Imrpoved ranking system
Moderator: Moderators
-
PhailQuail
- Posts: 35
- Joined: 17 Sep 2008, 15:54
Imrpoved ranking system
The ranking system in Spring is pitiful, I am sure the majority would agree.
I can understand why this feature hasn't been developed, my assumption is that its because the Spring engine and the lobbies are separately developed.
I believe the lobbies should handle the ranking system, but the only statistic they can get about the game is how long a player is in it, allowing a greater assortment of game statistics, primarily who won/lost would help solve this problem.
These additional statistics need to be returned to the lobby application somehow, perhaps saving the stats to a file that both the lobby and engine know the location of.
In any case, after the game each lobby client should send the outcome of the game to the lobby server where the server will make adjustments to the players ranks based on the outcome.
An obvious flaw with the above is the cheating thats possible due to the lobbies being opensource (or at least springlobby), to help counter this the server should make note of any game outcome conflicts.
Games shorter than a certain amount of time should not be counted.
In the lobby itself, each players overall rank should be displayed next to their username in userlist, when in the Pre-game room, the ranks displayed should be specific to the mod being played.
These per-mod ranks should be viewable by anyone, as an example these might be accessible by right-clicking a user and selecting "Rank details" the "outcome confliction" percentage should also be displayed along with the per-mod ranks.
I use the term "rank" very unspecifically, call it what you want, score, penis points, waffles, whatever!
What I am interested in knowing is how the "score" should be calculated,
Should points be taken from the losing players and given to the winners?
Should higher-scoring players be punished more by losing to a lower-scoring player?
Should lower-scoring players get more points than usual by beating a better player?
How should points be calculated for non-even teams?
How should players who crash/rage be punished?
How should players who crash/rage effect the rest of the team?
So somehow the outcome of the game needs to be extracted from the game, perhaps this can be done by reading the demos? but do they store the outcome?
I would expect some people would disagree with aspects of a ranking system, but seriously, is it less pathetic than the "minutes played" system?
I can understand why this feature hasn't been developed, my assumption is that its because the Spring engine and the lobbies are separately developed.
I believe the lobbies should handle the ranking system, but the only statistic they can get about the game is how long a player is in it, allowing a greater assortment of game statistics, primarily who won/lost would help solve this problem.
These additional statistics need to be returned to the lobby application somehow, perhaps saving the stats to a file that both the lobby and engine know the location of.
In any case, after the game each lobby client should send the outcome of the game to the lobby server where the server will make adjustments to the players ranks based on the outcome.
An obvious flaw with the above is the cheating thats possible due to the lobbies being opensource (or at least springlobby), to help counter this the server should make note of any game outcome conflicts.
Games shorter than a certain amount of time should not be counted.
In the lobby itself, each players overall rank should be displayed next to their username in userlist, when in the Pre-game room, the ranks displayed should be specific to the mod being played.
These per-mod ranks should be viewable by anyone, as an example these might be accessible by right-clicking a user and selecting "Rank details" the "outcome confliction" percentage should also be displayed along with the per-mod ranks.
I use the term "rank" very unspecifically, call it what you want, score, penis points, waffles, whatever!
What I am interested in knowing is how the "score" should be calculated,
Should points be taken from the losing players and given to the winners?
Should higher-scoring players be punished more by losing to a lower-scoring player?
Should lower-scoring players get more points than usual by beating a better player?
How should points be calculated for non-even teams?
How should players who crash/rage be punished?
How should players who crash/rage effect the rest of the team?
So somehow the outcome of the game needs to be extracted from the game, perhaps this can be done by reading the demos? but do they store the outcome?
I would expect some people would disagree with aspects of a ranking system, but seriously, is it less pathetic than the "minutes played" system?
Re: Imrpoved ranking system
I see no real solution to team games with random people and easy abuse of the system, so why waste time implementing a system that would be just as flawed as the current one?
-
Google_Frog
- Moderator
- Posts: 2464
- Joined: 12 Oct 2007, 09:24
Re: Imrpoved ranking system
This already exists. https://blendax.informatik.uni-bremen.d ... ng/ladder/ There are team ladders for BA, CA XTA and KP and I'm sure if you want another one you just have to ask. All you have to do to play in a ladder is register a ladder account and join a ladder game which can be set in the hosting options.
The team ranking system assumes that each player has contributed equally to winning/losing so it's not really any better than the current system. With the current system it's safe to assume 2 or 3 chevron players have no clue and super starts are good. Anything else is really a guess.
