How does Hive Skill work?
umdum9
London ON. Join Date: 2015-01-24 Member: 201013Members
London ON. Join Date: 2015-01-24 Member: 201013Members
In my last matched I scored about 7 deaths and 16 or 14 kills as well as countless assists. Despite all that my Hive Skill went down, was this because we lost the game? Or do you have to be impossibly good at the game getting 20 kills and only 2 deaths to even get your Hive Skill to move.
Comments
If the outcome is as the algorith predcts - nobodies score changes.
The idea is:
You do not get better score by having a sick kd (removes incentive from pubstomping). You do not get better score by being on the winning side of uneven team (removes incentive for stacking). You get better score by joining the weaker team and by moving it to victory (encourages you to actually fight towards the given winconditions).
If you would like an in depth explanation read here.
https://moultano.wordpress.com/2014/08/04/a-skill-ranking-system-for-natural-selection-2/
It is widely accepted that the system is broken.
Also I believe the gain or loss is proportonal to the difference between the average score on each team and your personal score. If you are at 2000 and win in a game that is 400 v 500 average, you probably won't get anything at all, even if you are on the underdog team.
Not at all. I think it's working pretty well. Yes it can still be improved in some ways but it's working well enough to be useful and considering the reactions when teams have too high average skill difference I'd say that it's widely accepted that the system works.
It is based on win/loss with multipliers like time passed playing and initial team skill difference. So there is a bunch of flaws that could be exploited and apparently are exploited. Some farming servers should be avoided. Not for those who are still there but for rookies.
I'm still waiting for the graphs i asked for. I believe we never gonna see that.
We could rename it but that wouldn't change the fact that ultimately it's a player "skill" rating as your ability to win rounds correlates undeniable in some way with your "skill".
Not everyone might like this approximation of "skill", but it's the commonly used one because any other mathematically model or description is ways to complex to handle to be used practically.
This. Could it use some improvements? Sure. Has anyone suggested a better alternative for a complete overhaul? Not really. There just aren't any unexploitable systems that give perfectly balanced games around.
People have too high expectations of the balancing system. As long as players on the server are not all similar in skill, even perfectly balanced teams on the surface can have disastrous results depending on who goes commander/gorge. Also, things like mistakes or practising unfamiliar lifeforms could turn the game very quickly. The hive skill is just a convenient scapegoat.
I don't agree with hiding the value. At least now we can point to and complain at how high/low the scores of certain players are. If you take that away and leave an invisible system, it will just foster distrust of the system. Its basically the difference between "It's alright except for these exceptions" and "The system is complete bullshit".
It has limitations. That is not the same as broken.
Simply put: You can't rate the individual skill by sampling team win/loose. It may be part of it but can't be all that matters.
Saying "any other mathematically model or description is way to complex to handle to be used practically." is a attempt to close the door on it. It's not that difficult and doesn't require PHDs to select proper samples. Even if it can be tricky to adjust it will be going in the right direction. I can't say that about the actual system which has come to be a farming contest. Totally the opposite of creating good games, good atmosphere and good grounds for rookies.
If you go to lichess.org (a relatively new popular open-source chess website) you can play a game (or review an old one) and request the site to analyze the game for you.
( NinjaEdit: It takes seconds to analyze an entire game, it's not time consuming at all. )
Here's an example of what that looks like: http://en.lichess.org/md0h3y4c/black#75
To the right, you can see each move recorded. The site will name first the general opening - then it will count all "inaccuracies", "mistakes" and "blunders" - and tell you which would have been the better move (or line to be more accurate). It also shows you exactly how much of an edge each player has at any point in time. These analysis' are a product of the stockfish engine - an engine that no human has beaten (afaik) without consulting an older iteration of stockfish.
My point with all of this is. If lichess wanted to, they certainly have the technology and the algorithms in place to accurately rank a players skill level based on the quality of each of their moves (A true skill evaluation). But they don't. No chess site does. They rank you on an elo system based on wins and losses - not unlike what we have for ns2 hive.
So a game, that is perfectly suited for the kind of system you promote - chooses NOT to. So why in the world, should a game with as many unpredictable variables, as NS2, use a system like that?
NinjaEdit#2: Also, who gives a crap if you have a particular skill - if you cannot exercise that skill to win a real game? Skill should translate directly to your abillity to win a game imo.
