The only reason I've thought that this might be a good measurement is that most of the time the players contributing the most to the team's win are at the top of the board due to the Score points - and very often they have poor KDR... so it's already counting things that aren't too abuse-able at least.
I understand what you mean, but idk, i often have a very good kdr, but not because i go out hunting for the kills. I go to the strategic points on the map eg system & dome and try to hold them. Not only does it attract alot of aliens to you it reduces their map control and allows potentially less good aimers to cap.
I guess it all depends on the intentions of the player. Are they just running around shooting stuff with no situational awareness or are they trying to be strategic.
I often don't build RT's, but i'll be stood there covering whilst someone else does.
A ranking system is never going to suit every players play style, especially in a game as complex as ns.
Score/min is so directly correlated to your kdr that I think it would be better than kdr in all ways for this game to balance skill rankings. If you play with certain people you notice they'll have a high player damage but abysmally low structure damage and build time.
In the ratio of psmg-sdmg-btime, where on marines I get 10000-30000-1:30 and the top fragger gets 15000-5000-0:20, I don't care about people saying he was the best player. I know that all he was is the most selfish player and that my game was played significantly better than his and I had more effect. I understand his frags are important, but so is solo capping and maintaining 1v1 victories.
This kind of kdr is better represented through score. And kdr is not the same for pressure and solo-capping.
I'm impressed whenever I see someone having done high structural damage and it certainly has a major impact on your team winning. It deserves just as much recognition as a rambo. Most of the newer players tend to put kdr on a higher pedestal since they come from more traditional FPS games. I don't really know whether your game play was significantly better or more effective than the rambo, though.
Imo, there's a time and place for different actions. For example (concerning pubs), if you know your team is mostly full of average or below-average players, surely you will consider skipping building base and naturals so that you can maximize your benefit to your team by killing skulks/gorges/structures early on. If you're one of the comm's more valuable players, he will probably tell you to do so. I tend to do this by default unless no one follows me on round start or if I'm somehow last out the door. At any rate, capping can be left to others if they're already heading toward the direction of uncapped res, you're closer to the frontlines, you're pretty sure said capper won't get ambushed while building, etc. If no one's in the vicinity to cap except you, of course it would be your responsibility to cap unless your comm has like 1 tres, someone already stated they would cap, whatever. I don't think even that hypothetical rambo would deliberately choose not to cap unless there was something more important to do. Even if said rambo was truly selfish, players are free to play however they like and the hive doesn't seem to take the more selfless actions into account. Therefore, I agree with you that score/min would be a better indicator of skill as opposed to kdr.
I think the truly skilled player is one who is capable of dealing damage to both players and structures while juggling tasks like building/welding, calling out intel, lane blocking, pinching higher lifeforms, being a field comm, scouting for PGs/tunnels, getting ninja PGs/tunnels up, etc. as the game state demands. It's been my aspiration since 500~ hours to become a better player and I've been able to reach 4k by following the above mindset of trying to do everything as humanly possible to benefit my team. It's based on weighing risk/rewards, making the best decision on the spot, and completing tasks quickly.
Of course, it helped that I didn't play competitively and I mostly play on large servers where engagements are much more frequent. My real hive skill is probably 3k at best. I think most competitive players deserve a much higher hive skill than pubstars. Perhaps taking win/loss out of the equation like MoFo1 suggested would help balance it out more. It's funny how the current rank #1 has such a high skill that is nearly impossible to reach through regular means. Perhaps he was just trying to make a point by exploiting some faulty mechanics. Can't wait to see how the new hive will work out.
Yes elo is personal, and as @Therius said, with enough games,
Which did and will NOT happen with the player base we have (334 on Avg all over the world -> diff timezones). Especially when we take under consideration the 18+ slots servers that act as a magnet. So basically you're saying it doesn't fit as a solution for the problem at hand.
for in the same manner when your ass is carried hard.
The 1800+ aren't the ones who get carried. They only rest in stacked games. Other than that they carry or are equally matched in game full of veterans.
Pissing on floor laughing. The problem isn't an equation ripped off from a math book. The equation will process the data as any function. The problem is the initial assumption that says basically, if the team wins all members of the team do provide positive actions towards victory. We all know it's far from the truth in general. Only the competitive do that. Given the competitive status isn't for a day or a rookie subscribing to NSL. oops.
The players that are hard carries have elos like 2800+. They can make up for a whole team of garbage. Their elos reflect this.
Even if one of the godlike does that for one game, and assuming (s)he can carry the game until the end (Which isn't possible). The guy just survived of a boring to death game, and will try to avoid that for ever. Basically, these godlike don't do that, they just move to another server, not even saying goodbye.
Players that aren't carries and cannot shift the tide of a game on their own do not have elos anywhere near that stratosphere.