The team ranking system assumes that each player has contributed equally to winning/losing so it's not really any better than the current system. With the current system it's safe to assume 2 or 3 chevron players have no clue and super starts are good. Anything else is really a guess.
Re: Imrpoved ranking system
Yeah, but for average lobby games, assuming all players in a team contirbutes equally just doesn't work, it works nice for clan games though, probably.
-
PhailQuail
- Posts: 35
- Joined: 17 Sep 2008, 15:54
Re: Imrpoved ranking system
Be optimistic, with a system like this in place, playing balanced team games with random people is much more likely.ZellSF wrote:I see no real solution to team games with random people and easy abuse of the system, so why waste time implementing a system that would be just as flawed as the current one?
And an option to disable this feature is very likely (every other RTS I've seen has it).
Not to sound like an asshole, but do you have a better suggestion?
Page Load Error, aside from the requirement that the player must randomly guess the URL to visit and not be freaked out by error and security warnings its not a very good idea, its implied that the only people who use that system are top players.Google_Frog wrote:This already exists. (...)
I doubt Big Bad Barry can use that website to steal by bank details, so why must it use encryption?
The current system is really bad, the most efficient method of "ranking up" involves being AFK and not actually playing.
-
Google_Frog
- Moderator
- Posts: 2464
- Joined: 12 Oct 2007, 09:24
Re: Imrpoved ranking system
They don't have to guess the URL. In TASClient it can be found under Help/Links -> Spring Ladder. Under options -> account there is an option to create a ladder account.PhailQuail wrote:Page Load Error, aside from the requirement that the player must randomly guess the URL to visit and not be freaked out by error and security warnings its not a very good idea, its implied that the only people who use that system are top players.
This system is in place and the fact that everyone isn't using it suggests that they don't want to put the extra effort in for a different ranking system.
In a perfectly balanced game there is a 50% chance that each side wins. That means that all players will have equal score leading to unbalanced games. It's an unstable system.Be optimistic, with a system like this in place, playing balanced team games with random people is much more likely.
The current system gives you a vague idea of how good a player is and it works most of the time. The most effective system I've found is the clan system, if you know of a good clan you can assume that a player with their tag is good. This doesn't work all the time but it does help.
Re: Imrpoved ranking system
No, not really. Your average game is 5vs5, the outcome of that just can't be used to judge how good a single player is.PhailQuail wrote:Be optimistic, with a system like this in place, playing balanced team games with random people is much more likely.ZellSF wrote:I see no real solution to team games with random people and easy abuse of the system, so why waste time implementing a system that would be just as flawed as the current one?
Yes, leaving the current system in place, it works, and no one's figured out a better system to replace it.Not to sound like an asshole, but do you have a better suggestion?
Setting up "fake" games to just win and increase your ranking with your system isn't exactly difficult either.The current system is really bad, the most efficient method of "ranking up" involves being AFK and not actually playing.
- Tribulexrenamed
- Posts: 775
- Joined: 22 Apr 2008, 19:06
Re: Imrpoved ranking system
Solution: Get some friends/clan
Every Ranking system is exploitable.
Every Ranking system is susceptible to smurfing.
Since the current ranking system is already so hard-set in place (with plenty of veterans) how to you expect to change over to a new system? Since it appears the problem is "fairness", how would this be "fair". Apparently the solution contradicts the predicament, and all solutions are thus false. Therefore, this thread is a useless ressurection of the same useless thread that has occurred about 100 times.
Every Ranking system is exploitable.
Every Ranking system is susceptible to smurfing.
Since the current ranking system is already so hard-set in place (with plenty of veterans) how to you expect to change over to a new system? Since it appears the problem is "fairness", how would this be "fair". Apparently the solution contradicts the predicament, and all solutions are thus false. Therefore, this thread is a useless ressurection of the same useless thread that has occurred about 100 times.
Re: Imrpoved ranking system
Do not discuss user friendliness in the forums.There are only a few people who will even listen.
Most will just tell you everything is fine and your wrong.
"spring is the most noob friendly game i know".
isn't the ladder broken anyway?
I remember Neddie saying something about him doing some work on it... .... ....
no clan ladders,no user profiles,smurfs everywhere...oh yeah...we are set to go.
Most will just tell you everything is fine and your wrong.
"spring is the most noob friendly game i know".
isn't the ladder broken anyway?
I remember Neddie saying something about him doing some work on it... .... ....
no clan ladders,no user profiles,smurfs everywhere...oh yeah...we are set to go.
