No stats system can be perfect, I apprecaite that. Which is why I can't help but think a purely time based one is better. 0-10 hours. 0.5 players 10-100 hours. 1.0 players 100-1000 hours. 1.5 players 1000+ hours. 2.0 players Imo that'd be way way easier, and it just takes into account experience. That counts for so so much.
The system needs some major changes, and until those are made, I don't disagree with you, it has serious problems. Even considering balancing based on a wildly fluctuating inaccurate rank that can be easily manipulated is insane. There's a reason that it was NEVER (outside of DD at least) used for that. Can you stop complaining about an alpha concept that is going to be drastically changed? It's really not constructive. @Trickster: Now there's a suggestion (2nd constructive one so far). That should be a part of the rank calculation. In any case, this system is going to be private for a reason. It was not meant to be (at least, imo) and should never been public. I'm hoping that this system will eventually yield a smarter version of the number balancer that gives veteran players more freedom while keeping the game fun for everyone (veterans included).
Since it's beginning to look constructive again, I guess I'll break down and discuss more... "You don't understand the system" - Correct. You haven't asked me any questions about "intended result" so you can't possibly know what the intended results are. Every conclusion you have drawn is tainted because you are treating the results as intended results. You also choose to believe that refining it further is either beyond me or something I do not intend to do... again a lack of understanding. "You don't understand the difference between alphas and final releases" - Apparently correct, because you still have yet to aknowledge that this is an alpha and it's results are not going to be as intended in all cases. A final release I could stand by the data it provides, this I cannot. I suppose that means I should just give up right? Since after all I didn't get it perfect on my first try. "You're ignorant" - About this plugin, correct. The definition of ignorance according to Wiki is "the state in which one lacks knowledge, is unaware of something or chooses to subjectively ignore information" which fits your dialogue with this subject matter to a T. If something looks messed up, such as Dubee being rated rather poorly, there are two possible answers... either it's an error/shortcoming, or it's intentional. Once you know which of the two, you can continue discussion. You however have chosen to take any undesired results and proclaim them as what this plugin is going to do, making your critisism quite unconstructive and creating a panic mentality to all those who read and do not understand the full story. Never seen a "cold heart fact", and there are no cold hard facts when the results are unintentional other than the observation that there is still work to be done. Do you see the constant theme here yet? Your score being so low is exactly why this is only an alpha, clearly things are not tuned right, but that doesn't mean it can't be done, it means more data is needed is all (for example diminished return on HMG kills). Freaking out about how skilled an alpha plugin rates you won't help, this is probably the fifth time in this thread you've stated the obvious. I get it, you shouldn't be so poorly rated, I agree. I don't know how many ways I can say that this system is not without it's bugs, that's why we're running test numbers on servers and not letting it impose it's "sense of balance". Also it is vital, absolutely vital, that everyone understands the difference between rating and skill. I believe no one can properly measure "skill" in a game such as this because almost everyones opinion of skill is different to some degree, however I believe players can be rated. I believe a rating can be assessed by a players ability to win games, or in other words his ability to help a team win (of course all exceptions such as last second joiners need covered). One of the problems with the previous code was that I underestimated the amount of point whoring that would go on, so there's a massive gap between the point whores (and people who have done well against the point whores) and everyone else. I have made tweaks and will continue to monitor. Once it is something I feel I can call "intended" then we can talk about if it's doing a good job or not. For now, bugs/bad-data should be pointed out, but chicken-little-the-sky-is-falling mentality is not constructive at all... I wouldn't impose a system that isn't working on servers, and even if I tried, the admins would have the common sense to disable it. It's not the end of the world by any measure. The problem with a system such as this is that virtually everyone will be in the 1.5 or 2 category, then the system is just the numbers balancer again. That's not worth the trouble. Also everyone right now would start at 0 hours, so it would take months before it even had people broken out into 1.5 or 2. After the months of waithing, it would ramp up new players, but it would treat long time players who suck as if they were twice as good as said new players. I can see a modifier within a more complete system that utilizes this information, but relying on it alone would fail to produce different results than mp_autoteambalance already provides. Look, there are two schools of thought with the current balancer (mp_autoteambalance). Either you feel it is enough to create fun/competitive games, or you don't. If you feel it's adaquate, then I really don't know what to say. Stacked games happen all the time (back and forth games, my favorite kind, are a rarity), and it would be awesome to be able to join the losing team and try to turn it around even if they have one or two more players than the winning team. If you feel mp_autoteambalance is inadaquate, this is the kind of system that could replace it without forcing everyones team. I'm not claiming to be able to develop something perfect, or something that measures a players absolute "skill", but I know I can make something better than "make them join the team with less players". If you want to be constructive about the concept (not the tainted data), let's talk about: Is a system that puts more importance on who joined a team than just how many joined a team is a good idea at it's core? This is the alpha code of that concept, nothing more.
