Holistic Matchmaking
Radix
Join Date: 2005-01-10 Member: 34654Members, Constellation
Join Date: 2005-01-10 Member: 34654Members, Constellation
Alright I've been spamming General enough for 3 forum members, so here's my implementation of matchamking in a nutshell:
1. Servers are not required to implement a matchmaking system *at all*.
2. Servers that do implement it will have the following elements:
- Players are scored only on servers that are implementing the system.
- Kills and deaths are quantified into usable data by a surrounding "shell" of variables that should outline the major environmental elements of the game. Main points would be a) upgrades on each side, b) weapon/lifeform states on both the killer and victim, c) skill level of the killer and victim d) heuristics to detect serious spikes in gameplay such as a large death rate combined with no deaths on the opposing team signalling camping (or producing diminishing returns for any other similar situation as a beneficial side effect). This is of course not an exhaustive list, and one cannot be created until the major elements of gameplay are instituted.
- Kills and deaths are measured as usable data by an arbitrary number given to the server by the admins. This arbitrary number is used as the acid test for what players can play on the server. It is kept in check by the fact that the admins have no control over what players will fall into its range, but only the fact that the players who do will have access or be denied access based on the matchmaking algorithm built into the game.
3. A player who achieves enough rank to play on a server whose index matches that player's rank will lose as much access as he gains. In other words, when you gain 400 points of rank over the course of 2 weeks of learning and good performance, you have two options:
a) Unlock a higher level server
b) keep access to the lowest level servers in your subset
If the player chooses a, his lowest range of servers will be quantified by a query through a master server and he will be locked out of them. In return he will gain access to a new subset of more advanced servers, <!--coloro:lime--><span style="color:lime"><!--/coloro-->which will not go away once he unlocks them even if he performs poorly at first<!--colorc--></span><!--/colorc-->.
If the player chooses b, nothing happens, but he may choose a at any time.
A player will get a preview of what servers will be lost and which will be gained before the decision is made.
----------------------------------------------------------
This system should be reasonably effective at quantifying skill. Obviously servers will differ, as will times of the day, player moods, the level of allergies in a given subnet and that subnet's respective impact on the skill environment as a whole, the collective unconscious, the alignment of the planets, but none of those factors, I think, are large enough to throw out matchmaking with a system such as the one proposed above. Despite the fact that the sytem is noisy, it's my opinion that it gives a decent approximation of skill which, being a fairly ephemeral thing to quantify on a macroscopic level, doesn't need total precision to begin with, and that a reasonable approximation of skill used to determine who to pair together for the best games, is better than throwing one's hands up in the air because the system cannot be ironed out completely without accepting some statistical noise.
1. Servers are not required to implement a matchmaking system *at all*.
2. Servers that do implement it will have the following elements:
- Players are scored only on servers that are implementing the system.
- Kills and deaths are quantified into usable data by a surrounding "shell" of variables that should outline the major environmental elements of the game. Main points would be a) upgrades on each side, b) weapon/lifeform states on both the killer and victim, c) skill level of the killer and victim d) heuristics to detect serious spikes in gameplay such as a large death rate combined with no deaths on the opposing team signalling camping (or producing diminishing returns for any other similar situation as a beneficial side effect). This is of course not an exhaustive list, and one cannot be created until the major elements of gameplay are instituted.
- Kills and deaths are measured as usable data by an arbitrary number given to the server by the admins. This arbitrary number is used as the acid test for what players can play on the server. It is kept in check by the fact that the admins have no control over what players will fall into its range, but only the fact that the players who do will have access or be denied access based on the matchmaking algorithm built into the game.
3. A player who achieves enough rank to play on a server whose index matches that player's rank will lose as much access as he gains. In other words, when you gain 400 points of rank over the course of 2 weeks of learning and good performance, you have two options:
a) Unlock a higher level server
b) keep access to the lowest level servers in your subset
If the player chooses a, his lowest range of servers will be quantified by a query through a master server and he will be locked out of them. In return he will gain access to a new subset of more advanced servers, <!--coloro:lime--><span style="color:lime"><!--/coloro-->which will not go away once he unlocks them even if he performs poorly at first<!--colorc--></span><!--/colorc-->.
If the player chooses b, nothing happens, but he may choose a at any time.
A player will get a preview of what servers will be lost and which will be gained before the decision is made.
