:
You're shot damage is, say 100 per shot, and you shoot for x seconds (as that is what we're trying to figure out-how long it takes), you're shot delay is .1 seconds, and you're opponent's ship armor is 5,000. That means:
0=5000-(100x/.1)
0=5000-1000x
1000x=5000
x=5 seconds.
It would, at that rate, take 5 seconds to kill your opponent with hitting every shot.
(NOTE: At this point, some of you are probably thinking 'but i'm not constantly shooting in a real battle.' True, but you are shooting for the most part when you have a chance at hitting them. As long as you shoot whenever you have a chance to hit them, this equation still works for time. Basically, for 'Time Shooting' you're just not constantly shooting during that time.)
Now to factor in accuracy. Accuracy (A) is simply a percentage: your number of shots over your number of hits. Since accuracy is a percentage of a whole, it will still go in the numerator of our fraction, but decrease the fraction as it is supposed to. So the basic revised equation would be:
0=hp-[(A*SD*TS)/ST]
All I did was multiply in the accuracy. Let's take this equation for a test ride now. Let's use the same example as last time, and start with the same equation. Let's also say you hit 25% of your shots. The following equations would be:
0=5000-[(100)(x)(.25)]/.1
0=5000-(25x/.1)
0=5000-250x
250x=5000
x=20
It would now take 20 seconds to kill your target. (NOTE: This would have been a lot easier-since you would only hit 1/4 of the time, we could have just multiplied the 5 seconds by four to get 20, but I wanted to prove that the equation still worked)
The equation is still obvious, following a basic train of thought. More importantly, all the variables I have used so far are concrete variables-they can all be found objectively. I haven't yet used a 'skill variable' or something that would change. Also, your shots are intelligently based. This fact may be taken for granted, however, some later equations, it is necessary to regard shots as if they were randomly generated.
I would once again like to acknowledge that I have so far left out many aspects of fighting. However, some of the aspects that one may think I have left out actually are inherently included in, such as the ‘seconds shooting’ aspect that I discussed earlier.
In this next equation, I spread apart accuracy to include multiple aspects. I will spread ‘accuracy’ to include not only your number of hits over your total shots, but also how well you’re enemy’s ship can dodge. In this next equation, I’ll simply change accuracy to # of hits over the # of shots taken.
0=hp-(N*SD*TS)/(ST*S)
Simple, again. And still using all concrete variables, though accuracy is a tough one to find. Now I’ll break off this train of thought for my next equation. In the following equation, the shot percentage (accuracy) is as though your shots were randomly generated all over your screen. In this equation, I take into account the size of the ship you are fighting, the distance you are from that ship, and your total screen size.
<edit>woah! the pi symbols came out really weird, so the little o's with the squigglies over them are actually pi symbols.<edit>
0=hp-(2*L*SD*TS)/(4š*D2*ST)
This equation may look really fancy with the pi and everything, but it’s actually not. This is just another way to factor in accuracy, without having to know in advance how many shots you are going to take and how accurate you will be. The things that were added in included:
4*Pi*D^2- I added in a pi and a 4 and a distance squared. Imagine your ship in space. It is surrounded by a 3-dimensional ‘celestial sphere,’ on which is everything you can see and shoot at is placed. You can only see about 1/8 of this sphere at a time, however, for the purpose of this equation, it doesn’t matter: the shots are randomly generated anyway. The distance to your enemy ship (D) would be the radius of this sphere. And the equation to find the surface area of you ‘celestial sphere’ is 4*pi*r^2. Which is 4*pi*D^2.
2*L- L is the length of your ship. Length indicates linear distance, but in order to find how much space a ship takes up, you need a 2-dimensional surface area figure. In order to find this, I had to find about an average for all ships between their length and surface area. This was difficult, seeing as though a valk has about the same length as a cent, but a much larger surface area. Assuming that a ship is at about a 45 degree angle towards you (which is probably about what it is most of the time) I estimated the average surface area to be the twice the length of the ship in square meters. This is the first estimation I have made so far.