Stacking hasn't really been an issue for a long time. I just think its wise you work on the bigger issues rather then this. So winning gets you points? Like if you join a team and the enemy commander sinks himself do I get points? There is so many factors and things in this mod that really you should focus of developing and fixing issues like balance and adding features instead of this which has not been an issue really since fn198 was banned.
I don't think you got it, nobody is supposed to use ratings like this, the only way you're right about me being ignorant, not knowing the difference between alpha and release is when this plugin has no real purpose yet, and might not ever be used for balancing. I've been continiously commenting on using a rating in this fashion, or intending to make a balancer in this fashion. Your alpha or the results of the ratings have never been my concern. Infact I couldn't care less if it's full of bugs, the principle of the system is wrong, in too many ways. Maybe if you stopped using the word "balancer", "skill" and all the other words, I would have accepted my ignorance in the matter, but you've given a name to this, you've given it a subject, you've explained how it currently works, and infact you're closer to being ignorant as you don't seem to have replied to any effects a "balancer" like this might have outside of what you thought to be considered as positive or negative. You've pretty much managed to dodge every comment I've made, saying I'm not pinpoint accurate, or have no clue what I'm on about, as if I'm in a different dimension where the word "team balancer" and "rating" have a complete different meaning. This charade of saying "we're wrong" will only continue as long as you manage to not give information that matters. So you base my ignorance on this? I have never proclaimed your undesired results at what this "plugin" is going to do. I have proclaimed the plugin, no matter what the results are, as a flawed system. It's actually quite funny, if you look back a page ago, I never mentioned any "data" from your alpha, you then replied saying that the "current data" is different. Aren't you doing exactly the thing you're blaming me of? If I knew that you were claiming my criticism as unconstructive as I don't seem to have any idea about it's full story, then I would've asked you much sooner what it really is to continue being constructive. So please do tell me, what is this plugin going to do? Is it your intention to turn this into a balancer based on ratings? Are you intending to use ratings for anything else other than balancing teams? Read up a few pages back, I've only been discussing the concept.
I just want to point out that stacking hasn't been an issue because our clans aren't quite as active, and neither is the community.
No system can fully account for all skillful actions, but a good system will correctly rank the relative skill of players. I'm pretty sure anyone who can do actions not measurable by a plugin will also be considered skillful by the things it can measure (k/d, points, w/l, etc.) The one thing that bugs me about spawwn's system is that playtime = greater 'skill' (i.e. why you and many of us vets are ranked 'low' in his system). The skill number should be divided by the playing time (or number of matches played). This separates those who are truly skilled from those who just play alot. Oh, its still an issue. Its just people can't say 'OMG BSID STAX, QQ ON THE FORUMS' because there is no link between the players on the stacked team. I'd say 9/10 matches I've played are skill stacked.
So what do you suggest Dizzy, that the current balancer is fine? That's an honest question, I haven't seen your recomendation. It's not about stacks so much as current team restrictions. For example, why don't we disable mp_autoteambalance and just say "keep games fair"? Well, simple answer to me is that it would be a mess because people wouldn't pay enough attention and games would get lopsided all the time, it would be chaos... not to mention stacks that would purposely happen. It would allow me to join the losing team for some 15v13 comeback attempts though, which is exponentially more fun to me than remaining in spec or joining the winning team. One of the hardest parts in a system that lets you sometimes join the overpopulated team is that you have to manage it so as not to go overboard. One solution is to calculate an imperfect but relyable and faily consistent number for valuing a player in calculations. Another is to flag certain players as "autobalance immune" and revoke their immunity if they abuse it. If servers were willing to do the latter, my inspiration to create a plugin would vanish, but until I have a way to help the losing team even if they are up on players, I will persue a system. **EDIT** Scardybob's suggestion: Thing is currently (doesn't mean it couldn't be changed) system relys on finite points. Yes, Dubee is rated low right now (a separate issue) but if he joins a game with 9 players of ratings 600-800 and wins as many as he loses, he would average with them, meaning they would all be 550-750 or whatever by the time it was done. Again the previous version (still active version on ELC) for gathering stats has many exploits that have been addressed in the curren version but this is still true. The main reason I see dubee still being low is that he hasn't played at all. He probably ended on a rather bad note as well. The last version was too volitile.