----------------------------------------------------------
This system should be reasonably effective at quantifying skill. Obviously servers will differ, as will times of the day, player moods, the level of allergies in a given subnet and that subnet's respective impact on the skill environment as a whole, the collective unconscious, the alignment of the planets, but none of those factors, I think, are large enough to throw out matchmaking with a system such as the one proposed above. Despite the fact that the sytem is noisy, it's my opinion that it gives a decent approximation of skill which, being a fairly ephemeral thing to quantify on a macroscopic level, doesn't need total precision to begin with, and that a reasonable approximation of skill used to determine who to pair together for the best games, is better than throwing one's hands up in the air because the system cannot be ironed out completely without accepting some statistical noise.
Comments
I do agree with your idea, however.
<!--quoteo--><div class='quotetop'>QUOTE</div><div class='quotemain'><!--quotec-->What exactly does 'holistic' mean to you?<!--QuoteEnd--></div><!--QuoteEEnd-->
<!--quoteo--><div class='quotetop'>QUOTE</div><div class='quotemain'><!--quotec-->–adjective
1. incorporating the concept of holism in theory or practice: holistic psychology.
2. identifying with principles of holism in a system of therapeutics, esp. one considered outside the mainstream of scientific medicine, as naturopathy or chiropractic, and usually involving nutritional measures.<!--QuoteEnd--></div><!--QuoteEEnd-->
He's using the first definition, not the original definition of holism as a religion.
<!--quoteo--><div class='quotetop'>QUOTE</div><div class='quotemain'><!--quotec-->the theory that the parts of any whole cannot exist and cannot be understood except in their relation to the whole; "holism holds that the whole is greater than the sum of its parts"; "holistic theory has been applied to ecology and language and mental states"<!--QuoteEnd--></div><!--QuoteEEnd-->
So traditionally it's been applied to "nature", but more recently it's been used as a philosophy when looking at systems.
Where holism applies to <b>radix</b>'s post is that he's not looking at matchmaking by itself and just slapping it in there, he's trying to predict potential pitfalls based on what we know about the game and design the system around that. Since we don't know much about the game "More Holistic Matchmaking" would probably be more descriptive. Regardless, the usage is correct as far as I can see.
A much more effective way is to allow server admins to self-label their server and for the developers to create a user-reserve slot system that will favor individuals that will follow the rules of the server. Since nearly every server in the community has a forum base, it would be up to the developers to decide who gets a reserve slot, and how to label their server. 3 recommended labels I would have are Competitive, Casual, and Open.
Wasting time on all of these stats system will just cause rifts in the community as well provide an illusion of effectiveness.
BINGO
While I think you could be right about this being inferior compared with other solutions that might have the same net effect with less effort, I can't help but to call some of your posts over-reactions.
For example, in your initial point, are you missing the fact that this would be an optional system, and that because of that element, most server admins who wanted to create a casual atmosphere would simply turn it off? This makes sense because it would increase their playerbase. In theory you can create a thought experiment where it's all over, and most of the servers are using it, but in reality the system would only likely apply to the best of the elite servers, and the greenest newb-friendly "starter servers". Is this, in your opinion, such a dangerous thing?
Here's yet another problem that I don't believe was raised yet: abusive server operators. People who run a server and either manipulate their rank somewhat subtly, for example by kicking certain players, or who outright cheat. We would need super-admins with powers to control what servers have rights to be a part of the system; would they ban the server from rankings? What if it was supposedly just an out-of-line server admin? Court is in session! Oh the drama. There are so many issues, some of which we haven't even thought of yet, which would utterly break the system. And this is a factor which would break the system. It's not just a matter of statistical fluctuations or problems which could be accounted for by a simple calculation.
BTW, bidding starts now on Ebay for my "unlocked" NS2 account! Comes with plush gorgie (lovingly used) ! All servers accessible!
I feel you though, Radix. It is soo tempting to want a ranking system that would not be broken. It would be beautiful if it could be done. But I don't think it's feasible.
It is a fair point, and I'm fairly sure that even I like the difficulty-icons implementation better at this point. Sometimes the best thing that can come out of a post like this is that the topic being proposed isn't the best answer.
Claiming that ranked servers would 'segregate the community' (what community exactly?) is completely ridiculous.
If anything, having 'favourite' servers to play on, having servers where you have reserve status, would be 'segregating' the (greater) community (players of NS2).