Why? Because elo is accurate.
Play 2 games in which your supposed to win and fail to win. Your ELO will move towards 0. Your aim didn't changed, your moves didn't changed, everything that is measurable didn't changed. You're the same now compared to the last hour. Still... the ELO moves. Especially when this skills of yours are precisely the things that made you... that forged the hero you are (hundreds of hours i suppose)...
Of course, this system can be and has been abused by players familiar with its nuance, but that does not mean you should throw the baby out with the bath water.
As long as it based only on the win/lose of a team (the same principles), you can.
I'm not saying I don't think it could be improved, but I'm saying if you play enough games, w/l will reflect accurately enough your likelihood to win. You can't honestly look at the hive dashboard's top players and tell me that their elo is not accurate.
Again "not enough games"... Plus the ELO of the top players is reflecting only one thing. They win more with their team (usually comp), leading the hive score towards infinity. AND / OR they play more, because playing and winning more games will just add to/subtract from the score. There is no limit. Humans can push the limits to an extent. But not up to infinity.
If you say: Well it's more like a sorting system. i answer : well why it's not working for the "middle skill players"?.
Why does it only consider the local scope instead of the global scope... An unfortunate competitive player can look like an average player on a public server while it's by far the best on the server.
Yes there are outliers, but again... The concept of elo in general principal works well enough.
Yes it works when it's used properly and using the appropriate initial assumptions. This is the prerequisite for anything you create to solve a problem. Sadly, I can't say this is done properly.
This was a post about me being bad at NS2, not a comment on the skill system..
When I made a thread in August 2014 commenting about a bug with the arrows on the scoreboard it turned into the longest thread yet on the skill system 16 pages. I ended up just changing the thread title to match the content of the thread.
And the award for the hardest-to-decipher post goes to...
I disagree, this is the easiest to read post he has ever written.
@Unclecrunch and @IeptBarakat I will try and put this in as simple terms as I can. First of all no one is saying hive is perfect, we all agree there are outliers, but the method is not as flawed as you think.
Using Win Loss does by its nature reflects other metrics such as KD, SPM, field comm ability, and anything else you can think of.
Look at the following correlation table.
The important thing to notice is that the killrate (KD) is almost as correlated to hive skill as the win rate (Win Loss) is. Even though KD is not measured directly it still is measured and is a big part of a players hive score. I am currently working on getting the correlations for SPM and other factors not shown in that graph.
If you think that KD, SPM, or anything else should be weighted heavier, you are wrong. They are weighted at an emergent rate. I have shown that KD is already highly weighed. If a skill does not contribute to a win, then it is not an important skill. If a skill does contribute to a win then the better you are at it, the higher your skill value will go.
"Play 2 games in which your supposed to win and fail to win. Your ELO will move towards 0. Your aim didn't changed, your moves didn't changed, everything that is measurable didn't changed."
This complain is really tiresome. You played 2 games and you hive score went down. You will play 2 more games and it will go right back up. For every game that you lose for unfair reasons, you will also win for unfair reasons.
Please provide an actual argument against why, "For every game that you lose for unfair reasons, you will also win for unfair reasons," is not true for the majority of players?
IeptBarakatThe most difficult name to speak ingame.Join Date: 2009-07-10Member: 68107Members, Constellation, NS2 Playtester, Squad Five Blue, NS2 Map Tester, Reinforced - Diamond, Reinforced - Shadow
Basically you're saying the W/L dependent hive score is buffering skill in a matter of chance and is okay as you assume the more you play the higher chance your team will win and it will balance out in the end.
No wonder why shuffling never results in a good game.
Not that I think it's necessary to point out any more inherent problems with the K/D and SpM skill systems. But I will anyway.
If you think that getting on a bad team is unfair with the current system, you're gonna have a BAD time with a K/D or SpM system.
It is inherently easier to get higher scores per minute and kills, if you are on the winning team. Not just because you will be leading in TECH, but you will also have the momentum and the map control.
So how do you account for that? You give the losing team an arbitrary score multiplier to compensate?
Any problem you raise with the current elo system, is not fixed by a K/D or SpM system.
The chance of losing unfairly is equal to the chance of winning unfairly for the vast majority of players.
This is not why shuffling does not always result in that vague term, "good games." That is mostly, but not entirely, a result from these factors.
1. NS2 is an asymmetric game. Aliens and marines need separate hive skill values.
2. The skill system is a bit too relativistic. Someone who has a high skill value on a no rookies pub server would have a higher skill value on a random pub server where the average skill is lesser.
3. New rookies start with a hive skill of 0, and returning players which are often rookies start from a hive skill of 1000.
4. Role selection. If a player is a really good fade, and has gained a high skill value, if he plays as commander his skill value is no longer reflective on his ability to contribute to a win.
