You as in (their) and you mean level? I for one love this being added(not because I added it) but because it makes me want to get better at the game, as well as others. "stacking and farming" is something you cannot easily control.
Why not? As has been said, implement the skill system the way, that you earn close to no skill when the enemy team has XXX skill less than the team you are playing on. This way you grab those ranking addicts at the balls and force them to play fair matches to get more skill. Easy as that.
And on top of this, you get a more accurate skill measurement for when the matchmaking is implemented.
@AceDude:
What is the problem on weighting the skill you get for a match on the skill the enemy team had? Or even better weight every kill / assist / death on the skill of the opponent you got.
Guys, i just came here interested in this new ranking system. IMO a very good choice, and with lots of potential. I managed to discern from this thread roughly how it works, although maybe i don't have all the details.
It seems the main discussion here is between AceDude and sotanaht, and correct me but it looks like you are looking for a comparative ranking system for games with more than 2 players? I agree with many things sotanaht has said here, i do not think estimating ranking from an equation of in-game stats is the way to go. It will introduce arbitrariness, possibility to manipulate stats and will not be accurate. I think there is a very good system that will accomplish the goals you have here, it is called TrueSkill and is developed by Microsoft engineers for multiplayer games (specifically Halo i think) and is an extension of the classic Elo-system.
Ace, please check this out, i hope it will solve your conundrum quite well.
For other people in the discussion:
The elegance of Elo-systems is that they completely ignore in-game stats, only the outcome of a match is relevant. Thus the players rankings between themselves (who is on the top of the leaderboard, who is second.. etc are compared to the skill of other players they have beaten or got beaten by.
Things like KDR-ratio will not play into this at all, since it is difficult to know how that stat affects match outcome. We will however need to make the assumption that the 'score' on the leaderboard is correct. That the total score shows your contribution to the game.
We do have a few issues to discuss for implementing a system like this i think, are we only going to rank complete matches for example? If not, hwo do we prevent player from dropping out to save their stats?
Oh, didn't know that. Well you can probably modify some parts of it, something in this direction should be the right way to go in any case. It must be possible to spin it somehow, but still utilize the principle.
What kind of system is used in LoL for example?
IMO we must go beyoind ranking simply win/loss of your team (like the discussion has been about so far) but also weight in rankings of players within the teams. Although i agree with sotanaht's conclusions (ranking by win/loss is the best system proposed so far) it will likely be very slow to converge towards a real ranking.
I'm thinking a mixed system:
1. The first component transfers a number of ELO points from the losing teams players to the winning teams players, who the points are taken from and who recieves them will be determined by the scoreboard rankings in the teams.
2. The second component transfers a number of ELO points between players within a team, based on their ELO rank and location on the scoreboard.
The parameters deciding the amount in component 1/2 are up to the designer here of course (Ace).
The result is that winning the match will be top priority as it adds to the pool of ELO points for your team, and to get points from this pool you need to be a major contributor to the victory (determined by scoreboard position, not total score). It is also important to play according to your skill compared to other players within your team, since there will be some point redistribution for a good scoreboard position.
I propose that players who have entered the game after start, or leave the game during the game will also participate in the scoring process, but will be accordingly weighted by playing time in the process, somehow.
I also think it is necessary for players to be able to determine if they want to be ranked or not in matches they play, to allow them to play casual games. It could be as simple as a checkbox, and they will be left out of the scoring process (not lose/gain any points). In essence be invisible to the ranking system.
I couldn't join a team on a server because 'Your Elo rating is too high'.
Who made that terrible servermod?
Hi that would be me,
and no that's not a bad mod, it's mainly meant to provide a way to split servers into different skill classes. There is only one server running it atm and that would be the rookie server of german-slaughterhouse.
If you have a Elo higher than 1650 you are not a rookie anymore and skilled enought to play on normal servers.
We don't call it noobs only, because that term would be wrong. PPl around 1500 are just avg./ new players, anything higher 1650 are advanced player and therefore should play on normal servers.
Well, it pissed me off when it was about the only populated server left yesterday night / this morning.
The server name only says 'rookie friendly', i suggest making it something more informative ('low skill only').
Well, it pissed me off when it was about the only populated server left yesterday night / this morning.
The server name only says 'rookie friendly', i suggest making it something more informative ('low skill only').
You know "low skill" sounds harsh. And it's not low it's avg. skill level what you could expect from a rookie friendly server.
What if I'm playing on a server where noone listens to my orders as a commander/field commander? This happens very often. You're also saying that skill should be highly dependent on the rest of the team. Bad idea.
Matchmaking... hmm... we're not talking about matchmaking yet.
How will Elo work any better for commanders than for groundplayers when both have the same team reliance issue?
1) yes
2) it should work, they both have the same team reliance issue
So does player skill have an Elo component? Seems weird that you say Elo doesn't work well for team-based games and skill shouldn't be dependent on the rest of the team, but then say that commanders (who don't choose their teams and whose success is largely determined by their team) will be ranked by Elo...
