@Pelargir, I should be fair to you here, I've lost a lerk down ventilation before now
A lerk? I know i'm a fade but, does a lerk not supposed to fly? Glad to know there are people worse than I am.
In my defense, it was a long time ago, not in your case. xD
@_INTER_ Thanks for the explanation although I think your missing the point of my original (now unwarranted) rant; that is a negative offset of scores is psychologically demoralizing to see even when you have worked tirelessly in the team, as a team, yet lost. Unwarranted due to the fact that I now know it was never intended or designed this way and it's in fact a bug. The ranking system revamp is fully welcomed if with a positive, or zero, offset irrespective of win or not.
@_INTER_ Thanks for the explanation although I think your missing the point of my original (now unwarranted) rant; that is a negative offset of scores is psychologically demoralizing to see even when you have worked tirelessly in the team, as a team, yet lost. Unwarranted due to the fact that I now know it was never intended or designed this way and it's in fact a bug. The ranking system revamp is fully welcomed if with a positive, or zero, offset irrespective of win or not.
I'm not sure what you want to say. You would like a skill system that only counts up? how could that work?
@Asraniel From what I'm reading that's the way the original formulas was designed -- it was centered around a positive offset like 400 so unlikely to ever receive a negative score however it's bugged. I hope that's the case as I'm strongly against the "CS:GO" style ranking system where you can go back to square one in terms of stats. A positive-only score system could absolutely be designed to work and is far more psychologically rewarding and stimulating than to see a negative amount on your rank.
@Asraniel From what I'm reading that's the way the original formulas was designed -- it was centered around a positive offset like 400 so unlikely to ever receive a negative score however it's bugged. I hope that's the case as I'm strongly against the "CS:GO" style ranking system where you can go back to square one in terms of stats. A positive-only score system could absolutely be designed to work and is far more psychologically rewarding and stimulating than to see a negative amount on your rank.
I could not yet find the blogpost that explains the new skill system, so i can't be totally sure. But having a positive only skill system would make no sense. How would you compare two players that did not play the same amount of games? Would a player that plays a lot have an infinite skill one day? Any reasonable skill system has to adapt the player skill up and down to work correctly. There is a separate stat, called "level" in hive that only goes upwards
@Asraniel and @garreh just in case you guys misunderstood each other:
garreh wants positive values for skill which still fluctuates, e.g. inbetween 100 and 700 instead of -300 and 300.
i think Asraniel interpreted that as a continuously increasing skill value (like the hive score) which basically increases without any limit.
@Pelargir, I should be fair to you here, I've lost a lerk down ventilation before now
A lerk? I know i'm a fade but, does a lerk not supposed to fly? Glad to know there are people worse than I am.
In my defense, it was a long time ago, not in your case. xD
Actually it was probably around the same time ago And it wasn't in a competitive match either :P
@Asraniel From what I'm reading that's the way the original formulas was designed -- it was centered around a positive offset like 400 so unlikely to ever receive a negative score however it's bugged. I hope that's the case as I'm strongly against the "CS:GO" style ranking system where you can go back to square one in terms of stats. A positive-only score system could absolutely be designed to work and is far more psychologically rewarding and stimulating than to see a negative amount on your rank.
I could not yet find the blogpost that explains the new skill system, so i can't be totally sure. But having a positive only skill system would make no sense. How would you compare two players that did not play the same amount of games? Would a player that plays a lot have an infinite skill one day? Any reasonable skill system has to adapt the player skill up and down to work correctly. There is a separate stat, called "level" in hive that only goes upwards
Since you're subtracting the skill values, any constant offset normally cancels out so it doesn't matter to the math. The only time this really matters is when teams have a different number of players. If there are negative skill values then the system is likely to predict that the team with more players is at a disadvantage if it's a low skill game, because it's sum will be more negative than the sum of the other team.
It makes sense to cap them at 0 I think. Skill values will drift up over time on average, but everything will still work correctly.
@Laosh'Ra Thanks buddy -- that's exactly what I'm trying to get at. Technically, it will work the same but in appearance it won't look like "omgz negative score you so bad!" type thing.
@Laosh'Ra Thanks buddy -- that's exactly what I'm trying to get at. Technically, it will work the same but in appearance it won't look like "omgz negative score you so bad!" type thing.
Then it's really difficult to understand what you mean exactly. I too thought you just wanted a system that only adjusts your skill upwards and never down. Could you expand on how is the CS:GO model bad?