5. Skill variance. When you have a server with players ranging from 500 up to 3000 skill there is a lot of skill variance. It is incredibly hard to balance with high skill variance. This skill variance also further enhances the problems caused by role selection.
6. Team selection. Both FET and shuffle try to keep you on the team you joined. It tries to balance with your preference in mind. This restricts the balance functions ability make the most balances teams possible.
7. Smurfs. One smurf can be so misrepresented he can really mess up the skill function. Not only does a smurf hurt balance, it also messes with everybody else skill value temporarily messing balance up even further. No, this is not a way that you can lose unfairly. The odds of you being on the team with the smurf, or against the team with the smurf are equal.
8. People switch after FET or shuffle was made.
IeptBarakatThe most difficult name to speak ingame.Join Date: 2009-07-10Member: 68107Members, Constellation, NS2 Playtester, Squad Five Blue, NS2 Map Tester, Reinforced - Diamond, Reinforced - Shadow
It may not have been the op's original intention with the thread, but criticizing the skill system in which is the reason why the thread was made in the first place is still a relative topic piece. However, I will concede on the matter for the fact that there hasn't been any good answers for making the current system better.
Three things can greatly improve the skill system, two of which will likely be done with hive 2.0. They do not fix all eight problems I listed, but these three things will have the biggest effect.
1. Separate skill ratings for marines and aliens.
2. Fix the fact that new rookies start at 0 and returning players start at 1000.
3. Increase the player base so that the there is less skill variance on a given server. If the player base numbers are improved enough, segregating skills would help this a lot. This segregation would also further improve the relativistic nature of hive.
I still would like for someone provide an valid argument against why, "For every game that you lose for unfair reasons, you will also win for unfair reasons," is not true for the majority of players?
I still would like for someone provide an valid argument against why, "For every game that you lose for unfair reasons, you will also win for unfair reasons," is not true for the majority of players?
Well.. If you account for the imbalance between factions, then I don't agree with this statement. However, that's hyperbole to the discussion.
Not only that, but using K/D or SpM encourages specific behaviors while Win Loss only encourages winning.
Actually.. Win/loss encourages stacking.
It also encourages f4ing mid game to switch to the winning team..
It could also encourage certain individuals to spend a day artificially lowering their score (troll comm, play bad on purpose, let their rookie friend play on their account, etc) so they get shuffled to the good team for a few days...
My point here is that any system could be abused..
I still think that kdr and spm should affect the score somewhat though. I've seen too many people with insanely high kdr, who end games with double the score of anyone on the server, yet their hive score is a fraction of what it should be.
I still think that kdr and spm should affect the score somewhat though. I've seen too many people with insanely high kdr, who end games with double the score of anyone on the server, yet their hive score is a fraction of what it should be.
They already do is major ways.
First read this little primer on what that means.
The main result of a correlation is called the correlation coefficient (or "r"). It ranges from -1.0 to +1.0. The closer r is to +1 or -1, the more closely the two variables are related.
If r is close to 0, it means there is no relationship between the variables. If r is positive, it means that as one variable gets larger the other gets larger. If r is negative it means that as one gets larger, the other gets smaller (often called an "inverse" correlation).
KDR correlates to hive skill at .451. SPM correlates to hive skill at 0.543. Win loss correlates to hive skill at .561. SPM is almost as important as Win loss.
I still would like for someone provide an valid argument against why, "For every game that you lose for unfair reasons, you will also win for unfair reasons," is not true for the majority of players?
Well.. If you account for the imbalance between factions, then I don't agree with this statement. However, that's hyperbole to the discussion.
People do have preferences of which team. Hive shows I play aliens 2/3 games. When I play aliens more, I get a higher hive score. When I play marines more I get a lower hive score.
Please provide an actual argument against why, "For every game that you lose for unfair reasons, you will also win for unfair reasons," is not true for the majority of players?
Let's assume "unfair" is "stacked". If you loose while you're supposed to win; you loose more than you will win in the opposite situation. I've seen people (2500+) on the top of the Hive list complaining and targeting others (be mean) because they just lost +150 points in one game. Everybody familiar with the system knew he will have to play a lot more games to get back where he was. Simply because he can't get a game without a stack and also because his factor is contributing to this. He will end up in the team supposed to win anyway.
Frankly I have other priorities than my appendix size when I play. But they don't!!! That's where it goes to ridiculously insane length. The more you encourage this, the more it'll ruin the game. And the more they will farm. Up to an extent the hive skill factor cannot be considered as a "rough indicator". It led to every single crap behavior players could do, and I'm not playing everyday, only the WE.
It is stupid to have no limits on a "factor". The game provides rules and limits you can't break (default speed etc.). Every human has limitation (reflexes, aim etc..). You don't magically get a bigger rack of teeth / gun. The hive score has only one limit : infinity.