Re: Imrpoved ranking system
I think a points system is far too finely graded. You can't measure a player's ability to that degree of accuracy, giving people scores is just going to lead to both gaming the system to gain extra points and people refusing to play against players with high scores. In my opinion there's pretty much 3 categories of players:
1) Players who are a liability to their team. Maybe they're new and they don't know how to play yet, maybe they're not so new but refuse to learn. Whatever the reason, they'll do things like ignore their team mates, fail to take any mexes, build guardians, teach with no resources, etc.
2) Competent players who understand the fundamentals and are at least capable of holding their side of a team game well enough for their team mates to win.
3) Outstanding players who, in a game with mainly category 2) competent players, are consistently able to swing the battle in their favour, whether by managing several things at once, organising the team into an effective strategy, or whatever.
I doubt I am alone in saying that all I want in a team game is my team to consist either of people in category 2), or at least to know who the category 1) players are and that there's an equal number on each team.
The current ladder system doesn't do a bad job at this really. It doesn't identify people in group 3) very well because there are both lots of people who have played nothing but speedmetal for hours on end, good players who haven't really played for very long and of course smurfs. But it's fairly easy to say that anyone without a silver star has a good chance of being in group 1 as long as they aren't a smurf.
If anything were to change I'd like to see a system with only these 3 ranks, where you are automatically rank 1 with less than say 10 hours in-game time, after that people can vote on your rating after a game. Regardless of votes your rank will tend towards rank 2 over time after you've played more than 10 hours, so you will have to be bad enough for people to consistently vote you down or good enough for people to consistently vote you up in order to be anything else. There won't be much point in smurfing because people will tend not to want rank 1s to be playing at all because it's a sign they are a liability.
If people still want their e-peen ladder rating then keep such a system separate and on top of the standard one so those of us who just want to identify who's going to be a liability to a team can do so without dealing with smurfing etc.
1) Players who are a liability to their team. Maybe they're new and they don't know how to play yet, maybe they're not so new but refuse to learn. Whatever the reason, they'll do things like ignore their team mates, fail to take any mexes, build guardians, teach with no resources, etc.
2) Competent players who understand the fundamentals and are at least capable of holding their side of a team game well enough for their team mates to win.
3) Outstanding players who, in a game with mainly category 2) competent players, are consistently able to swing the battle in their favour, whether by managing several things at once, organising the team into an effective strategy, or whatever.
I doubt I am alone in saying that all I want in a team game is my team to consist either of people in category 2), or at least to know who the category 1) players are and that there's an equal number on each team.
The current ladder system doesn't do a bad job at this really. It doesn't identify people in group 3) very well because there are both lots of people who have played nothing but speedmetal for hours on end, good players who haven't really played for very long and of course smurfs. But it's fairly easy to say that anyone without a silver star has a good chance of being in group 1 as long as they aren't a smurf.
If anything were to change I'd like to see a system with only these 3 ranks, where you are automatically rank 1 with less than say 10 hours in-game time, after that people can vote on your rating after a game. Regardless of votes your rank will tend towards rank 2 over time after you've played more than 10 hours, so you will have to be bad enough for people to consistently vote you down or good enough for people to consistently vote you up in order to be anything else. There won't be much point in smurfing because people will tend not to want rank 1s to be playing at all because it's a sign they are a liability.
If people still want their e-peen ladder rating then keep such a system separate and on top of the standard one so those of us who just want to identify who's going to be a liability to a team can do so without dealing with smurfing etc.
Re: Imrpoved ranking system
How about you suggest a system that might work rather than whine about how bad the current one is? It's not like everyone hates the idea of a new ranking system, but replacing the current system with a new system that will be equally flawed is just pointless.Gota wrote:Do not discuss user friendliness in the forums.There are only a few people who will even listen.
Most will just tell you everything is fine and your wrong.
"spring is the most noob friendly game i know".
isn't the ladder broken anyway?
I remember Neddie saying something about him doing some work on it... .... ....
no clan ladders,no user profiles,smurfs everywhere...oh yeah...we are set to go.
- Tribulexrenamed
- Posts: 775
- Joined: 22 Apr 2008, 19:06
Re: Imrpoved ranking system
Not only that, but the conversion between the old system and a new system is destined to be largely flawed, and source of much drama and dissolution in the community. You will be happy to know, however, that I dont really care either way. Just that ranking never effects me, and I never had a use or a problem with it, so i see no reason to invest effort in this meaningless endeavor.
