Hive stats issue - NSL players
CRaZyCAT_Rus
Russia Join Date: 2013-10-31 Member: 188899Members, Reinforced - Shadow, WC 2013 - Shadow
Russia Join Date: 2013-10-31 Member: 188899Members, Reinforced - Shadow, WC 2013 - Shadow
Don't know if such thread was created earlier.
I would like to say that Hive ranking system doesn't work for the competive players on 70-80%. The problem that Hive take into account NSL games. There are a number of different factors:
1) Unlike public games, NSL games doesn't depend on players individual skill so much but largely depends on teamplay and NS2 as well as CompMod balance condition and also team's strategy.
2) NSL matches not based on even teams shuffle but it is often high skill level play between teams with different average skill factor.
For example, Handwashers (div1) playing with DWG (div1). They going to play 2 maps, that means 4 games or even 3 maps (6 games). Suppose that DWG is a little stronger than HW (in competive play minor skill differences cause significant advantage). Then there are a big chance to result DWG 4 - 0 HW or 3 - 1. Every player of HW will lose at least 150 skill points. If such PCWs or official matches will repeat several times a week that will put this players lower than plenty of public heroes in hive stats. But the real skill of this HW guys will be much higher.
3) NSL playing by teams with permanent line-up. But usually teams consist of different skill level players. If team tends to lose, good players will constantly lose points as well as bad players. If team tends to win because of some carry, bad players will constantly gain points.
4) There is tendency to alien wins on low divs (independently of the player's real skill) through specificity of 6v6.
I think it is obvious that NSL events flaws Hive rank system. In fact Hive shouldn't take statistics from servers with NSL mod enabled or NSL mod should disable server's statistics sending function.
I hope for feedback from CDT in this thread.
I would like to say that Hive ranking system doesn't work for the competive players on 70-80%. The problem that Hive take into account NSL games. There are a number of different factors:
1) Unlike public games, NSL games doesn't depend on players individual skill so much but largely depends on teamplay and NS2 as well as CompMod balance condition and also team's strategy.
2) NSL matches not based on even teams shuffle but it is often high skill level play between teams with different average skill factor.
For example, Handwashers (div1) playing with DWG (div1). They going to play 2 maps, that means 4 games or even 3 maps (6 games). Suppose that DWG is a little stronger than HW (in competive play minor skill differences cause significant advantage). Then there are a big chance to result DWG 4 - 0 HW or 3 - 1. Every player of HW will lose at least 150 skill points. If such PCWs or official matches will repeat several times a week that will put this players lower than plenty of public heroes in hive stats. But the real skill of this HW guys will be much higher.
3) NSL playing by teams with permanent line-up. But usually teams consist of different skill level players. If team tends to lose, good players will constantly lose points as well as bad players. If team tends to win because of some carry, bad players will constantly gain points.
4) There is tendency to alien wins on low divs (independently of the player's real skill) through specificity of 6v6.
I think it is obvious that NSL events flaws Hive rank system. In fact Hive shouldn't take statistics from servers with NSL mod enabled or NSL mod should disable server's statistics sending function.
I hope for feedback from CDT in this thread.
Comments
Public:
The Hive is comprehensively useless, it's doesn't show any accurate nor useful stats. It's only use at this date is for reluctant public players to worship it like it's some sort of god, if something contradicts the -accurate- Hive numbers, it's false and should be burned, the attitude is just cancerous. Public knowledge has and always will be the following: "Typically, if team A beats team B, team A is obviously stacked", "If player 1 is better than player 2, player 1 is obviously hacking".
Competitive:
You should know better than to take the damn Hive stats into consideration.
Beside all that, I'm partially sure I've read the Hive is being reworked, thrilling.
To answer your points:
1) public games mostly do or should too depend on teamplay. no teamplay = no win.
2) If that is so, the system counts probability of win and Skill Points gain/penalty badly and as such would be equally unfair to pub players too.
3) Yes, looking at the math, the system is putting value on a team as a whole. I have my doubts about it being effective for individuals even in pub games (where people change teams frequently). But for your case yes, all players in a (unchanging) team will tend to have the same calculated Skill Point value, regardless of their individual merit. All players will get as many points as a imaginary player with team's average Skill would deserve.
4) The original public article about the system is aware of that and proposes do disbalance the SP award in favor of the weaker race (based on alien vs marine win/lose history on hive). I don't know, if it really got implemented in the game.
@Lamb No, the Hive and force even teams works pretty well now. Some adjustments will be made but nothing major. One problem is that it takes too long for players who start with a score of 0, like you, to get their score up.
And I fully disagree
They depend on a one player carry as much as any MOBA system
Put 10 players pretty new at the game and one carry. Do the same in the other team and give 10 players with 5 new players and 5 public players, what team gonna win? I don't know what games you've experienced but it never happens this way. Of course it helps a lot but sometimes, the side with a quite good player doesn't win the round. Because teamplay is more than important and a team that reveals some kind of teamplay has better chance to win.
Well, that's just in theory, only a few public servers can claim to have players who like to put ahead teamplay rather than walking alone around the map, chasing skulks or whatever.
But to answer the question, I'm quite sure the NSL mod could do that. But wouldn't be better to ask NSL providers servers to do so by themselves? Or, the other solution would be to win your matches, all the time.
The other points are valid, however.
But for those who play only on competitive, they may have a score that isn't relevant on public. Meaning that they can get a really bad score despite the fact they're actually better than public and more regular players.
This thread looks pretty relevant to me.
