PDA

View Full Version : Matchmaking System based on Stats



FreeBorN
21-01-2011, 11:56 PM
Dear RGC Community & Sir Rogers,

I would like to present a suggestion for all of you:
-Rank based matchmaking system-

The ONLY goal I would like the suggestion to accomplish:
-Better Game Experience for Everyone-

I will try to keep this post Updated with any new idea in the thread or with my own ideas
Last update: 2011.01.21

Disclaimer:
If there is anything about the following suggestion that (in your opinion) makes game experience worse, let us (the whole community & Sir Rogers) know it, so that aspect of the suggestion will be changed, I do NOT insist on any feature that turns out to be disadvantageous game experience-wise.

Therefore please feel free to post here!
I would like to collect every constructive idea about a potential Rank based matchmaking in this post. I am very ambitious, and will not stop with this suggestion until it's a very close to perfect concept so it deserves implementation. Furthermore I don't want this to be implemented only in months. I would be very happy if there was test-phase no later than in a few weeks.





Why Does RGC need a stat based matchmaking system?

Little history
As I mostly play the EU Public room (DR. Community and PlayDota inhouse rarely has any games on)
the reason below is only about EU Public, however I take the freedom to deduce that the situation is pretty much the same in other non-EU rooms.

EU Public room used to be a lot better when vast majority of its inhabitants were players who were fed up with the game quality that is usually present in Garena pubs.
These people, were ambitious high-skilled players, who are still here in EU Public, however mixed with a lot of other guys who happened to download the RGC client but they play the same way as Garena pubs usually go.
The client itself is very easy to use, which is a nice thing, but it also attracts people who has no competitive aspect in their game-play.


No need to hate, but to separate
I DO NOT condemn people who aren't serious players, these players:
-usually just auto attack on the lane
-don't estimate the threat of the opposing lane -> blatantly stay in stun/disable range -> feed
-don't recognize important situations on the (mini)map
-just want to play without much of an effort
-doesn't matter for them if their ISP is providing a LowQuality line and they keep DC-ing
-or they just simply leave too many times

Different Motivations, Different Results
There are players who constantly seek to improve, While
There are players who just want to have fun.
Skill level does not necessarily depend on which of the above 2 you aim to do.
Some players will perform a lot better even if they just wanna have fun, but some won't be as good even if they try.

Don't Mix Casuals with Serious players!
These different players playing with or against each other is not good:
-it creates unrealistically high points -> people who care will team up, while ignorant casuals will just loose
-different player in the same team can easily generate anger and flame, which is not good for the community.

Conclusion about separation
Players have to be separated based on how well they perform, and based on their own expectations towards themselves and their allies.





HOW TO do it then?

Beta Phase, Test Motivation -> More Points
There should be a separate room for this system [EU Pub Match Beta] as long as the system isn't stable.

Players should be motivated to try this new Matchmaking, therefore the games in the matchmaking system should reward more points than playing in the usual EU Public room. Everybody should be automatically vouched to the matchmaking room, and an easily noticeable Message in the EU Pub should prompt players to try the new system.

NOTE: POINT GAINS BELOW ONLY APPLY UNTIL IT'S IN BETA PHASE
As motivation players should gain more points per games : 7-10 (or whatever the community thinks is fair)
They should only loose 2 points if they loose.
Should be penalized the same -10 if they quit or DC.

Stats...
Your usual RGC stats show how important it is for you to stay in a game and try to win no matter what happens. I am saying that currently the best and simplest way to measure a players performance is based on the win & loss based current RGC statistics.

Trying to create an individual stat measurement is a VERY HARD & COMPLEX problem, implementing it demands way more time than Sir Rogers has. Individual stats are also hard to implement without inherently motivating players AGAINST team-play. Leave that project for the future.


Conclusion about stats
The present stat system has to be used for a new stat based matchmaking. No-one wants his stats deleted I guess.
However: The way people earn points has to be refined.
But: Penalizing leaves with a considerable amount of points (currently 10) has to be kept.





Basics concepts of the new stats based matchmaking
-It will have to separate players based on stat whenever it's possible (enough signing players are present)
-It will have to reward playing against same or higher stats
-It should only give minor stat gains when playing versus considerably lower stats


How AND When to separate players, The Adaptive Matchmaking
The more players who want to play the same time, the easier to make even matches. Therefore matchmaking of RGC has to be adaptive

What is this adaptive thing??? -> Explanation
RGC should never restrict 10 players from playing a DotA match if they are the ONLY ones on-line and they are of different stats, but if there are 1000 players available, they will be separated based on stats. Adaptive = it will depend on the number of players!

Friends forming Parties
There will always be friends who would like to play together. Separating them makes no sense, PERIOD.

Let them invite each other into a play-party (2 to 5 players) when the leader signs or un-signs the whole party signs or unsignes. The slots for the party would be reserved.

The stat of party would be the average of the party.

The leader of the party would be default captain.

In a Sentinel / Scourge team the captain will always be the leader with the bigger party. If there is a tie: 2+2+1, then the leader of the party with the higher party stat average will be the captain.