I think what a lot of people fail to understand is that we, here on the forum, represent a minority (of potential or past players of NS/2) NOT a majority. NOT everyone is looking for little community servers where they know everyone and have a bit of celebrity status. Most people (of which there seems to be a minority represented on these forums) will just want to play a good game.
I think a more likely possibility is that the developers would spend a lot of time on a system to balance servers based on stats, only to have hardly any server use it because it would break up communities. It would mean that many programming hours would be wasted, pretty much how nobody ever wanted to use phermones when they were introduced (but much worse because the stat-system would be more difficult to make).
If you really wanted to protect green newb-friendly servers, just supply the power to the newbs to get rid of the griefers who come in to ruin their fun. An accessible and functional voting system would take care of that. Likewise, the same voting system along with reserved slots would take care of unprepared players who disrupt "elite" games.
If the majority are going to be playing without the use of the skill-balancing system, then it is logical to look for a simpler solution to the problem of griefers or completely inept players. RSS, voting, and icons are the answer.
Here's why: Most casuals will accept that they are casual players. However, they will reject the concept of being a "newb". In short, very few server operators would label their server "newb" as compared to "casual".
On the flip side, I've known many NS players that would consider themselves pro, until it was time to back up their skills. Most of these players were professional in rhetoric rather than skill.
This is why I recommend the 3 self-ratings: Casual, Competitive, and Open. If a server community wants to welcome all players, leave it "open". If server operators want to restrict their clientelle, they can do that as well with the other labels combined with reserve slots to reward players for playing the style that server desires.
Newb has negative connotations, regardless of interpretation. Pro has negative connotations and leads to an inflated feeling of self importance. Fun can be interpreted as carte blanche to nob around rather than play.
I like where Firewater is going with the naming convention, but I think we should be striving for better than "Competitive" and "Casual". Casual servers can be just as competitive as any other, even if the calibre of their tactics are in doubt.
Maybe some in-universe slang could be cooked up for it?
Newb has negative connotations, regardless of interpretation. Pro has negative connotations and leads to an inflated feeling of self importance. Fun can be interpreted as carte blanche to nob around rather than play.
I like where Firewater is going with the naming convention, but I think we should be striving for better than "Competitive" and "Casual". Casual servers can be just as competitive as any other, even if the calibre of their tactics are in doubt.
Maybe some in-universe slang could be cooked up for it?<!--QuoteEnd--></div><!--QuoteEEnd-->
Actually I think the terms are rather fitting. Competetive doesn't have to mean clan owned or even that you're playing on a high level, just "we play serious here".
There's no reason, of course, that server operators can't name their servers "Casual" or "Competitive"; but these are just naming conventions. They could call it "Rob's Horror House of Pure Pwnage" or something equally ridiculous if they wanted to.
I reiterate - casual servers can be just as competitive as any other.
Its a minor naming quibble. One could just as easily label them as "TSA" ("Casual") or "Frontiersmen" ("Competitive").
Labeling the style of play of the server basically says that a player of that particular style is welcome. There can be competitive games on a casual server, and laid back games on a competitive server, but with the appropriate labeling, players of either style should not be banned (theoretically) as long they are playing in an appropriately labeled server.
Labeling a server by skill level has no concrete definition. What would be the difference between a low, intermediate, and high skill server? Would the results in one server generalize to another of the same "skill level"? If not, how does labeling by skill accomplish anything? The only way to label by skill is to create a stats system that would contain a stats system that will not be able to contain the validity to do so.
Labeling by playstyle allows the user to have an idea of what kind of environment they are going to expect. Casual servers would cater to players have a laid back approach to the game, and just play for the thrill of playing. Competitive servers would cater to the players that play to win, no matter what. Open servers (i.e. non-labeled) would welcome all players.
By keeping the labels based on playstyle as opposed to an unquantifiable measurement of skill allows the player to choose based on their personal preference, without the negative social effects that would exist with a skill level ranking. Server labels also acts as a fair warning as to what the server would expect from each individual player.
These severs will then appeal to the people who are not looking for a community, and can just label their servers open. Any server that is going to have reserve slots are going to need to be well admined this is true. I mean if the servers are truly "setup and forget" then why should it matter who plays on them? Those server companies are usually just setting up test servers so that people can test out the server and decide whether or not to purchase one.