NS2 is not Chess though, it would take plenty of work to create a system that can rate what the best move would be in a 3D environment.
Yes, w/l alone is not a good reflection of skill. However, the alternative is what? How do you measure positioning, timing and other such things? Oh, and you also need it to be difficult to manipulate and discourage farming. It's not that difficult? How about you come up with an alternative so that we can tear it down? I think the hive system has flaws, but it really is the best available option.
I don't know what servers you play on but I don't see why the system creates a "farming contest". People have been stacking the winning team since NS1. Its not new to the hive system and they don't really get much hive skill/score/rating out of it after awhile.
Kappa
The current system does not allow for stacking and rewards every role in the team. You can be a gorge that never kills anybody but heals and builds stuff, or a commander, or some sort of support marine that is very good at welding exos.. or you are just very good at communicating and help that way to make your team win.
There are so many roles in ns2 critical to success, you can not ever model them all in a formula. So the least biased way to do is simply to look at win/loss. If you started to look at k/d ratios, suddenly nobody would want to go gorge and everybody would play it like a deathmatch game instead of a teamplay game.
Also, no matter how ns2 evolves and the meta game evolves, that system covers all roles. If what you do helps the team win, then your hive skill goes up. Simple as that.
You can be for example a great shooter. your K/D is amazing, but you always insult your team, you don't help your teammates and only think about your own k/d ratio. In that case, maybe you loose a lot, but end up always at the top of the scoreboard. Well thats covered with the current system.
So yeah, sure there are things to improve. Like separating marine and alien skill values. Maybe even commander skill values. Looking at how fast you loose or win points etc. But overall its a good system.
Uhm. Funny that you cut the quote right there - because that was my exact point if you read the rest of the post.
I was trying to illustrate, that even in the best case scenario where we CAN calculate all the best moves, we still rely on a system that ranks us based on w/l.
Hive skill system causes autism.
That isnt fair to people who dont have 4500 elo
Quoting you but it's a general answer.
Again;
ELO system has been chosen for chess as it ranks players alone. For other reason as well. You can find it on internet. Also there are some chess federations that will remove points if you don't play on a regular basis... If you leave and get back, say, 2 years later, you get the initial 1.2k score. Guess who was a player. Chess is a game that can be analyzed perfectly as you said. It's a finite space with pieces that move always the same (ex: knight L sequence), turn system etc...
It's quite different in shape from NS indeed, so why on earth should we have the same system as chess ? In NS you can have many things that will contradict that ELO system when the only thing considered is W/L of the entire team. Let's say:
- A player is AFK
- A player willingly ruining the game (reverse stacking) to make his buddies farm points
- Stackers
- Rookies Farmers
So you can be at your best, it doesn't matter. F4/reconnect/surrender get +1. GGAs pointed out by many players before, you can be the rock star of the game you won't hold it long if you get stacked with a pile of rookies. 2 times the score of any player doesn't mean you're suddenly duplicated 2 times. Stacking get +1. GG
Joining fast. Everybody knows that FET do NOT put everyone to RR and start to reassign players. We also know it does not consider AFK people... So the faster you join with your clan buddies the more chance you have to stay together. Stacking get +1. GG
Implementation is a issue. There's a snapshot of the AVG team skill at the beginning of the game. If anything happens during the game like people come and go; there should be another snapshot. I think you get the point. Stacking and/or Farming get +1. GG
Take group A, B, C. If A&B play 1000 games maybe you will have a rough indicator of the skill. As soon A plays with C the system will have to adjust. Let's say a Australian connects to an American server; the balance it crippled. Question : How many games before it get to a proper value for each player ??? 1000 ??? 10000 ? Nobody dared to answer me on that one except : "you know,it's a rough indicator"... yeah right... Imagine that Australian was a godlike NS player. As soon as people start to loose points seeing this guy ruling the place -> Stacking get +1. GG
You can farm to infinity (and beyond!!!). While your human capabilities are limited the ELO can go up to infinite. Yes; gamers are human and humans are limited but not only. The game is limited too. A skulk has a finite speed, same for marine etc. Many thing are already sampled. Aim % (NO! NOT K/D RATIO) ; demolition (damage on structure), etc. All these things are samplED and are clearly related to tasks that should be performed by any OK NS player. You can clearly add this to a skill score if you so desire a single number. But at least the W/L would not be the only thing considered. And that changes everything. As long as it is based on your human capabilities you can have a skill number that is far more stable than what we have. I don't really aim differently. I'm still the same bastard skulk that i ever was. A rookie can aim better with time and training, but it won't change that much once he get there. And most important it won't go toward infinity.
