Hello, I think you should change the elosystem. i think its pretty annoying that if u lose against players with low elo and yourself has a pretty high elo you lose like -28. and this is a problem caus people have so many akas. so if example.. i want to drag down a players elo then i go make a new acc and then waits for him to join and hopfully beat him, thne he gets like -20.
my suggestion is that nomatter how high or low elo u have u always win or lose the same ammount of elo.. for example
a win gets 10
a lose gets -5
or whatever. i hope you concider it caus its so annoying when people abuse the elo system to benefit themself
Elo system
Moderator: Oversight Staff
-
nabo.
- Donator
- Posts: 11892
- Joined: Mon Jun 10, 2013 4:30 am
- Location: Dokdo, KOREA
- Has thanked: 44 times
- Been thanked: 158 times
Re: Elo system
First, I dont see why you should be winning more elo compared to when u lose... Win 10 elo, Lose 10 elo seems more reasonable.
Second, I dont see the problem with how the elo system is. Elo system is fine. Smurfing/sniping is the issue since it misrepresent stats. Numerous suggestions have been posted, but there havent been a really good one yet. If you stack: you have lower risk lower return. If you go against stack: you have higher risk, but higher return. Simple.
Third, If you good player you will have a high winning percentage, meaning you'll gain more elo than losing elo on avg. If you r losing more elo often, simply means you not so good as you think. Unless your elo skews the avg elo in game because your elo is either TOO high or TOO low, I dont see problem with elo system.
Second, I dont see the problem with how the elo system is. Elo system is fine. Smurfing/sniping is the issue since it misrepresent stats. Numerous suggestions have been posted, but there havent been a really good one yet. If you stack: you have lower risk lower return. If you go against stack: you have higher risk, but higher return. Simple.
Third, If you good player you will have a high winning percentage, meaning you'll gain more elo than losing elo on avg. If you r losing more elo often, simply means you not so good as you think. Unless your elo skews the avg elo in game because your elo is either TOO high or TOO low, I dont see problem with elo system.
-
Metall-Drago
- Treant
- Posts: 440
- Joined: Sat Sep 07, 2013 1:52 pm
Re: Elo system
You cannot fix smurfing unless forcing ppl to bind their wc3 acc to forum acc - which will reduce the player base drastically.
Removing the elo-system completely would be the only solution to this.
Removing the elo-system completely would be the only solution to this.
-
supersexyy
- Donator
- Posts: 3484
- Joined: Sat Sep 01, 2012 9:26 pm
- Has thanked: 12 times
- Been thanked: 39 times
Re: Elo system
Removing the elo system is a bad idea. It works good in theory, but in practice it will reduce the competitiveness - which makes gamers game.
-
greenmtn
- Forest Walker
- Posts: 126
- Joined: Sun May 19, 2013 7:07 pm
- Has thanked: 14 times
- Been thanked: 5 times
Re: Elo system
http://www.dota-void.com/score.php
The [void] Scoring System
What the score actually measures
The overall score attempts to quantify how much, on average, a player contributes to the outcome of his or her games. Unfortunately, DotA is a complex game, and there are innumerable ways for a player to contribute that can't be measured based on hero and creep stats alone. However, this method of scoring does attempt to reasonably judge player performance as a whole based on stats that can be measured. DotA is a team game, and because of this it's only logical to calculate a score that reflects team contribution.
How the [void] system differs from the ELO system
The [void] system differs from the ELO system in two main respects. First, the ELO system is based on wins and losses; the [void] system doesn't consider wins and losses. In a single player game (Chess, for example), wins and losses accurately describe a player's performance; however, wins and losses only vaguely describe a player's contribution to a team. Consider the player who has a weak team and yet still does well; he or she shouldn't be penalized by the performance of the rest of the team. And in a similar manner, the player who does poorly on a good team shouldn't be rewarded simply for having a good team.
The second major difference between the ELO system and the [void] system is the units used to measure the score. The ELO system uses an arbitrary base score (1000, for example), and players will either move up or down - and arbitrary amount - from this score. On the other hand, the [void] system is based on zero, and players will move up or down based on standard deviations. In fact, if you were to sum every single player's overall score on this server, you would get zero (by definition, the average player). For every strong (positive) player, there is an equally weak (negative) set of players - and vice versa.
Why the [void] system is better suited for DotA
One major drawback to having an arbitrary scoring system, like the ELO system, is score inflation. As more scores are added to the system, the average score will change; the base score will no longer be an accurate indication of average skill. With the [void] system though, the average player will always have a score of zero.