I'm not saying W/L shouldn't be taken under consideration, but it shouldn't be the only measurement. Every example has been given. 2000+ go gorge or commander, Unfortunate competitive player meets public server, carrying a game alone, not moving everyone in RR etc.
Hive skill system is like saying "you already read a book by just looking at the tittle". It may work for stupid online games, but not NS. Why ? Simply because the alien side is providing lifeforms accessible at the begining of the game. Those lifeforms have a clearly different impact on the game AND depending on WHO use it and WHEN. THIS is NOT predictable. Same goes for the marines when they can buy JP. But it's not at the beginning and that saves the skill system for the moment.
The best way is to profile players. Aim (even K/D if you want), damage on structures (sometime you only do that). Preferred lifeforms, life span as a specific lifeform. Ever been commander ?
On a side note; I would add a proximity measurement. If you're always close to your teammates it means you're coordinating (more of less). The competitive players would get points (or a 0-100% bar) even if they fail to win a game. That would be a counter to W/L score included as part of a new system. They would still loose games but they will be able to maintain their skill factor because they simply play well. closing the ")".
And that's the point. We don't need to know if someone has a bigger finger than the other, just to know what that someone can do, what he usually do in a game. Good aim, bad aim etc...
Then you will know how to even the teams for every single player by getting their profile and try to find his closest twin(s) / counter. It's even easier to mix rookies with vets. And eventually get them together, playing together (teamwork).
EX: You know a rookie always evolve gorge at start. Then you know he's gonna be pretty inactive (or static, fortifying). You shouldn't give the marine team more players than alien. Then in the opposite situation (rookies stay skulk, or even better, he goes Lerk). Then you know you can give an extra rookie to the marine team.
Same goes for "spear people" who always get in the front making as much as possible their way through it. You won't put all of them in the same team.
The only thing you need is a proper set of solvers (answering a single question by 0/1) organized in a tree. Find the matching couples and mix it. "How many potential rookie gorge at the beginning", etc.
All player would be able to get a solid idea of who can do what, and it will be stable for each player (in a game and looking at their own stats on the web). Behavior, habits, some go gorge some go fade, some are more versed on SG or mines, some are Exo... That is not "readable" with the actual skill factor number. It's why we have some completely crappy games where it was supposed to be even.
You won't see someone's aim going from 5% to 75% (Marine) other than using a 3rd party program. It may go from 15% to 25% maybe 30%... but it won't have a huge impact on the whole profile because of the other samples on skills. You just want to know what it will probably do in this game (most likely).
The only real random event that can happen is when someone change completely. He was always gorge, now he goes fade. We all know it's not happening in the real world. But we all know those 2 lifeforms have a different impact on the game.
Yes it means getting numbers we may not have but you seem to have plenty already.
Also there will be no "score" to grow. More like "you are doing this at the moment". And we gonna loose the pinkie size problem. The mood would improve on the servers.
The problem of balancing a NS2 game has never been about power (size), but more about how to use that power and IF it is used. And that's because of the last "if", the statistics are necessary. Just to get a better focused picture of the potential future game.
And if we could have a huddle system (as suggested before) the skill system could analyze an even better focused picture of what will happen in the next game. Simply because the players will "announce" what they will do. 4 Fades announced against rookies won't work right ? The system could signal this and maybe the alien players could decide not to have 3 fades to even things outs.
We won't avoid every bad game. But... at least we can have rookies playing alongside veteran for the greater good. Last time i checked ? Even the teams, still end up in a mess.
The detractors would say that it would make the same teams over and over. Well; this can be fixed easily if it were to happen. Storing information about the previous game isn't out of reach.
I strictly prefer a bunch of good games :
filled with fun
no size problem floating around ghost
helping the rookies eventually
playing more custom because people won't have anything to loose
Reconcile Competitive players vs Public play
with more girls (I guess they're hiding because of the mood other there)
by surrendering some randomness about team mixing.
...than going through the same crap that led to (and still do)
Bad player retention
Ego maniac people
Toxic people (sometimes including me when i loose my nerves on public server)
Rolling on rookies and make them quit for good
We don't need a powerful CDT/PT/UWE marketing stuff, just something in which we can be comfortable. Good candy doesn't need marketing. The players are the best marketing for a game.
Score/min is so directly correlated to your kdr that I think it would be better than kdr in all ways for this game to balance skill rankings.
It is indeed highly correlated at a rate of 0.718.
Here is how a lot of factors correlate for players greater than 100 hours. In order from highest to lowest.