If people want to setup servers and not have a community based around them that is fine. NS though is more social than the typical FPS thus it would be beneficial for the server's life to have a community around it.
To exclude this idea because of major server companies want to setup some servers is NOT a smart move.
Again, I must state that "casual" is a wide ranging term, and some will see it as "you don't automatically F4 when someone goofs" and others will see it as "gorge rush or leave".
I do wholly agree that there should be an indicator of how seriously people are taking the game, absolutely, but I feel that "casual" and "competitive" are just not great terms.
With regards to "setup and forget", the success of an online game can be correlated to the number of servers up. I mean, anyone could start a CO server and then just leave it to run. You just cant do that with NS Classic. Leaving them flagged as "open" is not an ideal situation, because there will generally be anarchy. This means the public perception of NS2 would be flawed, and that decreases the potential player base.
Which is why it DOES matter who plays on them.
I strongly feel that NS2 should be accessible and easy to run. Relying on a good admin or a good community should enhance the experience, and not instead be essential for the experience.
Its not a question of EXCLUDING the idea of server labels and RSlots, but a question of how we blend these together to guarantee a good game on the majority of servers.
Great! Because like I've said, it's a great feature and I have no problem with it making it into the game.
However, like I've said, it's no solution to the issues that Rated Servers seeks to address.
However, like I've said, it's no solution to the issues that Rated Servers seeks to address.<!--QuoteEnd--></div><!--QuoteEEnd-->
Rated servers based on what? Your flawed rating system idea?
Rated servers <b>seeks to</b> address certain issues - and the general feeling on the forums is that it's inadequate. RSS on the other hand, is completely <b>irrelevant</b> in regards to those issues.
Rated servers <b>seeks to</b> address certain issues - and the general feeling on the forums is that it's inadequate. RSS on the other hand, is completely <b>irrelevant</b> in regards to those issues.<!--QuoteEnd--></div><!--QuoteEEnd-->
Your rating system isn't broken for what reason now?
<!--quoteo--><div class='quotetop'>QUOTE</div><div class='quotemain'><!--quotec-->Rated servers <b>seeks to</b> address certain issues - and the general feeling on the forums is that it (ranked servers) is inadequate (to addressing those issues).<!--QuoteEnd--></div><!--QuoteEEnd-->
And somehow you don't seem to understand what 'seeks to' means.
Both globally rated and globally ranked servers fail to address the question of unenjoyable play. The best solution is local. This does not necessarily mean handing the responsibility to admins - it DOES mean that stats are restricted to the server they are earned on, which means comparative skill is much easier to calculate, and high tier players don't have to worry about outcompeting X thousand other players in order just ACCESS their favourite relaxation server.
At a local level, you're only looking to compete with another 60 or so players, which should be easy enough for any player. Easier still if the server runs RSlots and you're a good sportsman.
Previous Post: This post has been edited by Harimau: Apr 23 2008, 02:50 PM
Post after that: Apr 29 2008, 11:21 PM
You may retract that accusation.
I added the things in brackets () for clarification when I quoted myself, since it seemed that you had a lack of understanding.
I thought that was obvious.
-
Now, "segregation" is not a flaw in the ranked servers system - it is simply at odds with a certain <b>opinion</b>; coincidentally that is a <b>prevalent</b> opinion on these forums. (One which I, for one, do not wholly share.)
As for difficulties in quantifying skill or difficulties in quantifying other such things, I still personally think it's workable - when you have knowledge of the game flow and mechanics.
But, the problem now is whether it's worthwhile - addressing the issue is certainly worthwhile, of course - but this particular implementation may or may not be. And also given the prevalence of the aforementioned opinion, I'm no longer prepared to defend this particular idea. Instead I'm moving my support to other possible solutions (RSS not being a solution).
Also, consider that this game is free at the moment. Now, I'm a Constie, but my money was a donation to see development continue. It is not an equivalent to "buying" the game. When NS2 hits, which will be retail as far as I can see, then you have people paying money, and paying players don't want to be kicked out of NORMAL servers just because of their ability or lack thereof.
Rating and ranking cannot work on a global level, not when it is connected to balance. Yes, you certainly could "rank" people using a variant on the Co system, but you simply cannot take that system and translate it into a door pass for Classic.
I agree that these sorts or solutions should be discussed, because for every one we can discount, it leaves more time to focus on the remainder.