Smurfing ??? Should i explain ? hmmm ?
So Hive is exploitable in every way. But worse. It simply doesn't work. Who know a rookie (no smurfing & exploit involved) that went from 0 (or 1000) to +1.5K ? I doubt there is one. Even if he is good he will fall in one of the case mentioned before.
"This is the best we have"... NO! This is the only thing that was implemented. It's different. Being the best is when there is a competition.
Well... Hive and FET did something that is more armful than good to NS. It clearly encouraged stacking and bad behavior and now it's rookie farming. It led to some great steam comment on the community. Words like "Toxic" are accurate when i think about it. Sad story.
I'm not against FET and ranking systems as long as they are properly done/shaped and implemented. I clearly see people farming points on rookies lately and they probably want to keep it like that. I just don't have a problem with my dick which is clearly not the case seeing whats happening every week end when i drop by. I exclude girls playing NS (haaan he knows
PS: Also sorting the scoreboard by names would ultimately make those insecure boys so pissed off that you should do it right now. Cherry on top, baby... Cherry on top.
After all that value is only supposed to be used for the server browser skill icon and the force even teams vote. Thats it.
It strictly doesn't work at all. Not enough players. Not enough games, not enough the "right thing".
FET/Hive system look like this:
Every time i get to play, which isn't often. Somebody throw a FET vote (usually a rookie), while the clan people already joined a team. => bloodshed, rookie whine "stacked". No farmer say a word about it. Next round. Same routine.
It's dumb, FET act as a decoy to make them believe they gonna have a balanced game. Everyone who knows how to turn FET to its own advantage use it as a bait. 100% since 1 month, it's bad game after bad game for them. In fact everybody. The rookies don't learn anything, there is no challenge (knife-butter), nothing come out of it for anyone.
It's broken, accept it, change it, let's move on. FET/Hive is responsible for stacking and emptying a server in approximately 3 rounds for what i witnessed so far the last 2 Week Ends.
If one thing had to be listed on the Trello i would set it top priority. Bug and performances can wait a little. Making a better environment (all aspect of it; from FET to player retention, rookie training camp facility, etc) would clearly improve things for new comers. Since NS2 is out, i can't remember anything that was done specifically for rookies (and player retention by extent). I sorry to say but performance improvements didn't help player retention.
Right now the rookie story didn't changed a bit. Download, install, connect to a server, get obliterated, disconnect after a random number of games under 20 hours, uninstall. They all believed in FET. too bad...
Like I was saying, having a general indication of skill is useful game information. Who do you watch out for, who do you expect can get shit done on your own team, far better than just going in blind and getting blindsided by some 2500 you might have been able to handle with some strategy, the same strategy that would have been wasted on a lucky 1000. Knowing people by name helps, but some people like to change their name and score helps more regardless.
Not in terms of stacking. Hell no one really deliberately stacks this game. Normally the closest thing to a deliberate stack is when a couple of pro-level buddies want to be on the same team, but hive score has nothing to do with that. Hive score comes in when the teams are already 1500 vs 400 because the no-longer-green rookies all jumped on marine and nobody wants to face the natural stack to even it out. Can't really blame people for that, no one wants to be on the team that's OBVIOUSLY going to lose, and why should they anyway? Better to end that game quickly and try again, rather than get a couple competent people to switch just to drag out the inevitable turtle.
Anytime you tag a player with a value, it becomes personal. In their mind it's a reflection of their ability, even if technically it isn't. Higher means you're good, lower means you're not good. It gets attached to self-worth. It's that simple. Just look at all the posts from people saying 'I lost XX points in that game, WTF!'.
If you hide it, this won't happen anymore. It will also stop people doing things like leaving a team when they think it's going to lose to preserve their score. People are gaming the hive rating, instead of the game itself.
Hide it.