W/L - Win-Loss-Ratio
K/D - Kill-Death-Ratio
Skill - Hive Skill
SPM - Score per minute
time - hive time on record
A <-> B : C - corr(A,B)=C
K/D <-> SPM : 0.718
Skill <-> W/L : 0.580
W/L <-> K/D : 0.546
Skill <-> SPM : 0.494
Skill <-> K/D : 0.521
Skill <-> time : 0.476
W/L <-> SPM : 0.434
K/D <-> time : 0.193
SPM <-> time : 0.155
W/L <-> time : 0.107
If you don't know what a correlation coeffecient is, read the following.
The main result of a correlation is called the correlation coefficient (or "r"). It ranges from -1.0 to +1.0. The closer r is to +1 or -1, the more closely the two variables are related.
If r is close to 0, it means there is no relationship between the variables. If r is positive, it means that as one variable gets larger the other gets larger. If r is negative it means that as one gets larger, the other gets smaller (often called an "inverse" correlation)
Soul_RiderMod BeanJoin Date: 2004-06-19Member: 29388Members, Constellation, Squad Five Blue
Score per minute does include gorge building, and to a degree commanders as well I believe, so might be the better option than k/d, purely because it takes more of the game scoring mechanics into play.
Soul_RiderMod BeanJoin Date: 2004-06-19Member: 29388Members, Constellation, Squad Five Blue
edited February 2016
It does, but as a personal skill score, it doesn't represent anything interesting. Currently your hive skill represents the weight of your winning potential.
I've been thinking about it, but is it feasible to vary the no. of points gained/loss from a win/loss according to percentage of score of team?
For example, if you are the top ranking player on the team, you should theoretically be getting the highest score too. So if you don't get the percentage expected of your score's percentage, you get less points from a win and lose more from a loss. If you get more than expected for your score, you get more points from a win and less from a loss.
Of course some things that are important are not tracked by scores, but the advantage of this is that it may lead to a faster convergence, which helps mitigate the not enough games played some what.
It does not help with factors such as skill being relative to the server you play on, but lets not kid ourselves. No system will account for that. Again, I think you guys have overly high expectations of the hive system. It is best used as a rough gauge, not as a balancing tool. There are other factors outside of a player's skill evaluation that contribute to unbalanced games such as disparity in hive scores on the server that will not go away even with a perfect skill evaluation tool. Sometimes a bad mix of players is just that and not stacked teams.
Regardless, negative scores should not occur when it is "hard limited to zero".
All people with negative scores played their last match around august of 2014. That's when we had the first implementation of hive skill, which was highly bugged. Some days later it was corrected and everyones skill was set to 1000 (but this value is only updated when you play your next match)
If I recall correctly, a few weeks later it was added the hard limit at 0 at the same time that it was decided to make rookies start with skill 0
People are unable to understand that you can conjure up the most wild scenarios and experience the most vivid anecdotes which would seem to work against the foundations of the skill system, but this has nothing to do with the big picture. In any statistical system, there will be outliers, there will be scenarios that twist the numbers in an unfair way. But in the big picture, these balance each other out as they work for a player as much as they work against them, given enough games.
Coming up with examples of individual situations where the skill system is wrong holds absolutely no weight.
Soul_RiderMod BeanJoin Date: 2004-06-19Member: 29388Members, Constellation, Squad Five Blue
Of course, I think people seem to forget that numbers can never truly account for the human element. The skill is just a weighted potential, a decent guess, but nothing more.
Which is why I'd rather the base was something more interesting to a person, like SPM. If scoring could be better weighted and distributed within the game, it would make an accurate measure of your skill based on something interesting, rather than win/loss
Comments
I understand what you mean, but idk, i often have a very good kdr, but not because i go out hunting for the kills. I go to the strategic points on the map eg system & dome and try to hold them. Not only does it attract alot of aliens to you it reduces their map control and allows potentially less good aimers to cap.
I guess it all depends on the intentions of the player. Are they just running around shooting stuff with no situational awareness or are they trying to be strategic.
I often don't build RT's, but i'll be stood there covering whilst someone else does.
A ranking system is never going to suit every players play style, especially in a game as complex as ns.
Imo, there's a time and place for different actions. For example (concerning pubs), if you know your team is mostly full of average or below-average players, surely you will consider skipping building base and naturals so that you can maximize your benefit to your team by killing skulks/gorges/structures early on. If you're one of the comm's more valuable players, he will probably tell you to do so. I tend to do this by default unless no one follows me on round start or if I'm somehow last out the door. At any rate, capping can be left to others if they're already heading toward the direction of uncapped res, you're closer to the frontlines, you're pretty sure said capper won't get ambushed while building, etc. If no one's in the vicinity to cap except you, of course it would be your responsibility to cap unless your comm has like 1 tres, someone already stated they would cap, whatever. I don't think even that hypothetical rambo would deliberately choose not to cap unless there was something more important to do. Even if said rambo was truly selfish, players are free to play however they like and the hive doesn't seem to take the more selfless actions into account. Therefore, I agree with you that score/min would be a better indicator of skill as opposed to kdr.