Another drawback of the ELO system is the difficulty of comparing two scores relative to one another. As an example, a player with a score of 1500 is not necessarily the opposite of a player with a score of 500. But in the [void] system, a player at +1, for example, is exactly opposite in skill level of a player at -1.
Because the [void] system has these strengths, it is a better method of quantifying a player's skill level than a system like ELO.
How the score is calculated
The first step in calculating a player's score is to calculate a raw score for each player at the end of every game. Kills, deaths, and assists are the main contributing factors to the raw score, but creep stats are also accounted for. At the end of a game, each player is assigned a raw score based on the aforementioned stats. Next, both the average and standard deviation are calculated with all ten raw scores. Once this is done, each player is assigned a game-specific score which is the number of standard deviations he or she deviated from the mean. Finally, a player's overall score is simply the average of all his or her game scores.
The formula was posted by Undefined; admin of dota-void: http://www.dota-void.com/forum/index.ph ... 16.30.html
There seems to be some confusion on the score calculation - maybe I can help clear it up. First, read this. Now, here's how the score is calculated.
Score is based on kills, deaths, assists, and creep stats (not neutrals though). I'll explain why these particular values are used later on, but for now here's what each event adds to your raw score.
kill: +1.000
death: -0.952
assist: +0.800
creepkill: +0.040
creepdeny: +0.100
So to get a raw score then is easy. You just multiply a player's k/d/a/cs by those constants and return whatever number comes out:
Code: [Select]
RETURN (((kills / 1.0) + (deaths / -1.05) + (assists / 1.25)) / 1.0) + (((creepkills / 12.5) + (creepdenies / 5.0)) / 2.0);
At the end of a game then, you've got 10 raw scores. They're ugly: unbounded, not normalized, not very meaningful between different games. So the solution is to find some basic statistical properties of this set of scores. First, find the mean and standard deviation of these 10 scores. (Add 'em all up, divide by 10 - that's the mean. Take the square root of the sum of the squared differences of each raw score from the mean - that's the standard deviation. The mean and standard deviation are very useful in many contexts, and so the database can calculate these values by simply calling "avg()" and "std()".)
Ok now you've got 10 raw scores, a mean, and a standard deviation. The DV score is defined as the number of standard deviations that the raw score is from the mean. (In other words, it's the plain old standard score.) This standard score is a player's score in that particular game.
Getting a player's overall score is wicked easy - just take the average of all the individual game scores.
Hopefully I've convinced you that the formula works. Now let's look at why the min/max scores are -3/+3, respectively. To do this, let's consider a very lame game. Blue kills one creep and the game ends. So blue's raw score is +0.040, the other 9 players have a score of exactly 0. (Do you see why? Work out the formula above.)
Next we need the summary statistics. Mean = +0.040 / 10 = +0.004. Standard deviation = 0.012. Now, get the standard score for each player: (score - mean) / stddev
Blue: (0.040 - 0.004) / 0.012 = +3.000
Everyone else: (0.000 - 0.004) / 0.012 = -0.333
Strange huh? The magical number 3 arises as a result of there being 10 players in the game. (Image a version of DotA with 12 players. The guy with a single creep kill still has a raw score of +0.040, the mean is now 0.00333, and the stddev is now 0.01106. So in this trippy 12-man dota, the max score is now 3.316.)
There's one last sanity check that we can do with these 10 standardized scores. Add them all together, and you should get exactly zero. (+3 -.333 * 9) = 0. It's not just in this silly example either. For any game, the sum of all 10 scores will add to zero. Which proves that, as stated in the score page I linked above, the sum of all player scores server-wide including all games is zero. And hence, the average player is, by definition, assigned a score of zero.
About those magic k/d/a/cs multipliers. Way back in the beginning of DV, there was no score. I had the formula, but I needed the constants. To get them, I took the player database that I had at the time (which was something like 3,000 games - pretty small, but not too small), and I got averages for all k/d/a/cs. Essentially I asked "on average, how many kills does a player get for each death?"; and "how many kills for each assist?"; and so on. In other words, what's the average k/k ratio, k/-d ratio, k/a ratio, k/ck ratio, and k/cd ratio. These ratios (rounding a bit) are the source of the magic numbers. k/k = 1, k/-d = -0.952, and k/a = 0.8. As for creep stats, there's one extra step: divide their score contribution by four. (My reasoning was that creep k/d stats aren't quite as important as hero k/d/a stats.) So, k/ck/4 = 0.04 and k/cd/4 = 0.1.
Of course, these ratios are constantly fluctuating as new data gets added to the database. To try to keep things fair though, I've held the numbers constant since day 1. So if you played here 2 years ago, your score was calculated the same way as is today - no shenanigans.