The Matchmaking itslef
The sign button would no longer directly put you into a game. This future sign would put you or your party on a short 30-90 sec search time (the exact time can be shorter, its just an idea).

IN the search time RGC would optimalize a match and when the optimal match has been found it would send it to only those 10 players that have been selected for the match, and it would say something like "Your match has been hosted"

Criteria of Optimization
-Don't try to find a perfect match (keep the search time low)
-Minimal stat variance amongst the players in the match
-Secondarily: try to include similar sized sub-parties on Sentinel/Scourge size, example: try to avoid 1+1+1+1+1 vs 4+1 or even worse 1+1+1+1+1 vs 5.

Briefly about a sophisticated Match optimization
There could be certain evaluation methods, such as how important is sub-team size vs stats difference.
Example: It's better to make a 2+3 vs 1+4 match with a stats average difference of 50, like 1700 vs 1650 average teams.
While it's a lot worse to make a 5 vs 1+1+1+1+1 with a stat difference of 10 Like 1680 vs 1690.

I know it can get very complex to make sophisticated evaluation methods, but still it's not impossible to evaluate bigger sub-teams -> better co-operation within the team -> virtual bonus stats for those teams when deciding the who will play who.





The new Reward System

Let's Normalize those points
I have no exact evaluation in mind but here are the following guidelines:

-People with lower points in the winning team should earn a bit more than their team-mates with high points

-Teams with lower stats (remember that matchmaking can't be perfect) should earn more if they win than teams with the higher stats.

Same for losses:

-People with lower points in the loosing team should loose less points, while their higher point mates should loose a bit more.

-Team with lower stats should loose less points in a match while higher stats teams should loose a bit more.

Thresholds
There should be either soft or hard thresholds in the Reward system.

There should be a limit on the following things:
- very low point player won't loose any points for match-loss if his points differ greatly from the average.
- very high point player won't win any further points if his point greatly differ from the average.
- EXCEPTION: Leavers should ALWAYS be penalized with - points. However there could be a separate threshold for DC-ing (like you cannot go under 1400 if your net sucks). But you can go indefinitely low if you Quit.

Leave or DC?
As long as there is no reconnection system leaving and Disconnecting should be separately penalized:
-Give more penalty to the player who quits Warcarft 3 (voluntary leave) (-15)
-Give less penalty to the player who totally looses connection (-10)

You can say: "But people can <<Plug>> to cheat"
I say: " So what, there will be sure some ridiculous pluggers, who will loose less points by plugging but majority won't even know, and at least Quitters will be penalized more.





Desired Results
Players would finally play with the same kinds of players as themselves, whenever there are enough players available.

Still if there are very few players on-line, RGC would still make matches, but matches wouldn't be so perfectly optimized -> people will not be able to gain or loose lots of points thanks to the new reward system in imbalanced matches.

Casual low point players would play with casual low points players, and as they are not so serious they won't really care if someone quits or will have to quite because of other real-life activities, or if they loose.

But people who invest a lot of effort into the game will play exciting and close matches with other persistent players, and the quality of those games will be a lot better!

NeC.madchen
23-01-2011, 06:03 PM
Even if this was suggested before and rejected, because rogers dont have time, this is realy good text. Even if rogers dont have time, when it comes, this should be a base for building system.

aystro
28-01-2011, 01:09 PM
Even if this was suggested before and rejected, because rogers dont have time, this is realy good text. Even if rogers dont have time, when it comes, this should be a base for building system.

If he doesn't have time then why is this bot still online? With all the respect I have to say that from all the bots/clients I have played on, this has to be the one with the biggest amount of amateurs/ feeders/ dc's . It's impossible to play seriously. Sure you can stay until the end of the game but you already know after 5 min who is going to win and who will be the mega feeder(s) that will ruin the game for the rest.
I actually stopped playing here after being here for a couple of weeks, because it's unbearable. The point system is lacking purpose other than the initial impression of the players skill. I do think that it is incredibly important to change the system if this bot wants to grow in class and in quantity.

Mephisto
28-01-2011, 07:19 PM
If he doesn't have time then why is this bot still online? With all the respect I have to say that from all the bots/clients I have played on, this has to be the one with the biggest amount of amateurs/ feeders/ dc's . It's impossible to play seriously. Sure you can stay until the end of the game but you already know after 5 min who is going to win and who will be the mega feeder(s) that will ruin the game for the rest.
I actually stopped playing here after being here for a couple of weeks, because it's unbearable. The point system is lacking purpose other than the initial impression of the players skill. I do think that it is incredibly important to change the system if this bot wants to grow in class and in quantity.

let me tell you that roger work alone here, and he doesn't have time cz he is working on another thing that's more important, fixing bugs, doing his best to reduce disconnect problems, anti hacking, he wants to get as close to perfection as possible, he have some great plans for reducing the global leaver and skill difference problem, but it requires so much work that he can't do it alone. 10x for ur understanding and we really appreciate your support guys.


If he doesn't have time then why is this bot still online?

let me answer on this question: Bots are online for simple reason "roger working on it" that's why he doesn't have time
and i will tell u another thing that RGC will stay online cz i think there is many ppl who adore it and they will never leave it.