A time based rating can't possibly hope to catch up with the years of experience some players have already logged. What's more, a quick learner with experience in other FPS' can be as good or better, with excellent aim, as a veteran in an hour or so. A points based one (or kills based one) will emphasise certain tasks and discredit other ones. For example, in both, the difficult task of tank hunting as grenadier is very poorly rated. However, if you are to get a "rough" look at how good people are, the points or kills one is going to be far more accurate than the time one can ever hope to be. Over a long period of time, this "rough" look will become an accurate cross-section of the skill of the people that play. It can't accomodate different play styles, such as dubees, right away, but after playing for a bit and his team winning more than half the time because of his input, he'll rise to where he should be.
The system could be used to suggest balance, but don't force it on pubbers. Giving more information is almost always a good idea, but not when it starts dictating how the game should turn out. Just giving players ranks for instance is done in 100's of games, but it's the players themselves that decide what the balance is. Look at any game with a lobby and ratings and you'll see that none of them force "balance". Infact this isn't even enforced in the most serious tournaments, if it were things like "noob farming" would never exist. There are games that filter out players on ranking, but the ranking is usually linear and based on simple things, and in the context of those games it makes sense (1v1 etc). I can understand if for some the current balancer is lacking, but it's the most forgiving kind of balancer, and of course there should be a balancer. I personally think the number balancer is fine, want to play with someone? Just wait till the teams even out. Want to stack? Well hey, you're a dick, but its your choice. I'm all for ratings, but they should be used as your personal advisor, not a governor.
I appreciate you taking a very serious and realistic approach. Now let's get to hammering this out then. One cvar is the allowable gap, I believe this will satisfy your concern of it dictating how a game should turn out. At a gap of 0 it's always making you join the lowest rated team... total nazi. At a gap of 20% it's fairly forgiving (one team has 2000 points (10 fresh players) other has 2400 points (8 fresh players, and two average vets @ 400points each), you could join either team. At 80% you're basically letting everyone run wild. I'm confident we can find a percent that works, I think 20-40% but do not yet know. Also note, since it's a perccent, as you get into higher valued players the freedom remains. So same example but everyone is more skilled... 4000 points (10 vets at 400) vs 4800 points (8 vets at 400 and 2 superstars at 800) you can still choose either team. @Blizzerd Formula is pretty simple for allowing or disallowing you joining a given team. if (Enemy team score)+(Enemy team score * gap) < (Friendly team score) you can't join. So enemy team score = 5000, gap = 25%... if the team you want to join (friendly team) has a combined rating of more than 6250, you wouldn't be able to join.
Then you'd still be balancing on a rating, which I'm against, and have explained in detail (and others have too) why it's far worse than a balancer that treats everyone equally. Instead of trying to come up with a different way to "balance", why not come up with a completely different approach to joining a team, like lets say changing the start phase of the game or whatever.
That's why I'm saying divide it by matches played. So the 'skill' would be calculated as Skill = Finite Points/Matches Played. Therefore, someone like Dubee who is skilled but doesn't play as much as someone like apwall who isn't but plays a lot more wouldn't call apwall the 'skilled' player and Dubee the 'unskilled' one. I'm not sure which games you play, but ones that don't enforce some type of 'balance', whether by numbers or skills, regularly turn into pub stomps. For the same reason some people enjoy griefing, some people enjoy steamrolling newbs. Unfortunately, it only takes a few of these types of people to cause a skill stack in Empires.
No one I've seen has explained (in detail or otherwise) that a system watching "who" and not just "how many" join is far worse than a system only watching "how many". Regardless, you have a suggestion here but lack detail. I'm all ears for alternatives. What do you mean change the start phase? We can try it, I have a feeling it leaves exploits though. I really like the idea that it makes you a shark to keep a high rating (every game diminishes your effective rating) but I think it would break with finite points. Let me give an example... 10 of us have ratings of 500 and 2 games played. 8 of us leave, the 2 play 5 games of 1v1 District to kill some time, they do "equally well" so their average is the same, but now our effective average is less than 100 when others rejoin. Now it reads that the 8 have effective ratings of 250 and the 2 have 71. Also, everyone approaches zero rather quickly unless many points are introduced all the time, making it no longer finite.