BeigeAlertTexasJoin Date: 2013-08-08Member: 186657Members, Super Administrators, Forum Admins, NS2 Developer, NS2 Playtester, Squad Five Blue, Squad Five Silver, NS2 Map Tester, Reinforced - Diamond, Reinforced - Shadow, Subnautica Playtester, Pistachionauts
Does the system take into account people joining/leaving? I've had games that start stacked on one side, then due to people leaving and joining, ends up being stacked the other way.
Does the system take into account people joining/leaving? I've had games that start stacked on one side, then due to people leaving and joining, ends up being stacked the other way.
Yes. Score is scalled based on how much time you played durring the match.
What about if you only play on the week ends ? If stats are bounded with the number of games played, it will leave a gap between skill announced and real skill.
What about the role issue (gorge / commander) ?
I don't really understand your question about only playing on the weekends. But I don't think it matters at all. Sure, if you don't play that often then it will take longer for you skill to converge, but that's about it.
The system only takes wins and losses into account. So if you are a gorge that's good enough to make you team win then your skill will increase. The same goes for being commander.
Wow... Only win / loose? herr... And how do you remove stacking ?
Every player plays for fun and Win because Win is better... So if it is a Win/Lost game skill system; all players will do more stacking on the public servers. No doubt.
The skill system will not distinguish the top of the line players from others, because they probably ended up on the "dumb team" +2 or N times more than it should be in a raw. Even if they do a 100:50:1 on the scoreboard but had a base rush, they just loose XP.
Then when you try to balance it with automatically it just end up in a mess the same way as before and with more stacking. I see it already happening. It's cool that i only play on Sunday and Saturday. I clearly see the changes from week to the next.
The -347 score is utterly stupid. Let's say you played a real intense 1 hour game, racked up a score of 800 yet your team lose. So suddenly your hive reports a negative score.
1. How is that fair exactly? To all those players on the team that fought their hardest -- lost yet had lots of personal victories and absolutely zero reward, in fact negative reward.
I agree but would not talk about reward. XP should go up depending on victory but not only. Especially at a high skill level. Things can turn upside down so fast that win or loose is like flipping a coin sometimes.
2. What incentive is there to continue playing on that server even though you know it's unlikely your win, because you know your end up with a negative score on the hive?
None. Quitters are gonna leave faster, and move to another game faster.
3. What's stopping someone from leaving the server, rejoining, and waiting in the ready room (as far as I know the score is 'finalised' when either team wins/loses, so if your not on the server at that point the hive won't report it.)
I really loved the old ns2 where regardless of if your team won or not your personal efforts towards the goal of winning was recognized. Now they are just shrugged off.
I understand everyone receives the same treatment, so it balances itself out
On the contrary. It will end up bad. Stackers at top, then... others.
-- but psychologically it's quite demoralizing to receive a negative score and feel like your going backwards. If the -ve score centered the distribution around a non-zero offset e.g. 100 this would greatly help.
@UncleRunch Stackers dont get much skill, when they win but loose a lot when they lost the round. Therefor they wont end up on top unless they win a lot of games without loosing once. So it balances out. This behaviour motivates joining the weaker team.
@UncleCrunch - FFS! Why are you responding to garreh's quotes - it has already been established that he didn't understand how the skill system works. His objections are unwarrented, and so are yours. Please, take the time to read the VERY DETAILED write up on how the skill system works, before you make this wall of text, demonstrating your complete ignorance on the matter.
Please mods, for the love of sanity, close this god damn thread already!
Quote from the link above:
Properties of the Basic Model
1. If one team smashes another, but the teams were stacked to make that outcome almost certain, nobody’s skill level changes at the end of the round. Only unexpected victories change skill levels.
2. Nothing you do during the round other than winning the game has any effect on your skill level. This means that there’s no incentive to rack up kills at the end of the game rather than finishing it, or to play skulk instead of gorge, or to go offense rather than build.
3. The effect on your score is determined by the time you spent playing the round rather than your points, with the beginning of the game weighted much higher than the end. This means that it doesn’t harm you to join a game that is already lost, because you’ll have played for a very small fraction of the weighted time spent in the game.
4. If the two teams have a different number of players, this is compensated for automatically by normalizing by the total time across both teams rather than just within a team, and the model predicts that the team with more players will win.
5. Larger games contribute less to each player’s skill level because the presumption is that each player has a smaller effect on the outcome of the game.
2. Nothing you do during the round other than winning the game has any effect on your skill level. This means that there’s no incentive to rack up kills at the end of the game rather than finishing it, or to play skulk instead of gorge, or to go offense rather than build.