The [void] Scoring System
What the score actually measures
The overall score attempts to quantify how much, on average, a player contributes to the outcome of his or her games. Unfortunately, DotA is a complex game, and there are innumerable ways for a player to contribute that can't be measured based on hero and creep stats alone. However, this method of scoring does attempt to reasonably judge player performance as a whole based on stats that can be measured. DotA is a team game, and because of this it's only logical to calculate a score that reflects team contribution.
How the [void] system differs from the ELO system
The [void] system differs from the ELO system in two main respects. First, the ELO system is based on wins and losses; the [void] system doesn't consider wins and losses. In a single player game (Chess, for example), wins and losses accurately describe a player's performance; however, wins and losses only vaguely describe a player's contribution to a team. Consider the player who has a weak team and yet still does well; he or she shouldn't be penalized by the performance of the rest of the team. And in a similar manner, the player who does poorly on a good team shouldn't be rewarded simply for having a good team.
The second major difference between the ELO system and the [void] system is the units used to measure the score. The ELO system uses an arbitrary base score (1000, for example), and players will either move up or down - and arbitrary amount - from this score. On the other hand, the [void] system is based on zero, and players will move up or down based on standard deviations. In fact, if you were to sum every single player's overall score on this server, you would get zero (by definition, the average player). For every strong (positive) player, there is an equally weak (negative) set of players - and vice versa.
Why the [void] system is better suited for DotA
One major drawback to having an arbitrary scoring system, like the ELO system, is score inflation. As more scores are added to the system, the average score will change; the base score will no longer be an accurate indication of average skill. With the [void] system though, the average player will always have a score of zero.
Another drawback of the ELO system is the difficulty of comparing two scores relative to one another. As an example, a player with a score of 1500 is not necessarily the opposite of a player with a score of 500. But in the [void] system, a player at +1, for example, is exactly opposite in skill level of a player at -1.
Because the [void] system has these strengths, it is a better method of quantifying a player's skill level than a system like ELO.
How the score is calculated
The first step in calculating a player's score is to calculate a raw score for each player at the end of every game. Kills, deaths, and assists are the main contributing factors to the raw score, but creep stats are also accounted for. At the end of a game, each player is assigned a raw score based on the aforementioned stats. Next, both the average and standard deviation are calculated with all ten raw scores. Once this is done, each player is assigned a game-specific score which is the number of standard deviations he or she deviated from the mean. Finally, a player's overall score is simply the average of all his or her game scores.
The formula was posted by Undefined; admin of dota-void: http://www.dota-void.com/forum/index.ph ... 16.30.html
There seems to be some confusion on the score calculation - maybe I can help clear it up. First, read this. Now, here's how the score is calculated.
Score is based on kills, deaths, assists, and creep stats (not neutrals though). I'll explain why these particular values are used later on, but for now here's what each event adds to your raw score.
kill: +1.000
death: -0.952
assist: +0.800
creepkill: +0.040
creepdeny: +0.100
So to get a raw score then is easy. You just multiply a player's k/d/a/cs by those constants and return whatever number comes out:
Code: [Select]
RETURN (((kills / 1.0) + (deaths / -1.05) + (assists / 1.25)) / 1.0) + (((creepkills / 12.5) + (creepdenies / 5.0)) / 2.0);
At the end of a game then, you've got 10 raw scores. They're ugly: unbounded, not normalized, not very meaningful between different games. So the solution is to find some basic statistical properties of this set of scores. First, find the mean and standard deviation of these 10 scores. (Add 'em all up, divide by 10 - that's the mean. Take the square root of the sum of the squared differences of each raw score from the mean - that's the standard deviation. The mean and standard deviation are very useful in many contexts, and so the database can calculate these values by simply calling "avg()" and "std()".)
Ok now you've got 10 raw scores, a mean, and a standard deviation. The DV score is defined as the number of standard deviations that the raw score is from the mean. (In other words, it's the plain old standard score.) This standard score is a player's score in that particular game.
Getting a player's overall score is wicked easy - just take the average of all the individual game scores.
Hopefully I've convinced you that the formula works. Now let's look at why the min/max scores are -3/+3, respectively. To do this, let's consider a very lame game. Blue kills one creep and the game ends. So blue's raw score is +0.040, the other 9 players have a score of exactly 0. (Do you see why? Work out the formula above.)