I think the truly skilled player is one who is capable of dealing damage to both players and structures while juggling tasks like building/welding, calling out intel, lane blocking, pinching higher lifeforms, being a field comm, scouting for PGs/tunnels, getting ninja PGs/tunnels up, etc. as the game state demands. It's been my aspiration since 500~ hours to become a better player and I've been able to reach 4k by following the above mindset of trying to do everything as humanly possible to benefit my team. It's based on weighing risk/rewards, making the best decision on the spot, and completing tasks quickly.
Of course, it helped that I didn't play competitively and I mostly play on large servers where engagements are much more frequent. My real hive skill is probably 3k at best. I think most competitive players deserve a much higher hive skill than pubstars. Perhaps taking win/loss out of the equation like MoFo1 suggested would help balance it out more. It's funny how the current rank #1 has such a high skill that is nearly impossible to reach through regular means. Perhaps he was just trying to make a point by exploiting some faulty mechanics. Can't wait to see how the new hive will work out.
Play 2 games in which your supposed to win and fail to win. Your ELO will move towards 0. Your aim didn't changed, your moves didn't changed, everything that is measurable didn't changed. You're the same now compared to the last hour. Still... the ELO moves. Especially when this skills of yours are precisely the things that made you... that forged the hero you are (hundreds of hours i suppose)...
As long as it based only on the win/lose of a team (the same principles), you can.
Again "not enough games"... Plus the ELO of the top players is reflecting only one thing. They win more with their team (usually comp), leading the hive score towards infinity. AND / OR they play more, because playing and winning more games will just add to/subtract from the score. There is no limit. Humans can push the limits to an extent. But not up to infinity.
If you say: Well it's more like a sorting system. i answer : well why it's not working for the "middle skill players"?.
Why does it only consider the local scope instead of the global scope... An unfortunate competitive player can look like an average player on a public server while it's by far the best on the server.
Yes it works when it's used properly and using the appropriate initial assumptions. This is the prerequisite for anything you create to solve a problem. Sadly, I can't say this is done properly.
When I made a thread in August 2014 commenting about a bug with the arrows on the scoreboard it turned into the longest thread yet on the skill system 16 pages. I ended up just changing the thread title to match the content of the thread.
I disagree, this is the easiest to read post he has ever written.
@Unclecrunch and @IeptBarakat I will try and put this in as simple terms as I can. First of all no one is saying hive is perfect, we all agree there are outliers, but the method is not as flawed as you think.
Using Win Loss does by its nature reflects other metrics such as KD, SPM, field comm ability, and anything else you can think of.
Look at the following correlation table.
The important thing to notice is that the killrate (KD) is almost as correlated to hive skill as the win rate (Win Loss) is. Even though KD is not measured directly it still is measured and is a big part of a players hive score. I am currently working on getting the correlations for SPM and other factors not shown in that graph.
If you think that KD, SPM, or anything else should be weighted heavier, you are wrong. They are weighted at an emergent rate. I have shown that KD is already highly weighed. If a skill does not contribute to a win, then it is not an important skill. If a skill does contribute to a win then the better you are at it, the higher your skill value will go.
"Play 2 games in which your supposed to win and fail to win. Your ELO will move towards 0. Your aim didn't changed, your moves didn't changed, everything that is measurable didn't changed."
This complain is really tiresome. You played 2 games and you hive score went down. You will play 2 more games and it will go right back up. For every game that you lose for unfair reasons, you will also win for unfair reasons.
Please provide an actual argument against why, "For every game that you lose for unfair reasons, you will also win for unfair reasons," is not true for the majority of players?
No wonder why shuffling never results in a good game.
If you think that getting on a bad team is unfair with the current system, you're gonna have a BAD time with a K/D or SpM system.
It is inherently easier to get higher scores per minute and kills, if you are on the winning team. Not just because you will be leading in TECH, but you will also have the momentum and the map control.
So how do you account for that? You give the losing team an arbitrary score multiplier to compensate?
Any problem you raise with the current elo system, is not fixed by a K/D or SpM system.
This is not why shuffling does not always result in that vague term, "good games." That is mostly, but not entirely, a result from these factors.
1. NS2 is an asymmetric game. Aliens and marines need separate hive skill values.
2. The skill system is a bit too relativistic. Someone who has a high skill value on a no rookies pub server would have a higher skill value on a random pub server where the average skill is lesser.
3. New rookies start with a hive skill of 0, and returning players which are often rookies start from a hive skill of 1000.
4. Role selection. If a player is a really good fade, and has gained a high skill value, if he plays as commander his skill value is no longer reflective on his ability to contribute to a win.
5. Skill variance. When you have a server with players ranging from 500 up to 3000 skill there is a lot of skill variance. It is incredibly hard to balance with high skill variance. This skill variance also further enhances the problems caused by role selection.