2. Nothing you do during the round other than winning the game has any effect on your skill level. This means that there’s no incentive to rack up kills at the end of the game rather than finishing it, or to play skulk instead of gorge, or to go offense rather than build.
And how do you balance games with that ?
That particular property, is not about balance, it's about incentive to play to win rather than farming for skill points. Think of it as social engineering.
3. The effect on your score is determined by the time you spent playing the round rather than your points, with the beginning of the game weighted much higher than the end. This means that it doesn’t harm you to join a game that is already lost, because you’ll have played for a very small fraction of the weighted time spent in the game.
I wonder if this one is implemented properly; P needs to take into account that the team had one player short for a while or one player left midgame and then a new player filled the spot. And not just use the old P calculated from the start or the new P with the new player. It should be some mix right? We'll see how @AceDude etc. solved it (hopefully this weekend).
2. Nothing you do during the round other than winning the game has any effect on your skill level. This means that there’s no incentive to rack up kills at the end of the game rather than finishing it, or to play skulk instead of gorge, or to go offense rather than build.
And how do you balance games with that ?
Look up the Elo rating system for an example. You win, you gain points, you lose, you drop points. If you're rated higher than your opponent, you won't gain much points if you win, but you can drop lots of points if you lose. Play enough games and in the end, the rating is a good representation of a players skill. The NS2 skill system designed on thesame principle.
Look up the Elo rating system for an example. You win, you gain points, you lose, you drop points. If you're rated higher than your opponent, you won't gain much points if you win, but you can drop lots of points if you lose. Play enough games and in the end, the rating is a good representation of a players skill. The NS2 skill system designed on thesame principle.
I don't think this new team balancing mechanic is working out so well, or at least I haven't gotten to try it much. The vote force even teams has a tarnished reputation and everyone hates it. It rarely passes when brought up and I lost count how many times someone yells out "Don't, it sucks", "Vote stack, lol guys!", etc.
I don't think this new team balancing mechanic is working out so well, or at least I haven't gotten to try it much. The vote force even teams has a tarnished reputation and everyone hates it. It rarely passes when brought up and I lost count how many times someone yells out "Don't, it sucks", "Vote stack, lol guys!", etc.
It's because in it's current implementation it's broken. There is no explanation, bar wrong implementation/maths, for someone to go from +360 points to -360 points (720 point loss) in one game, particularly when that player is carrying a losing team. It might be possible for someone to have -360 in one game and that effect their global score level through an average, but it appears the hive is only showing your global current skill level as the points you gained or lost in the last round.
If this implementation / maths behind the new skill system is NOT broken in the mind of CDT, then it's efficacy is precisely 0.
Does the system recalculate the probability of a team winning or losing every time a player leaves or joins the (ongoing) game?
AFIAK, it computes it at the end of the game after recording all the players that left and joined.
I think there are some things that will need to be tweaked having to do with how it responds to teams with different number of players, but this shouldn't have a huge effect.
Look up the Elo rating system for an example. You win, you gain points, you lose, you drop points. If you're rated higher than your opponent, you won't gain much points if you win, but you can drop lots of points if you lose. Play enough games and in the end, the rating is a good representation of a players skill. The NS2 skill system designed on thesame principle.
And we don't see it on the score board...
nice...
I assume that if skill values were visible on the scoreboard before the system has been operational long enough to have a reasonable chance of generating accurate values, then there would be a whole bunch of complaints by people who expect it to work immediately.
I'm okay with not seeing them right away, because it avoids the inevitable shitposts and threads from newly registered members who don't understand that these systems require large amounts of data before it can become even vaguely representative of a player's skill. Also, and perhaps more importantly, it avoids people crying stack in-game based on skill-values before they even mean anything.
I would actually understand if skill is never implemented into the vanilla ns2 scoreboard, even if I would prefer it to happen eventually. Force even teams can always be used to balance a game. And mods can always be created for server operators who prefer these skill values to be visible.
Comments
A lerk? I know i'm a fade but, does a lerk not supposed to fly? Glad to know there are people worse than I am.
In my defense, it was a long time ago, not in your case. xD
I'm not sure what you want to say. You would like a skill system that only counts up? how could that work?
I could not yet find the blogpost that explains the new skill system, so i can't be totally sure. But having a positive only skill system would make no sense. How would you compare two players that did not play the same amount of games? Would a player that plays a lot have an infinite skill one day? Any reasonable skill system has to adapt the player skill up and down to work correctly. There is a separate stat, called "level" in hive that only goes upwards
garreh wants positive values for skill which still fluctuates, e.g. inbetween 100 and 700 instead of -300 and 300.
i think Asraniel interpreted that as a continuously increasing skill value (like the hive score) which basically increases without any limit.