Next we need the summary statistics. Mean = +0.040 / 10 = +0.004. Standard deviation = 0.012. Now, get the standard score for each player: (score - mean) / stddev
Blue: (0.040 - 0.004) / 0.012 = +3.000
Everyone else: (0.000 - 0.004) / 0.012 = -0.333
Strange huh? The magical number 3 arises as a result of there being 10 players in the game. (Image a version of DotA with 12 players. The guy with a single creep kill still has a raw score of +0.040, the mean is now 0.00333, and the stddev is now 0.01106. So in this trippy 12-man dota, the max score is now 3.316.)
There's one last sanity check that we can do with these 10 standardized scores. Add them all together, and you should get exactly zero. (+3 -.333 * 9) = 0. It's not just in this silly example either. For any game, the sum of all 10 scores will add to zero. Which proves that, as stated in the score page I linked above, the sum of all player scores server-wide including all games is zero. And hence, the average player is, by definition, assigned a score of zero.
About those magic k/d/a/cs multipliers. Way back in the beginning of DV, there was no score. I had the formula, but I needed the constants. To get them, I took the player database that I had at the time (which was something like 3,000 games - pretty small, but not too small), and I got averages for all k/d/a/cs. Essentially I asked "on average, how many kills does a player get for each death?"; and "how many kills for each assist?"; and so on. In other words, what's the average k/k ratio, k/-d ratio, k/a ratio, k/ck ratio, and k/cd ratio. These ratios (rounding a bit) are the source of the magic numbers. k/k = 1, k/-d = -0.952, and k/a = 0.8. As for creep stats, there's one extra step: divide their score contribution by four. (My reasoning was that creep k/d stats aren't quite as important as hero k/d/a stats.) So, k/ck/4 = 0.04 and k/cd/4 = 0.1.
Of course, these ratios are constantly fluctuating as new data gets added to the database. To try to keep things fair though, I've held the numbers constant since day 1. So if you played here 2 years ago, your score was calculated the same way as is today - no shenanigans.
Re: Elo system
Well, void's system is so unfair to supports. I remember some time ago I played amazing with dazzle, I won the game for my team, kept graving my carry and securing his kills, wasted all my money to wards, dust and dewarding and in the end i had fckin negative score(-0,200 i think) with 3-6-24 wtf.
Next game, I picked invoker, went mid, stomped opponent team, lost thanks to my team not wanting to push early and feeding opponent phantom lancer, but I ended up with the best score in game when i couldn't carry my team to victory http://www.dota-void.com/game.php?gameid=227165
And then it's this one http://www.dota-void.com/game.php?gameid=296930 where Im owning with mort, and in the most crucial moment i get alttabbed thanks to antivirus, only to see when i returned in game that we had lost. I was the reason we lost, but still, best score because i had a carry lol.
They had to add the purchase of wards and shit to the total score if they wanted to be fair to the players skill, because it's like them encouraging u to play only carry if u want to have positive score.
Next game, I picked invoker, went mid, stomped opponent team, lost thanks to my team not wanting to push early and feeding opponent phantom lancer, but I ended up with the best score in game when i couldn't carry my team to victory http://www.dota-void.com/game.php?gameid=227165
And then it's this one http://www.dota-void.com/game.php?gameid=296930 where Im owning with mort, and in the most crucial moment i get alttabbed thanks to antivirus, only to see when i returned in game that we had lost. I was the reason we lost, but still, best score because i had a carry lol.
They had to add the purchase of wards and shit to the total score if they wanted to be fair to the players skill, because it's like them encouraging u to play only carry if u want to have positive score.
-
child
Re: Elo system
wahts support? u mean sucky player who needs to buy wards tps and die for my carry wd??? u rly need to call solo mid or feed more
[spoiler=]
[/spoiler]
[spoiler=]
[/spoiler]Re: Elo system
nabo. wrote:First, I dont see why you should be winning more elo compared to when u lose... Win 10 elo, Lose 10 elo seems more reasonable.
Second, I dont see the problem with how the elo system is. Elo system is fine. Smurfing/sniping is the issue since it misrepresent stats. Numerous suggestions have been posted, but there havent been a really good one yet. If you stack: you have lower risk lower return. If you go against stack: you have higher risk, but higher return. Simple.
Third, If you good player you will have a high winning percentage, meaning you'll gain more elo than losing elo on avg. If you r losing more elo often, simply means you not so good as you think. Unless your elo skews the avg elo in game because your elo is either TOO high or TOO low, I dont see problem with elo system.
You dont really see my point, what im saying is that how the elo system is today you can stagg a team with new accs to either boost or ruin for an other player. for example. player X has 1300 elo, player y and q joins the game with new accounts ( both player y and q are pros) wich means that player X gets a pro team with low elo averages wich also mean when they probably win player X gets alot more Elo then normal. I think this is a problem. Its probably impossible but the best thing would be that you can only have 1 account at Entservers.