What if I'm playing on a server where noone listens to my orders as a commander/field commander? This happens very often. You're also saying that skill should be highly dependent on the rest of the team. Bad idea.
Matchmaking... hmm... we're not talking about matchmaking yet.
How will Elo work any better for commanders than for groundplayers when both have the same team reliance issue?
1) yes
2) it should work, they both have the same team reliance issue
So does player skill have an Elo component? Seems weird that you say Elo doesn't work well for team-based games and skill shouldn't be dependent on the rest of the team, but then say that commanders (who don't choose their teams and whose success is largely determined by their team) will be ranked by Elo...
True, but Elo-type systems depend on a large number of games to be stable and one must assume that the factor of unbalanced teams will not be a factor after commanding ~50 games.
However if using player-based Elo systems one could easily make teh commander-Elo converge (stabilize) more quickly by weighting the points based on their teammembers Elo. The designer would, however, need to decide how large the component of a good team will be versus the skill of the commander.
The amount of skill points you gain or lose from a match should probably be at least clamped then, so you can't gain anything past a certain point by farming rookies. And can't lose plenty of points just because the enemy team decides to farm your out-teched ass instead of ending the game.
Average of all games played is the only fair way to go. That's the point, it's average. You may often perform better or worse than that, that's not the point.
What if I'm playing on a server where noone listens to my orders as a commander/field commander? This happens very often. You're also saying that skill should be highly dependent on the rest of the team. Bad idea.
Matchmaking... hmm... we're not talking about matchmaking yet.
How will Elo work any better for commanders than for groundplayers when both have the same team reliance issue?
1) yes
2) it should work, they both have the same team reliance issue
hate to burst your bubble, but that will exasperate the problem of starting a game in pubs. sometimes you already have the problem of nobody wanting to jump in hive without throwing in negative incentives like lowered score.
Average of all games played is the only fair way to go. That's the point, it's average. You may often perform better or worse than that, that's not the point.
I disagree with this point due to the following nuance: I got my PC in 2011 and started playing ZPS as my first FPS. For a few months my K/D was probably around 1 - I mean I was starting not only as a noob to ZPS but a noob to using a mouse in an FPS in general. Then in the last few months I played the game it was around about 20 - double the second best person. Yet according to my entire average I wasn't in the top ten. My point being that I improved at an extraordinarily exponential rate (and was playing with the same people all of the time because I was part of a community) but because I was terrible at the beginning it made me out just to look like a decent player. Not that anyone really paid attention to it but I suppose matchmaking wasn't an issue in that game as it is here, so the anecdote is pertinent.
So perhaps it should only count your last x number of games to be more representative of how you are now (even though, of course, the whole system is hardly representative as it is in its current form).
as long as the last X round is something like 50-100 or it will be statistically meaningless. Because K:D actually has no place in a correctly configured 'chance to positively impact team success rate' statistic, which is actually all we should be looking at.
"No you didn't skill rating is an average of the skill from every game you've had recorded. You probably had a really bad game and it dropped as such. "
So you DID have a "bad" round then. That is to say, you got less skill of a skill rating that round than your average.
EDIT: Not surprising, since as a gorge the only time you get a lot of points is end game bile base rape.
really depends on how derpy the other team is. playing a lot of gorge to tone it down for newer players. you can get some lerk numbers with a proper gorge wall.
Am I alone in thinking it might be neat to see an average skill number when you look at the hive stats? Something for people to compare themselves to, maybe something to work towards or maybe give a small amount of gratification to those above average.
Yes, as it would make the attitude of the community even worse than it is (if that is possible).
Er - because, I went gorge, we won, then my 'skill' went down - I think I can handle that level of causality, thank you.
it's actually hard to tell sometimes, does it update immediately? give it the benefit of the doubt im saying... that being said, it's completely believable that the system penalizes gorging
Comments
Why not? As has been said, implement the skill system the way, that you earn close to no skill when the enemy team has XXX skill less than the team you are playing on. This way you grab those ranking addicts at the balls and force them to play fair matches to get more skill. Easy as that.
And on top of this, you get a more accurate skill measurement for when the matchmaking is implemented.
@AceDude:
What is the problem on weighting the skill you get for a match on the skill the enemy team had? Or even better weight every kill / assist / death on the skill of the opponent you got.
It seems the main discussion here is between AceDude and sotanaht, and correct me but it looks like you are looking for a comparative ranking system for games with more than 2 players? I agree with many things sotanaht has said here, i do not think estimating ranking from an equation of in-game stats is the way to go. It will introduce arbitrariness, possibility to manipulate stats and will not be accurate. I think there is a very good system that will accomplish the goals you have here, it is called TrueSkill and is developed by Microsoft engineers for multiplayer games (specifically Halo i think) and is an extension of the classic Elo-system.
Ace, please check this out, i hope it will solve your conundrum quite well.
http://research.microsoft.com/en-us/projects/trueskill/
http://machinelearning.wustl.edu/mlpapers/paper_files/NIPS2006_688.pdf
For other people in the discussion:
The elegance of Elo-systems is that they completely ignore in-game stats, only the outcome of a match is relevant. Thus the players rankings between themselves (who is on the top of the leaderboard, who is second.. etc are compared to the skill of other players they have beaten or got beaten by.