6. Team selection. Both FET and shuffle try to keep you on the team you joined. It tries to balance with your preference in mind. This restricts the balance functions ability make the most balances teams possible.
7. Smurfs. One smurf can be so misrepresented he can really mess up the skill function. Not only does a smurf hurt balance, it also messes with everybody else skill value temporarily messing balance up even further. No, this is not a way that you can lose unfairly. The odds of you being on the team with the smurf, or against the team with the smurf are equal.
8. People switch after FET or shuffle was made.
Whats funnier is how it is always the same people arguing the same points. It makes me question not only my use of time, but this forum communities.
Not only that, but using K/D or SpM encourages specific behaviors while Win Loss only encourages winning.
1. Separate skill ratings for marines and aliens.
2. Fix the fact that new rookies start at 0 and returning players start at 1000.
3. Increase the player base so that the there is less skill variance on a given server. If the player base numbers are improved enough, segregating skills would help this a lot. This segregation would also further improve the relativistic nature of hive.
I still would like for someone provide an valid argument against why, "For every game that you lose for unfair reasons, you will also win for unfair reasons," is not true for the majority of players?
Actually.. Win/loss encourages stacking.
It also encourages f4ing mid game to switch to the winning team..
It could also encourage certain individuals to spend a day artificially lowering their score (troll comm, play bad on purpose, let their rookie friend play on their account, etc) so they get shuffled to the good team for a few days...
My point here is that any system could be abused..
I still think that kdr and spm should affect the score somewhat though. I've seen too many people with insanely high kdr, who end games with double the score of anyone on the server, yet their hive score is a fraction of what it should be.
They already do is major ways.
First read this little primer on what that means.
The main result of a correlation is called the correlation coefficient (or "r"). It ranges from -1.0 to +1.0. The closer r is to +1 or -1, the more closely the two variables are related.
If r is close to 0, it means there is no relationship between the variables. If r is positive, it means that as one variable gets larger the other gets larger. If r is negative it means that as one gets larger, the other gets smaller (often called an "inverse" correlation).
KDR correlates to hive skill at .451. SPM correlates to hive skill at 0.543. Win loss correlates to hive skill at .561. SPM is almost as important as Win loss.
People do have preferences of which team. Hive shows I play aliens 2/3 games. When I play aliens more, I get a higher hive score. When I play marines more I get a lower hive score.
Let's assume "unfair" is "stacked". If you loose while you're supposed to win; you loose more than you will win in the opposite situation. I've seen people (2500+) on the top of the Hive list complaining and targeting others (be mean) because they just lost +150 points in one game. Everybody familiar with the system knew he will have to play a lot more games to get back where he was. Simply because he can't get a game without a stack and also because his factor is contributing to this. He will end up in the team supposed to win anyway.
Frankly I have other priorities than my appendix size when I play. But they don't!!! That's where it goes to ridiculously insane length. The more you encourage this, the more it'll ruin the game. And the more they will farm. Up to an extent the hive skill factor cannot be considered as a "rough indicator". It led to every single crap behavior players could do, and I'm not playing everyday, only the WE.
It is stupid to have no limits on a "factor". The game provides rules and limits you can't break (default speed etc.). Every human has limitation (reflexes, aim etc..). You don't magically get a bigger rack of teeth / gun. The hive score has only one limit : infinity.
I'm not saying W/L shouldn't be taken under consideration, but it shouldn't be the only measurement. Every example has been given. 2000+ go gorge or commander, Unfortunate competitive player meets public server, carrying a game alone, not moving everyone in RR etc.
Hive skill system is like saying "you already read a book by just looking at the tittle". It may work for stupid online games, but not NS. Why ? Simply because the alien side is providing lifeforms accessible at the begining of the game. Those lifeforms have a clearly different impact on the game AND depending on WHO use it and WHEN. THIS is NOT predictable. Same goes for the marines when they can buy JP. But it's not at the beginning and that saves the skill system for the moment.
The best way is to profile players. Aim (even K/D if you want), damage on structures (sometime you only do that). Preferred lifeforms, life span as a specific lifeform. Ever been commander ?
On a side note; I would add a proximity measurement. If you're always close to your teammates it means you're coordinating (more of less). The competitive players would get points (or a 0-100% bar) even if they fail to win a game. That would be a counter to W/L score included as part of a new system. They would still loose games but they will be able to maintain their skill factor because they simply play well. closing the ")".
And that's the point. We don't need to know if someone has a bigger finger than the other, just to know what that someone can do, what he usually do in a game. Good aim, bad aim etc...
Then you will know how to even the teams for every single player by getting their profile and try to find his closest twin(s) / counter. It's even easier to mix rookies with vets. And eventually get them together, playing together (teamwork).