So now we’ve got 2/4 so the equation simplifies to:
0=hp-(L*SD*TS)/(2*š*D2*ST)
Ok, I realize that an equation that assumes that randomly generated shots will be taken is not very useful, but this is just a building block for the next large equation that takes into account player skill. For this equation, I’ll just state it then prove it.
0=hp-(10*L*SD*TS*PS)/(š*D2*ST*ps)
This equation is basically the same as the last one, except it factors in player skill and takes away the concept of randomly generated shots. The player skill factor is, I approximated the same whether the player is shooting or dodging. Basically, I assumed that if a player is extremely accurate, they must be extremely good at dodging too. I know this is not necessarily true, but for the most part it will be about correct. Next, to get rid of the ‘randomly generated shot’ factor, I multiplied the top equation by 40. It may seem like a lot, but here is my reasoning: with a randomly generated shot, it assumes you shoot anywhere in your ‘celestial sphere.’ Now, I assumed that a player will only shoot at least when their opponent is inside they’re field of view, which is 1/8 as small an area as the whole sphere. Then I decided that within one’s field of view, if it were divided into 5 equal sized boxes with the enemy in the middle, a player would shoot into 2 of those boxes. This increases the factor by 2/5, making the numerator of the final equation divided by 2/40. Multiply by the reciprocal and you get 20 times the original numerator. However, there was a 2 in the denominator, so I simplified and got a final multiple of 10. Now the shots are intelligently based.
This is what I have worked out so far. I realize that I have neglected to factor in many, many things such as mass, thrust, multiple weapons, battery usage, etc., but I plan on working out some more equations to factor these in.
My goal of this little project was not to create an entirely accurate predictor of battles (that would be impossible), but to simply better understand the way fighting works, and what affects what. Through this, I have learned many things that help me in terms of weapon and ship choices, including that low delay on weapons is important, and that spraying with energy weapons isn’t very effective.
Thanks for reading all this, and please reply! If you have found some flaws in my algebra, or in my reasoning, please tell me. Also, if you have any suggestions as to other things to factor in, post ‘em. I hope to come up with more stuff.
hey Celebrim, are you reading this? :)
tboyz, can you code? you may want to try and run some scripts or simulations based on these equations and see what sorts of results you get. i previously tried my hand at this sort of modeling (though with some simplifying assumptions) and found the results quite surprising.
tboyz, can you code? you may want to try and run some scripts or simulations based on these equations and see what sorts of results you get. i previously tried my hand at this sort of modeling (though with some simplifying assumptions) and found the results quite surprising.
no, i can't code...but if someone could help me run scripts or simulations...PLEASE DO!!! =p
This hurts my head...
/me runs off crying
Nice work though. :p
/me runs off crying
Nice work though. :p
Nice effort. :) I think you are mixing assumptions and statistics too much and might have better results with purely statistical approach. Gathering those statistics might be difficult.
or rather, you can take a serious of tests to find out your pilot skill factor... except... you don't know anyone else's skill factor... hmm.
Perhaps it would have to be like with rankings for Go used to be in the 17th century. You build from the top down, find the guy no one can beat, (Eldrad/Niki I think has dibs currently), then work your way down.
Perhaps it would have to be like with rankings for Go used to be in the 17th century. You build from the top down, find the guy no one can beat, (Eldrad/Niki I think has dibs currently), then work your way down.
In other news, I've developed a formula that will determine your intellectual ability and potential earnings for the rest of your life! More at 12!
Shape- Good idea with the rankings. However, that would require a large-scale competition and the rankings would change often. I actually have the pilot skill number of about 30 players. I used the responses from my "Vendetta Community Survey" in the community projects section to find these. (In those responses, you're at the top =p, though you might have some competition if Eldrad or Niki responded).
So what you have now is an equation that will tell you how quickly you can kill another pilot.
Running some quick tests now..
You have two "D"s as variables. Might want to chekc that out. Im assuming the D in the final equation is distance.
Ok.
Assuming you are both using CentIIIs and a single neut3, fighting at a distance of 150m (seems standard), with pilot skill assumed to be equal, it would take about 3 hours for the fight to end.