- CarRepairer
- Cursed Zero-K Developer
- Posts: 3359
- Joined: 07 Nov 2007, 21:48
Re: Imrpoved ranking system
I attribute 90% of this problem to spring's default chat window. Solved by IceUI's scrollable chat.Scikar wrote: Whatever the reason, they'll do things like ignore their team mates,
Re: Imrpoved ranking system
The question is why do all browsers scream "Warning! This site is encrypted!". As if it's a bad thing. (No, I don't want an answer.)PhailQuail wrote:Page Load Error, aside from the requirement that the player must randomly guess the URL to visit and not be freaked out by error and security warnings its not a very good idea, its implied that the only people who use that system are top players.
I doubt Big Bad Barry can use that website to steal by bank details, so why must it use encryption?
Anyway, the auto-reports from the lobby are broken atm due to Spring changes. I'm working to fix that and implement a few new featuers while I'm at it. No eta though.
Re: Imrpoved ranking system
I have an idea that I would like to share. I am fully aware of the fact that it will never be incorporated into Spring during the lifetime of our sun, and I completely understand that it is easily exploitable even by a 10-year-old kid. Not to mention that it is technically impossible to be implemented into Spring to begin with, although it is based on Spring's gameplay. However, I will post it here anyway, hoping that it may inspire some new ideas for ranking systems in games other than Spring's.
This is basically a ranking system based on points called "Rank Points" (abbr. R-Points) Obviously players with more R-Points have higher ranks. I have to make this designation because there is another point system that I will explain later.
Initialization:
Each new player starts with one R-Point.
Earning Points:
Players can earn R-Points in several ways:
In-game time: gives very small amount of R-point per unit time only. This acts as "consolation prize" only and should only have minimal effect on ranking.
In-game contribution:
This makes up for the major part of R-points. Damage dealt and net metal/energy/units sent are all considered ("net" means that players can't send resources/units back and forth to gain more points). The next part is a bit complicated, so I will explain it step by step.
After a game ends, each team will be awarded by a certain amount of R-points that is proportional to the sum of all opponent's current R-points.
The R-points then will be distributed to each individual player in the team according to the proportion of contribution they have made as a team member. E.g. the more damage you dealt to your enemies compared to your fellow teammates, the more R-points you will be rewarded. This ratio is calculated by a "contribution points system". This point system exists in each game only, and will not affect ranking directly.
Damage dealing:
Contribution Points (abbr. C-Points) awarded are proportional to the amount of damage you have dealt to your enemies. Whom you have dealt damage to is also considered. Killing units of a high-ranked player will be considered more valuable than dealing the same amount of damage to a noob.
Resource/Unit Sharing:
The same goes for resource/unit sharing (not by the number of units you gave, but by their total cost). Of course these figures should be properly scaled, so that giving resources or units equivalent to one peewee will only be equivalent to dealing, say, 300 damage.
After all C-points are calculated, the portion of R-points a player gets is equal to the C-point the player has divided by the sum of C-points the whole team has.
Finally, winning a game will double the R-points you earn in this game.
After all points have been rewarded and saved into the database, the matrix will be normalized, meaning that the player with highest point will be scaled back to 1, and all other players' points are also scaled accordingly. E.g if there are 3 players in the system with {2, 1.5, 0.8} points respectively, they will then have {1, 0.75, 0.4} points respectively after normalization. Otherwise I can imagine the R-points exploding exponentially after a month or so.
This is just a preliminary concept and not even remotely functional. It still can't handle special cases such as players quitting before the game ends, and doesn't award players who don't contribute much in the statistics but did help the team significantly e.g. not directly engaging in battles but built 50 bombers to kill an adv. fusion, crippling the enemies' econ (and half of their base) thus turning the tide altogether.
This is basically a ranking system based on points called "Rank Points" (abbr. R-Points) Obviously players with more R-Points have higher ranks. I have to make this designation because there is another point system that I will explain later.
Initialization:
Each new player starts with one R-Point.
Earning Points:
Players can earn R-Points in several ways:
In-game time: gives very small amount of R-point per unit time only. This acts as "consolation prize" only and should only have minimal effect on ranking.
In-game contribution:
This makes up for the major part of R-points. Damage dealt and net metal/energy/units sent are all considered ("net" means that players can't send resources/units back and forth to gain more points). The next part is a bit complicated, so I will explain it step by step.
After a game ends, each team will be awarded by a certain amount of R-points that is proportional to the sum of all opponent's current R-points.