Actually it was probably around the same time ago And it wasn't in a competitive match either :P
http://moultano.wordpress.com/2014/08/04/a-skill-ranking-system-for-natural-selection-2/
Since you're subtracting the skill values, any constant offset normally cancels out so it doesn't matter to the math. The only time this really matters is when teams have a different number of players. If there are negative skill values then the system is likely to predict that the team with more players is at a disadvantage if it's a low skill game, because it's sum will be more negative than the sum of the other team.
It makes sense to cap them at 0 I think. Skill values will drift up over time on average, but everything will still work correctly.
Then it's really difficult to understand what you mean exactly. I too thought you just wanted a system that only adjusts your skill upwards and never down. Could you expand on how is the CS:GO model bad?
Wow... Only win / loose? herr... And how do you remove stacking ?
Every player plays for fun and Win because Win is better... So if it is a Win/Lost game skill system; all players will do more stacking on the public servers. No doubt.
The skill system will not distinguish the top of the line players from others, because they probably ended up on the "dumb team" +2 or N times more than it should be in a raw. Even if they do a 100:50:1 on the scoreboard but had a base rush, they just loose XP.
Then when you try to balance it with automatically it just end up in a mess the same way as before and with more stacking. I see it already happening. It's cool that i only play on Sunday and Saturday. I clearly see the changes from week to the next.
I agree but would not talk about reward. XP should go up depending on victory but not only. Especially at a high skill level. Things can turn upside down so fast that win or loose is like flipping a coin sometimes.
None. Quitters are gonna leave faster, and move to another game faster.
On the contrary. It will end up bad. Stackers at top, then... others.
Please mods, for the love of sanity, close this god damn thread already!
Quote from the link above:
1. If one team smashes another, but the teams were stacked to make that outcome almost certain, nobody’s skill level changes at the end of the round. Only unexpected victories change skill levels.
2. Nothing you do during the round other than winning the game has any effect on your skill level. This means that there’s no incentive to rack up kills at the end of the game rather than finishing it, or to play skulk instead of gorge, or to go offense rather than build.
3. The effect on your score is determined by the time you spent playing the round rather than your points, with the beginning of the game weighted much higher than the end. This means that it doesn’t harm you to join a game that is already lost, because you’ll have played for a very small fraction of the weighted time spent in the game.
4. If the two teams have a different number of players, this is compensated for automatically by normalizing by the total time across both teams rather than just within a team, and the model predicts that the team with more players will win.
5. Larger games contribute less to each player’s skill level because the presumption is that each player has a smaller effect on the outcome of the game.
And how do you balance games with that ?
That particular property, is not about balance, it's about incentive to play to win rather than farming for skill points. Think of it as social engineering.
Look up the Elo rating system for an example. You win, you gain points, you lose, you drop points. If you're rated higher than your opponent, you won't gain much points if you win, but you can drop lots of points if you lose. Play enough games and in the end, the rating is a good representation of a players skill. The NS2 skill system designed on thesame principle.
And we don't see it on the score board...
nice...
It's because in it's current implementation it's broken. There is no explanation, bar wrong implementation/maths, for someone to go from +360 points to -360 points (720 point loss) in one game, particularly when that player is carrying a losing team. It might be possible for someone to have -360 in one game and that effect their global score level through an average, but it appears the hive is only showing your global current skill level as the points you gained or lost in the last round.
If this implementation / maths behind the new skill system is NOT broken in the mind of CDT, then it's efficacy is precisely 0.
@Obraxis will post the system source code pretty soon. I want to refactor it before and show it to @moultano.
AFIAK, it computes it at the end of the game after recording all the players that left and joined.
I think there are some things that will need to be tweaked having to do with how it responds to teams with different number of players, but this shouldn't have a huge effect.
"looks like" does not sound confident.
I assume that if skill values were visible on the scoreboard before the system has been operational long enough to have a reasonable chance of generating accurate values, then there would be a whole bunch of complaints by people who expect it to work immediately.
I'm okay with not seeing them right away, because it avoids the inevitable shitposts and threads from newly registered members who don't understand that these systems require large amounts of data before it can become even vaguely representative of a player's skill. Also, and perhaps more importantly, it avoids people crying stack in-game based on skill-values before they even mean anything.
I would actually understand if skill is never implemented into the vanilla ns2 scoreboard, even if I would prefer it to happen eventually. Force even teams can always be used to balance a game. And mods can always be created for server operators who prefer these skill values to be visible.