Sorry spawwn, I accidently pressed the wrong button and removed your post... I've reverted it, sorta. anyhow, heres my reply: I don't have a detailed suggestion, I think the balancer is fine, but if you really wish to change it, I'd say theres more ways than to try and rate players. And I'm not sure if you've formulated your reply in a way that makes a system that watches "who" as well as "how many" as a favorable system, you're trying to translate "who" into "how many" (correct me if I'm wrong). The question to that would be, is the way you're going to translate that valid? If your system intends to think of rating as "skill" then I'd say, its far from valid as it cannot draw that conclusion from some kills and wins. This I have explained and you should read back why. So yes, in this case, watching "who" and translating that in "how many" is worse than just looking at "how many" as the teams will always need numbers to achieve a balance. Where the least skilled player could still build a barracks right next to their base, if you didn't have that player, you couldn't build that barracks. It's that simple, this might not be true in all cases, but a rating is probably going to be less true than sheer numbers. So one good guy could kill 10 engineers behind a wall, but he couldn't kill 10 engis spread across the map in the same time. (no art of war quotes please) I think the best choice you can make here is drop the ball on trying to "balance" but instead improve the ratings themselves and assign those ratings to players, then allow them to give information on how it thinks the teams are balanced, and suggest players what team to join etc. Like every player has their rating next to their name, and it would then analyze which team would be best to join for a certain player and highlight the join button of that specific team. Or for instance always try and balance out to even teams (in numbers!) and suggest from the start, assuming the players that are unassigned will be playing that round, how the teams could be best split up. Unbalanced in numbers has a few nasty side effects and unless your system is practically non existant (everyone has an avarage rating) you'll have to deal with the possibility that your balancer will actually unbalance the game and possibly ruin games based on that unbalance. I think you should drop that part all together. Dividing the good players across teams is fine, but not evening out in numbers is bad, and even then the division shouldn't be forced. If you want to give certain players extra options and not necessarily allow the teams to be balanced, why not focus on those options over trying to benefit some players at the cost of others. If someone has a rating of 600 and is able to join any team he likes, the next person shouldn't be in a disadvantage. The way I see it, someone will always be in a disadvantage in this scenario, whether his rating is high or low.
Well, it's not really trying to convert "who" into "how many", but one could see it that way... it's trying to convert each player to a value (even the most worthless player having a positive number) and then only allow certain extremes in that rating between the teams. Teams do not always need even (or even near even) team size to achieve balance, this may be where we disagree... but listen to come EPIC campfire stories of 10v40 where the 10 won (Trickster has said many times it was among the most fun he's ever had). The simple answer is "if you don't want the numbers stacked against you, don't join the team with all the vets in a mixed skill server". You don't have to be number stacked against, you could join the other team and you would achieve balance in numbers. This gives you the freedom to do whichever you like (assuming it assigns rating effectively). A single player shouldn't be worth 10 other players, unless the single player is among the top 5 in Empires and the 10 are all beyond worthless... in which case sure... make it 2 of them for 2v20 and it is probably fair, or the 2 can choose not to join the same team, and it would be 11v11 with even teams. I guess my point here is that this plugin actually gives more choice, not less. No one is making you join the team with the other skilled players... the numbers balancer either does (because the skilled team has less players) which can create skill stacks with no counterweight, or it lets you do whatever you want if the numbers are even which may or may not screw up the "team skill" that we all long to keep balanced. Basically you seem to describe a balancer that just watches and suggests... That's already there, just turn off mp_autoteambalance so when a side gets skill stacked players have the option to outnumber them, and I'll set up a command for admins to watch the ratings as an advisor to punish people going beyond their subjective limits. The main reason I implemented a "kills result in rating changes" function is so the system doesn't create disbalance. If your team is winning you're likely to be killing the other team, it's one of the few ways to take ground. So the teams may have been "balanced" at the beginning, but 5 minutes in the ratings may have shifted, and now players would need to join the team dying the most. If by chance your team is losing yet somehow getting kills it could put more opposition on you... this could be a mistake but again it's very early stages, I'm looking for better ways of measuring how a game is going, but I really like this idea as well... if you turtle up and are providing stern resistance (lots of killing even though you are the losing team) then numbers would actually begin to stack against you, finishing the match sooner, which addresses a major problem with Empires (teams have lost 10-20 minutes before they actually lose). "If someone has a rating of 600 and is able to join any team he likes, the next person shouldn't be in a disadvantage. The way I see it, someone will always be in a disadvantage in this scenario, whether his rating is high or low." I don't know what you mean. Are you saying that because a 600 rating player joins NF you cannot also join NF immediately and that makes you unhappy? Because with the current balancer if they join NF you can't join NF until someone joins BE... so at that moment things are identical. Really it just means the more you care about freedom to choose, the longer you should wait to see the teams unfold. This is something the more skilled players should be doing anyway, to help do their part to keep games competitive. If the plugin is working as intended, the only times you would not see a numbers-balanced game is if people chose to skill stack a side, and they would get number-stacked against for their troubles... what would be wrong with that?