Something is wrong either in how I did the equation or you still need to factor a lot more stuff in ;)
Here we go:
0=6700hp - (10m * 600dps * xs * 50)/(pi * 150m^2 * .14s * 50)
Your units don't end up balanced either, which is an indication of a problem. You get something like 0 = hp - (damage/meter * second). HP would have to be in terms of damage/meter seconds for that to work in physics.
Running some quick tests now..
You have two "D"s as variables. Might want to chekc that out. Im assuming the D in the final equation is distance.
Ok.
Assuming you are both using CentIIIs and a single neut3, fighting at a distance of 150m (seems standard), with pilot skill assumed to be equal, it would take about 3 hours for the fight to end.
Something is wrong either in how I did the equation or you still need to factor a lot more stuff in ;)
Here we go:
0=6700hp - (10m * 600dps * xs * 50)/(pi * 150m^2 * .14s * 50)
Your units don't end up balanced either, which is an indication of a problem. You get something like 0 = hp - (damage/meter * second). HP would have to be in terms of damage/meter seconds for that to work in physics.
I came up with a couple different approaches to this. My aim was to find arbitrary 'goodness' numbers for each weapon by scanning through thousands of 'kills' in the database I've collected. Each kill is obviously affected by the pilot's skill and the weapons/ship/etc he's using. So I performed gradient descent to find all these things in a massive system of equations.
It sorta worked but the results weren't all that great - it mainly had to do with my extremely simplistic model of the pilot-ship-weapons system. It occasionally found local minima where, for instance, the mining beam was considered a good weapon, because I just added up numbers for whatever it was they were carrying. And that was another problem - it's hard to solve this without simply finding lots of silly local minima.
I've also applied the ELO rating system ( http://en.wikipedia.org/wiki/ELO_rating_system -- used for chess, go, and a bunch of other games) to the duel statistics database. That's also pretty interesting, but I didn't include any weapon handicaps or anything. John came up with the idea of separating duelers into classes depending on what license levels they have, and then listing ratings within each class (plus an 'open' class which includes everybody). For this I'm of the opinion that simpler is probably better.
There are only a couple of free parameters in the governing equations (K and '400' in the Wikipedia article) and I was able to tweak it to give only a very small reward to one particular player who created a dummy newbie character and killed him over and over. If I tried to actually penalize him, it screwed with everyone else's stats.
I've been meaning to publish these results and add a section to the website with duel ranks, but we need to patch the duel code on the server to update the ranks dynamically and I hadn't had an opportunity to do so yet. I will soon™.
Back to the original post, I urge you to convert to a form where you can find a probability that player A will kill player B given your predictor parameters. That's what I came up with, and that's how the ELO system works (I later found out). Then you can just work to minimize the probabilistic error between your predicted results and the actual results.
A more accurate prediction equation in this form would definitely make the weapon balance equations a lot more accurate. I've wanted to do some work in this direction as well. Please come up with a form that handles multiple weapons. :P Maybe you can just factor out everything that isn't pilot skill and sum up the weapons.
Unfortunately, I'm not collecting statistics on number of hitpoints of damage and how many shots were fired and how long the battle took. That assumes duel conditions and wouldn't work in the case of prolonged battle among several people vying, for instance, for xithricite cargo. So I just record who-killed-whom and carrying what (including, for instance, xithricite).
It sorta worked but the results weren't all that great - it mainly had to do with my extremely simplistic model of the pilot-ship-weapons system. It occasionally found local minima where, for instance, the mining beam was considered a good weapon, because I just added up numbers for whatever it was they were carrying. And that was another problem - it's hard to solve this without simply finding lots of silly local minima.
I've also applied the ELO rating system ( http://en.wikipedia.org/wiki/ELO_rating_system -- used for chess, go, and a bunch of other games) to the duel statistics database. That's also pretty interesting, but I didn't include any weapon handicaps or anything. John came up with the idea of separating duelers into classes depending on what license levels they have, and then listing ratings within each class (plus an 'open' class which includes everybody). For this I'm of the opinion that simpler is probably better.