Things like KDR-ratio will not play into this at all, since it is difficult to know how that stat affects match outcome. We will however need to make the assumption that the 'score' on the leaderboard is correct. That the total score shows your contribution to the game.
We do have a few issues to discuss for implementing a system like this i think, are we only going to rank complete matches for example? If not, hwo do we prevent player from dropping out to save their stats?
http://www.google.com/patents/US20090227313
Oh, didn't know that. Well you can probably modify some parts of it, something in this direction should be the right way to go in any case. It must be possible to spin it somehow, but still utilize the principle.
What kind of system is used in LoL for example?
IMO we must go beyoind ranking simply win/loss of your team (like the discussion has been about so far) but also weight in rankings of players within the teams. Although i agree with sotanaht's conclusions (ranking by win/loss is the best system proposed so far) it will likely be very slow to converge towards a real ranking.
I'm thinking a mixed system:
1. The first component transfers a number of ELO points from the losing teams players to the winning teams players, who the points are taken from and who recieves them will be determined by the scoreboard rankings in the teams.
2. The second component transfers a number of ELO points between players within a team, based on their ELO rank and location on the scoreboard.
The parameters deciding the amount in component 1/2 are up to the designer here of course (Ace).
The result is that winning the match will be top priority as it adds to the pool of ELO points for your team, and to get points from this pool you need to be a major contributor to the victory (determined by scoreboard position, not total score). It is also important to play according to your skill compared to other players within your team, since there will be some point redistribution for a good scoreboard position.
I propose that players who have entered the game after start, or leave the game during the game will also participate in the scoring process, but will be accordingly weighted by playing time in the process, somehow.
I also think it is necessary for players to be able to determine if they want to be ranked or not in matches they play, to allow them to play casual games. It could be as simple as a checkbox, and they will be left out of the scoring process (not lose/gain any points). In essence be invisible to the ranking system.
Who made that terrible servermod?
Hi that would be me,
and no that's not a bad mod, it's mainly meant to provide a way to split servers into different skill classes. There is only one server running it atm and that would be the rookie server of german-slaughterhouse.
If you have a Elo higher than 1650 you are not a rookie anymore and skilled enought to play on normal servers.
We don't call it noobs only, because that term would be wrong. PPl around 1500 are just avg./ new players, anything higher 1650 are advanced player and therefore should play on normal servers.
The server name only says 'rookie friendly', i suggest making it something more informative ('low skill only').
You know "low skill" sounds harsh. And it's not low it's avg. skill level what you could expect from a rookie friendly server.
So does player skill have an Elo component? Seems weird that you say Elo doesn't work well for team-based games and skill shouldn't be dependent on the rest of the team, but then say that commanders (who don't choose their teams and whose success is largely determined by their team) will be ranked by Elo...
True, but Elo-type systems depend on a large number of games to be stable and one must assume that the factor of unbalanced teams will not be a factor after commanding ~50 games.
However if using player-based Elo systems one could easily make teh commander-Elo converge (stabilize) more quickly by weighting the points based on their teammembers Elo. The designer would, however, need to decide how large the component of a good team will be versus the skill of the commander.
hate to burst your bubble, but that will exasperate the problem of starting a game in pubs. sometimes you already have the problem of nobody wanting to jump in hive without throwing in negative incentives like lowered score.
I don't think it is. I think it is an average of your last 20 games. I have seen mine go down despite the latest game being higher than average.
I disagree with this point due to the following nuance: I got my PC in 2011 and started playing ZPS as my first FPS. For a few months my K/D was probably around 1 - I mean I was starting not only as a noob to ZPS but a noob to using a mouse in an FPS in general. Then in the last few months I played the game it was around about 20 - double the second best person. Yet according to my entire average I wasn't in the top ten. My point being that I improved at an extraordinarily exponential rate (and was playing with the same people all of the time because I was part of a community) but because I was terrible at the beginning it made me out just to look like a decent player. Not that anyone really paid attention to it but I suppose matchmaking wasn't an issue in that game as it is here, so the anecdote is pertinent.
So perhaps it should only count your last x number of games to be more representative of how you are now (even though, of course, the whole system is hardly representative as it is in its current form).
I actually find myself preferring that... For the current setup anyway.
Or I Gorged, won, then went down 30 points.
EDIT: Not surprising, since as a gorge the only time you get a lot of points is end game bile base rape.
really depends on how derpy the other team is. playing a lot of gorge to tone it down for newer players. you can get some lerk numbers with a proper gorge wall.
You can also spam Gorge Tunnels... the only "bigger" amount of skill points you get in one go.
Er - because, I went gorge, we won, then my 'skill' went down - I think I can handle that level of causality, thank you.
it's actually hard to tell sometimes, does it update immediately? give it the benefit of the doubt im saying... that being said, it's completely believable that the system penalizes gorging