EX: You know a rookie always evolve gorge at start. Then you know he's gonna be pretty inactive (or static, fortifying). You shouldn't give the marine team more players than alien. Then in the opposite situation (rookies stay skulk, or even better, he goes Lerk). Then you know you can give an extra rookie to the marine team.
Same goes for "spear people" who always get in the front making as much as possible their way through it. You won't put all of them in the same team.
The only thing you need is a proper set of solvers (answering a single question by 0/1) organized in a tree. Find the matching couples and mix it. "How many potential rookie gorge at the beginning", etc.
All player would be able to get a solid idea of who can do what, and it will be stable for each player (in a game and looking at their own stats on the web). Behavior, habits, some go gorge some go fade, some are more versed on SG or mines, some are Exo... That is not "readable" with the actual skill factor number. It's why we have some completely crappy games where it was supposed to be even.
You won't see someone's aim going from 5% to 75% (Marine) other than using a 3rd party program. It may go from 15% to 25% maybe 30%... but it won't have a huge impact on the whole profile because of the other samples on skills. You just want to know what it will probably do in this game (most likely).
The only real random event that can happen is when someone change completely. He was always gorge, now he goes fade. We all know it's not happening in the real world. But we all know those 2 lifeforms have a different impact on the game.
Yes it means getting numbers we may not have but you seem to have plenty already.
Also there will be no "score" to grow. More like "you are doing this at the moment". And we gonna loose the pinkie size problem. The mood would improve on the servers.
The problem of balancing a NS2 game has never been about power (size), but more about how to use that power and IF it is used. And that's because of the last "if", the statistics are necessary. Just to get a better focused picture of the potential future game.
And if we could have a huddle system (as suggested before) the skill system could analyze an even better focused picture of what will happen in the next game. Simply because the players will "announce" what they will do. 4 Fades announced against rookies won't work right ? The system could signal this and maybe the alien players could decide not to have 3 fades to even things outs.
We won't avoid every bad game. But... at least we can have rookies playing alongside veteran for the greater good. Last time i checked ? Even the teams, still end up in a mess.
The detractors would say that it would make the same teams over and over. Well; this can be fixed easily if it were to happen. Storing information about the previous game isn't out of reach.
I strictly prefer a bunch of good games :
- filled with fun
- no size problem floating around ghost
- helping the rookies eventually
- playing more custom because people won't have anything to loose
- Reconcile Competitive players vs Public play
- with more girls (I guess they're hiding because of the mood other there)
by surrendering some randomness about team mixing....than going through the same crap that led to (and still do)
We don't need a powerful CDT/PT/UWE marketing stuff, just something in which we can be comfortable. Good candy doesn't need marketing. The players are the best marketing for a game.
Here is how a lot of factors correlate for players greater than 100 hours. In order from highest to lowest.
If you don't know what a correlation coeffecient is, read the following.
The main result of a correlation is called the correlation coefficient (or "r"). It ranges from -1.0 to +1.0. The closer r is to +1 or -1, the more closely the two variables are related.
If r is close to 0, it means there is no relationship between the variables. If r is positive, it means that as one variable gets larger the other gets larger. If r is negative it means that as one gets larger, the other gets smaller (often called an "inverse" correlation)
That's boring.
Ohh, and when mine reaches 0, can it go negative?
*reads*
*doesnt understand*
*awesomes*
For example, if you are the top ranking player on the team, you should theoretically be getting the highest score too. So if you don't get the percentage expected of your score's percentage, you get less points from a win and lose more from a loss. If you get more than expected for your score, you get more points from a win and less from a loss.
Of course some things that are important are not tracked by scores, but the advantage of this is that it may lead to a faster convergence, which helps mitigate the not enough games played some what.
It does not help with factors such as skill being relative to the server you play on, but lets not kid ourselves. No system will account for that. Again, I think you guys have overly high expectations of the hive system. It is best used as a rough gauge, not as a balancing tool. There are other factors outside of a player's skill evaluation that contribute to unbalanced games such as disparity in hive scores on the server that will not go away even with a perfect skill evaluation tool. Sometimes a bad mix of players is just that and not stacked teams.
Edit - Hive is borked, Nordic.
All people with negative scores played their last match around august of 2014. That's when we had the first implementation of hive skill, which was highly bugged. Some days later it was corrected and everyones skill was set to 1000 (but this value is only updated when you play your next match)
If I recall correctly, a few weeks later it was added the hard limit at 0 at the same time that it was decided to make rookies start with skill 0
Coming up with examples of individual situations where the skill system is wrong holds absolutely no weight.
Which is why I'd rather the base was something more interesting to a person, like SPM. If scoring could be better weighted and distributed within the game, it would make an accurate measure of your skill based on something interesting, rather than win/loss