There are only a couple of free parameters in the governing equations (K and '400' in the Wikipedia article) and I was able to tweak it to give only a very small reward to one particular player who created a dummy newbie character and killed him over and over. If I tried to actually penalize him, it screwed with everyone else's stats.
I've been meaning to publish these results and add a section to the website with duel ranks, but we need to patch the duel code on the server to update the ranks dynamically and I hadn't had an opportunity to do so yet. I will soon™.
Back to the original post, I urge you to convert to a form where you can find a probability that player A will kill player B given your predictor parameters. That's what I came up with, and that's how the ELO system works (I later found out). Then you can just work to minimize the probabilistic error between your predicted results and the actual results.
A more accurate prediction equation in this form would definitely make the weapon balance equations a lot more accurate. I've wanted to do some work in this direction as well. Please come up with a form that handles multiple weapons. :P Maybe you can just factor out everything that isn't pilot skill and sum up the weapons.
Unfortunately, I'm not collecting statistics on number of hitpoints of damage and how many shots were fired and how long the battle took. That assumes duel conditions and wouldn't work in the case of prolonged battle among several people vying, for instance, for xithricite cargo. So I just record who-killed-whom and carrying what (including, for instance, xithricite).
me gets a major headache.
eccc... add my grey matter to the spreading stain on the far wall... #-@
(oh, and that was supposed to be a smiley, lol)
(oh, and that was supposed to be a smiley, lol)
Mining beam a good weapon .... /me rofl
Actually, they don't use the ELO ratings for go, except on Yahoo.
The Go ranking system is all based around the amount of handicap you can give another player. If you're 1 kyu and you can give another player 8 stones consistently, then they're considered to be at 9 kyu. likewise a 3 dan gives a 1 kyu 3 stones, etc...
Doubt this could be applied to ratings here though. There's a fair amount of room for people to make mistakes in the duels here.
EDIT: a1k0n: isn't separating players into classes by license levels going to prefer the usage of certain weps. Wouldn't grouping players by pk's make more sense, or by pks/deaths? you get an idea for the people who are successful with those weps.
Cuz right now there really isn't a big correlation between kills and levels. Once people hit level 6 they usually have most of the weps they need to cause havoc.
The Go ranking system is all based around the amount of handicap you can give another player. If you're 1 kyu and you can give another player 8 stones consistently, then they're considered to be at 9 kyu. likewise a 3 dan gives a 1 kyu 3 stones, etc...
Doubt this could be applied to ratings here though. There's a fair amount of room for people to make mistakes in the duels here.
EDIT: a1k0n: isn't separating players into classes by license levels going to prefer the usage of certain weps. Wouldn't grouping players by pk's make more sense, or by pks/deaths? you get an idea for the people who are successful with those weps.
Cuz right now there really isn't a big correlation between kills and levels. Once people hit level 6 they usually have most of the weps they need to cause havoc.
Shape, yes, each class will probably end up with its favored weapons. That's what usually happens in any classed competition.
There doesn't need to be any correlation between kills/weapons. The point is you'll have skilled people in each class. There's no "beginner" class - we're not classing based on skill, just equipment available to the player.
There doesn't need to be any correlation between kills/weapons. The point is you'll have skilled people in each class. There's no "beginner" class - we're not classing based on skill, just equipment available to the player.
This reminds me of that show NUMB3RS! Yay!
@Starfisher- "Assuming you are both using CentIIIs and a single neut3, fighting at a distance of 150m (seems standard), with pilot skill assumed to be equal, it would take about 3 hours for the fight to end"
Have you ever fought a player thats the exact same skill level as you in a CentIII each of you with a single neut at a distance of 150m? It would take a veeeerrry long time. However, you are correct that I end up with different units-i have to fix that. But the equation seems to give logical results anyway...hmm...