@mattji104 I suppose one player can carry the game, lets call him lets say.... commander? Otherwise unless some player who outskills everyone on the server by factor of ten shows up, the teams of healthy demographic need to show some measure of teamplay to win.
Okay, let's say it in other words. I should have said that people who mainly play competitive.
You're a player who plays 65% of your games with your own NSL team and 35% of the other games on public servers.
Thus, it's relevant because their score will be clearly not suitable with what they really claim. Of course, you can also join a rookie server and stomp those playing on it but it seems obvious that is not a solution, it causes even more problems.
What I wanna raise here is that despite the Force Random Teams based on the Hive score, those that play competitive most of the time and decide to join a few public servers will necessarily cause unbalanced games. Why? Because those players get the same score than any players that do not play a lot or who lose most of their rounds. Makes no sense and only the competitive players who regularly play on public servers (70% vs. 30% with their team) can effectively be counted in regard of their real skill level. And no, that is not a random problem that happens sometimes. I haven't played plenty of public games lately but the Force Random Teams or any other votes based on the Hive stats may cause this issue.
The only solution I can see right now is to disable Hive stats on NSL modded servers.
Personally, i didn't play anything but scrims to go from 1000 to +2000, and i have no interest to play the game just to "rank up", but it's definately more helpful for the FET votes for me to have that +2k rating than skew the teams up even more by having 1k.
There is a good chance it won't lead to that many unbalanced teams. Some comp player will get more SP than he deserves, some less, cancelling each other out a bit. There is still pretty much 50 % chance comp/skilled players get separated to the two teams leading to sort of a balanced match. They should join opposite teams on their own to help that.
@Ots Exactly, the ONLY use of the system as implemented should be FET. Its usefullness is not in being good source of which individual is better than other, but in that it can balance teams somewhat better than random. I think the Skill Points were never ment to be read by humans. I guess at some point they decided for transparency sake to show the numbers. But making a public leaderboard from it was I think not a fortunate decision. It constantly and understandably leads to threads like this one.
The Skill Points have naturally a higher error the more exceptional you are too. If you are the 1 %er do not expect it to be extremely precise.
That's my Finnish nickname.
Doesn't work for what usage? Doesn't work as an indicator of skill? Doesn't work as a way to balance teams in pubs?
Hive scores are pretty much useless for anything right now. They're not a good indicator of skill because the pools of public and competitive players don't overlap enough to have scores be relevant... even if you believe that the only factor to being a "skilled" player is a win. The score is boiled down to too simple a factor.
After that, the biggest problem is that there are players at the top like NovoRei who admittedly game the system/hive score to overinflate their own rating. The system's easily exploitable and was exploited in the past and not reset.
All it's good for now is meaningless e-peen puffing and confusing people into wondering why forced even teams don't work. It's basically the same as using badges as a skill indicator.
why can't I write like this
... Why would you want to? I think he makes use of complex sentence structures... But it doesn't really help if the reader gets lost going round the bends he takes.
The point of writing is to communicate...
Why people usually don't bother posting in the forums.
Direct and simple enough?
Extremely effective?
Many players use FET and think it better than just randomize. Russian community collated NS2Stats and Hive shuffle recently and found Hive better. Sometimes on our servers people do re-shuffle 2-3 times to reach good team balance. Many people complaining about bad ranking of NSL-players. Why just agree that NSL games interferes Hive ranking and don't discussing the senseless of the theme.
Hey, I got turned around on some of your loops. And my reading comprehension is fairly decent. You gotta write like nobody else knows what you're thinking... And everybody else has English as a second language
1. Hive doesn't differentiate players who always play together. It treats them as one unit. If teams are shuffled all the time, this issue goes away, which isn't true for competitive games.
2. NSL players don't play against public players much.
Skill values are always relative to the people you play against. The hope is that those people shuffle enough, and play against enough other people that the skill values converge to something globally useful. For NSL players, you can decompose their skill values into 3 quantities.
1. The player's skill - the average skill of their team.
2. The average skill of their team - the average skill of nsl.
3. The average skill in nsl - the average skill of all players.
The skill that an NSL player will converge to is 1 + 2 + 3.
Hive varies in how well it estimates these quantities.
1. Not at all
2. Very well
3. OK.
Hive doesn't differentiate players within a team, so if players always play together, their skill will go up and down together. Between teams, hive works just like ELO, so will work just as well. Between groups of players, the scores will only be comparable if there are enough games between those players. The NSL players who also pub are in essence taking skill from the pub pool and adding it to the NSL pool to lift the average skill of all NSL players. Over time this should converge to the correct values between the two groups, but the rate depends on the overlap of the two groups.
@Zefram It's important to keep in mind the original motivation for basing the skill system only on wins and losses. Anything else creates a perverse incentive. The only accurate measure of whether a team performed well is that they won, and the only accurate measure of whether a player contributed to their team is that their team won. Anything else rewards something that can be abused (farming kills, building in a corner, etc..) Forcing a critical beacon is 0 points.
This idea is known as http://en.wikipedia.org/wiki/Goodhart's_law
"When a measure becomes a target, it ceases to be a good measure."
We had a system that rewarded players for kills, and it ruined many many games. Marines in the endgame would just spawn camp until the server died. Whatever complaints you have about Hive, at least it no longer kills servers just by existing.
If having accurate skill values for NSL at the player level rather than at the team level is important for some reason, the best way to accomplish that is to pug. If you want your skill level to be accurate relative to pubs, the best way to accomplish that is to pub.