-
BlackSpade
- Forest Walker
- Posts: 140
- Joined: Fri Feb 01, 2013 8:50 am
- Been thanked: 1 time
Re: Elo system
greenmtn wrote:http://www.dota-void.com/score.php
How the score is calculated
The first step in calculating a player's score is to calculate a raw score for each player at the end of every game. Kills, deaths, and assists are the main contributing factors to the raw score, but creep stats are also accounted for. At the end of a game, each player is assigned a raw score based on the aforementioned stats. Next, both the average and standard deviation are calculated with all ten raw scores. Once this is done, each player is assigned a game-specific score which is the number of standard deviations he or she deviated from the mean. Finally, a player's overall score is simply the average of all his or her game scores.
This has to be the most retarded shit ever...... It's like baiting every players to play non support heroes and this would make players to play selfish with 0 team work.
From experience, there is no correct system to evaluate players skill (only thing that can evaluate players skill is from other players and not bunch of meaningless stats numbers)
far as elo or any other stats system goes! to me they are all pointless. every stats system has flaws because dota isn't single player based game.
Re: Elo system
There is also the TSR rating system which is supposed to better represent an individuals score in team games. I think it was first made for halo or something. The Elo system is for chess, which is meant for 1v1.
-
EdgeOfChaos
Re: Elo system
The first step in calculating a player's score is to calculate a raw score for each player at the end of every game. Kills, deaths, and assists are the main contributing factors to the raw score, but creep stats are also accounted for. At the end of a game, each player is assigned a raw score based on the aforementioned stats. Next, both the average and standard deviation are calculated with all ten raw scores. Once this is done, each player is assigned a game-specific score which is the number of standard deviations he or she deviated from the mean. Finally, a player's overall score is simply the average of all his or her game scores.
This may be the dumbest way of scoring I've ever had the misfortune of hearing. If you really want me to give every reason why it sucks, I can, but it should suffice to say that kills does not equal skill. This promotes KSing, mass carry picking, and team abandonment.
Example 1: I am axe. I can either choose to blink in, blade mail, and call all 5, probably resulting in my death but tanking for my carries giving them valuable time to crush them, or I can choose to wait until I see a low hp hero, blink in, ult, and run. I will choose the first if I want to win and the second if I want kills. Void's system will directly lead to the second happening.
Example 2: I go 3-0 mid. My team all sucks and gives up 10 kills, feeding them. They have a bara, I am sf. I can choose to keep playing and probably suffer deaths from the fed heroes on the enemy team, or I can choose to afk in base, keep my 3-0, and accept the loss. I will choose the second on void's system.
Example 3: I am centaur. I'm full hp. I see an allies mortdred at low hp running away from three heroes who are coming to kill her. I can either go back and let her blinkstrike to me and stomp them or I can let her die. If I go back, there's a chance they will catch us anyways and I will die in addition to her. Not worth the risk on void, better to let her die.
See why that's a harmful policy?
-
DreathKnight
- Protector of Nature
- Posts: 2140
- Joined: Sat Jun 30, 2012 4:52 am
- Location: Aiur
- Contact:
Re: Elo system
MrCool wrote:There is also the TSR rating system which is supposed to better represent an individuals score in team games. I think it was first made for halo or something. The Elo system is for chess, which is meant for 1v1.
uakf.b wrote:The only difference between TSR and ELO is that TSR changes more quickly if you have less games played. Neither takes into account "performance" during the game; the only metric that matters is winning/losing, because that is the goal of the game.
https://entgaming.net/forum/viewtopic.php?f=64&t=6256&p=28137&hilit=tsr#p28137
http://en.wikipedia.org/wiki/TrueSkill
-
Diablo_
- Protector of Nature
- Posts: 3180
- Joined: Sat Apr 06, 2013 10:26 pm
- Has thanked: 55 times
- Been thanked: 145 times
Re: Elo system
The only thing which matters are wins and losses, kills won't win games, assists won't win games, creep kills won't win games. A victory wins a game.
-----
LIHL player parser, a tool to automatically parse LIHL players' Elo and create reports for it: CLICK
LIHL player parser, a tool to automatically parse LIHL players' Elo and create reports for it: CLICK
-
EdgeOfChaos
Re: Elo system
I think TrueSkill is patented anyways, so you couldn't get it even if you wanted to.
And, diablo and blackspade speak the truth. Listen to them!
And, diablo and blackspade speak the truth. Listen to them!
Return to “Suggestion Archive”
Who is online
Users browsing this forum: No registered users and 17 guests