@Shape+a1k0n- The thing about grouping players, and just generally ranking players and especially weapons, is that fighting styles come greatly into play. Some players, who can't manage to get more than 25 player kills, may be able to beat a vet who has +500 pk's simply because the n00b's fighting style exploits the vet's weaknesses. Would the n00b then get a large boost in their ranking, and the vet's drop?
Also, when it comes to effectiveness of weapons, ships and fighting styles also have a huge affect. And neut III, a great all-around weapon for most people, is practically useless against a prom or another large ship, when it would tear a wraith apart. However, to some players, it might be just the opposite. The variance of fighting styles is one of the greatest parts of PvP in this game, but makes it hard to predict or rank anything.
Pings also throw things off. As one post on this forum discussed, a ship such as a centurion with a 100 ping is almost unhittable. And a player with +150 ping can't shoot for beans. The devs have done a great job minimizing the affects of a slow connection, but in an online game, it's unavoidable.
Finally, binds and tricks throw the whole thing off. A player with an uber auto-dodge bind may easily kill one without the bind. Also, a player who knows little tricks (such as assigning a seperate mouse button to each S weapon port in a 2 S weapon port ship in order to "rapid fire") will have that extra advantage when maybe their true skill level would be less.
Despite these hinderences, I also still believe there's a way to rank players. If the 'group' idea was tweaked a little to also include things such as a player's ping, it may work out well. I'll begin to look for a way to find the affectiveness of weapons and what is important to make weapons fair. I'll post my findings in another string.
Have you ever fought a player thats the exact same skill level as you in a CentIII each of you with a single neut at a distance of 150m? It would take a veeeerrry long time. However, you are correct that I end up with different units-i have to fix that. But the equation seems to give logical results anyway...hmm...
@Shape+a1k0n- The thing about grouping players, and just generally ranking players and especially weapons, is that fighting styles come greatly into play. Some players, who can't manage to get more than 25 player kills, may be able to beat a vet who has +500 pk's simply because the n00b's fighting style exploits the vet's weaknesses. Would the n00b then get a large boost in their ranking, and the vet's drop?
Also, when it comes to effectiveness of weapons, ships and fighting styles also have a huge affect. And neut III, a great all-around weapon for most people, is practically useless against a prom or another large ship, when it would tear a wraith apart. However, to some players, it might be just the opposite. The variance of fighting styles is one of the greatest parts of PvP in this game, but makes it hard to predict or rank anything.
Pings also throw things off. As one post on this forum discussed, a ship such as a centurion with a 100 ping is almost unhittable. And a player with +150 ping can't shoot for beans. The devs have done a great job minimizing the affects of a slow connection, but in an online game, it's unavoidable.
Finally, binds and tricks throw the whole thing off. A player with an uber auto-dodge bind may easily kill one without the bind. Also, a player who knows little tricks (such as assigning a seperate mouse button to each S weapon port in a 2 S weapon port ship in order to "rapid fire") will have that extra advantage when maybe their true skill level would be less.
Despite these hinderences, I also still believe there's a way to rank players. If the 'group' idea was tweaked a little to also include things such as a player's ping, it may work out well. I'll begin to look for a way to find the affectiveness of weapons and what is important to make weapons fair. I'll post my findings in another string.
tboyz007 - Yes, actually. Well, maybe not exactly equal skill - but we were so close as to make the distinction either way impossible. took about fifteen minutes. I lost. I guess I'd technically be the lesser party.
You're equation may work in cases of highly disparate skill levels, but it still needs a lot of factors - shots per battery, for instance, would be a good next addition.
Actually I think this will always fail for equal skill levels and ships, so my example sucked :P.
You're equation may work in cases of highly disparate skill levels, but it still needs a lot of factors - shots per battery, for instance, would be a good next addition.
Actually I think this will always fail for equal skill levels and ships, so my example sucked :P.
As I'm running through simple tests of this equation on my graphing calculator (the only handy tool that can do this stuff quickly) I'm realizing that my equation is even more basic and obvious than I thought.