The R-points then will be distributed to each individual player in the team according to the proportion of contribution they have made as a team member. E.g. the more damage you dealt to your enemies compared to your fellow teammates, the more R-points you will be rewarded. This ratio is calculated by a "contribution points system". This point system exists in each game only, and will not affect ranking directly.
Damage dealing:
Contribution Points (abbr. C-Points) awarded are proportional to the amount of damage you have dealt to your enemies. Whom you have dealt damage to is also considered. Killing units of a high-ranked player will be considered more valuable than dealing the same amount of damage to a noob.
Resource/Unit Sharing:
The same goes for resource/unit sharing (not by the number of units you gave, but by their total cost). Of course these figures should be properly scaled, so that giving resources or units equivalent to one peewee will only be equivalent to dealing, say, 300 damage.
After all C-points are calculated, the portion of R-points a player gets is equal to the C-point the player has divided by the sum of C-points the whole team has.
Finally, winning a game will double the R-points you earn in this game.
After all points have been rewarded and saved into the database, the matrix will be normalized, meaning that the player with highest point will be scaled back to 1, and all other players' points are also scaled accordingly. E.g if there are 3 players in the system with {2, 1.5, 0.8} points respectively, they will then have {1, 0.75, 0.4} points respectively after normalization. Otherwise I can imagine the R-points exploding exponentially after a month or so.
This is just a preliminary concept and not even remotely functional. It still can't handle special cases such as players quitting before the game ends, and doesn't award players who don't contribute much in the statistics but did help the team significantly e.g. not directly engaging in battles but built 50 bombers to kill an adv. fusion, crippling the enemies' econ (and half of their base) thus turning the tide altogether.
Re: Imrpoved ranking system
"official" dedicated servers could track stats without a client being able to spoof and cheat
also, the upcoming lobby server will track many more stats than tasserver does.
also, the upcoming lobby server will track many more stats than tasserver does.
- Pressure Line
- Posts: 2283
- Joined: 21 May 2007, 02:09
Re: Imrpoved ranking system
agreed.Tribulex wrote:Just that ranking never effects me, and I never had a use or a problem with it, so i see no reason to invest effort in this meaningless endeavor.
Re: Imrpoved ranking system
I don't feel that a ranking system is an improvement over no ranking system, and I don't feel that the above suggested ranking systems are necessarily improvements over the current arbitrary time-based ranking system.
The arguments against ranking are self-evident and effectively incontrovertible. Ranking systems promote elitism. They can always be gamed. They discriminate against alternative forms of play. They cannot be applied monolithically to multiple games effectively. Even if they were comparatively effective in balancing they would simply supersede basic social interaction we should encourage, i.e. getting to know your fellow players in your timezone.
An optimistic view on this topic is, as on many, foolish and naive. Observation of any popularized title with an online ranking system related to win/loss or the accumulation of points though competitive play reveals a destructive result. Casual players are discouraged. Alternative players are discouraged. Victories are hoarded. Stacking though the loopholes provided by balance mechanisms is rampant. The play population of the game is pared away and you are left with an arbitrarily stratified player base. All well and good if you are at the top, but getting there so to speak is made artificially difficult.
Even if we desired the culture which would result from a "serious" rating system we don't have the people to support it. 1100 unique active players, only 160 or so regulars, spread across the games on the engine... as it is we hardly have enough people for regular play.
The arguments against ranking are self-evident and effectively incontrovertible. Ranking systems promote elitism. They can always be gamed. They discriminate against alternative forms of play. They cannot be applied monolithically to multiple games effectively. Even if they were comparatively effective in balancing they would simply supersede basic social interaction we should encourage, i.e. getting to know your fellow players in your timezone.
An optimistic view on this topic is, as on many, foolish and naive. Observation of any popularized title with an online ranking system related to win/loss or the accumulation of points though competitive play reveals a destructive result. Casual players are discouraged. Alternative players are discouraged. Victories are hoarded. Stacking though the loopholes provided by balance mechanisms is rampant. The play population of the game is pared away and you are left with an arbitrarily stratified player base. All well and good if you are at the top, but getting there so to speak is made artificially difficult.
Even if we desired the culture which would result from a "serious" rating system we don't have the people to support it. 1100 unique active players, only 160 or so regulars, spread across the games on the engine... as it is we hardly have enough people for regular play.
Re: Imrpoved ranking system
very nice written, neddiedrow! 
Re: Imrpoved ranking system
Yep, +1 to that.
Only improvement I think would be nice is splitting out the ingame time over the mods/games a user has played.
Only improvement I think would be nice is splitting out the ingame time over the mods/games a user has played.