First, Starfisher- two things. 1. you're right, it shouldn't take two hours to finish off a duel ever. 2. It doesn't according to my equation. I ran your situation through my calculator, and it came up with 1105.6 seconds, which is roughly equal to 18 minutes (hehe, pretty darn close to you're real life example where it took 15 minutes). I might be wrong on my calculation however, so I'll check it again. Anyone else wanna test this out.
Next, I was looking at the equation itself, and realized what basic principles it implied. Note that the following conclusions are based entirely on the real facts of the game, as long as all other factors are equal (including pilot skill). Basically, it means that under completely equal circumstances, they'd be true.
1. As distance to the enemy ship increases, the length of time of the battle increases exponentially.
2. Shot delay and the length of the battle are directly related.
3. Ship length and length of battle are inversly related.
4. Shot damage and length of battle are inversly related.
Noticing a pattern? All the factors on the numerator are (obviously) inversly related to the length of the battle. All the factors on the denominator are (obviously) directly related to the length of the battle. I didn't need to run any tests to find that out, it's just simple fractions. But it does give some nice (but obvious) insight into the game. Anyway, I'm beginning to work on a formula to judge weapon efficiency. However, for spray weapons such as the gatlings and flechette, i need some description of how they spray. Anyone? a1k0n? Thanks.
First, Starfisher- two things. 1. you're right, it shouldn't take two hours to finish off a duel ever. 2. It doesn't according to my equation. I ran your situation through my calculator, and it came up with 1105.6 seconds, which is roughly equal to 18 minutes (hehe, pretty darn close to you're real life example where it took 15 minutes). I might be wrong on my calculation however, so I'll check it again. Anyone else wanna test this out.
Next, I was looking at the equation itself, and realized what basic principles it implied. Note that the following conclusions are based entirely on the real facts of the game, as long as all other factors are equal (including pilot skill). Basically, it means that under completely equal circumstances, they'd be true.
1. As distance to the enemy ship increases, the length of time of the battle increases exponentially.
2. Shot delay and the length of the battle are directly related.
3. Ship length and length of battle are inversly related.
4. Shot damage and length of battle are inversly related.
Noticing a pattern? All the factors on the numerator are (obviously) inversly related to the length of the battle. All the factors on the denominator are (obviously) directly related to the length of the battle. I didn't need to run any tests to find that out, it's just simple fractions. But it does give some nice (but obvious) insight into the game. Anyway, I'm beginning to work on a formula to judge weapon efficiency. However, for spray weapons such as the gatlings and flechette, i need some description of how they spray. Anyone? a1k0n? Thanks.
Two randomly-chosen angles, call them theta and phi. theta is -180 to 180 and phi is 0 to whatever cone angle the weapon uses. The forward vector of the shot is then computed with two simple [[cos -sin][sin cos]] rotation matrices (or one, multiplied out).
You probably want to come up with a probability density function with the weapon cone angle as a parameter. Thing is, the 'spray' effect can actually increase the probability of a hit of someone who is dodging optimally, so the approximation will be rough.
Edit: actually, looking at the code, it doesn't use two angles, and though it could be considered equivalent, the probability density of what I described is probably different from what it actually does: it picks a random unit vector, then multiplies that by a random number from 0 to 1, then multiplies that by the 'inaccuracy' factor (which is 0.05 for the large port gatling turret), then perturbs the quaternion defining the laser's orientation by that vector. So it has more of a uniform distribution of randomness than what I originally implied. It also can 'roll' the laser around the ship's forward axis, but that won't really affect anything.
You probably want to come up with a probability density function with the weapon cone angle as a parameter. Thing is, the 'spray' effect can actually increase the probability of a hit of someone who is dodging optimally, so the approximation will be rough.
Edit: actually, looking at the code, it doesn't use two angles, and though it could be considered equivalent, the probability density of what I described is probably different from what it actually does: it picks a random unit vector, then multiplies that by a random number from 0 to 1, then multiplies that by the 'inaccuracy' factor (which is 0.05 for the large port gatling turret), then perturbs the quaternion defining the laser's orientation by that vector. So it has more of a uniform distribution of randomness than what I originally implied. It also can 'roll' the laser around the ship's forward axis, but that won't really affect anything.