Matchmaking Algorithm for Normal Warzones

Maybe you need to start looking at yourself, and how you act? Also using ad hominem attacks and then crying about it are indications you know your position is very weak.

please try to keep it civil

This coming from the guy that insults people in damn near every comment he posts...

The irony... it's thick in here.

I'm trying to be constructive... the most constructive thing you've done all day is insulting others just to boost your own internet ego. You came into a thread on a random idea and just started spewing hate like Linda Blair.

PS... I haven't attacked anyone. You're the one walking around like the fat kid at recess.

Anyways, I'll take this elsewhere so we don't derail something that a constructive member of the forums created.

Since its quite simple mathematical procedures (for the servers to compute) theres no reason to even make the time limits and range of rating of the players that are invited to a WZ dependable on such things as number of players logged in on the server, number of players quing for WZ , the mean rating value of queing players etc.
That way the timelimits would be different on different servers at different time of day wich would almost guarantee a WZ within a bearable time. Of course it would also make those warzones more uneven when low on players I guess.

You are the only one that has used personal attacks, and you continue to do so. The amount of projection from your posts is stunning. Obviously people are going to react in kind. For some reason if I ever defend myself on these forums I get an infraction and if I insult the mods I just get banned, so I'm just going to ignore you from now on. There is absolutely no other way to deal with someone who stoops to such low levels as you do.

Maybe I missed it but I don't recall seeing you state a criteria by which to judge different scheduling algorithms. I want to know what your criteria is for judging whether one algorithm is better than another.

How can anyone claim one match making algorithm (e.g. yours) is better than another (e.g. the current) without first stating the criteria of measuring (comparing) them?

Presumably a metric for evaluating match making would look at the community as a whole (e.g. the entire set of regular WZs matches played on a server).

Let me make a space travel / rocket analogy: You have gone into wonderful detail about what fuel the rocket should use, how many stages it should have, materials for various components but I haven't heard what the mission is. Are we going to Mars? Jupiter? The Earth/Sun L4 point? Are we sending a robot or a human crew of 5 people?

In OP you say
I paraphrase the above as "OP wants to make a better match making algorithm" but that is way too vague a requirement to start with (IMO).

Your algorithm is not the goal, right? Some desired outcome is the goal and your algorithm is a means. Right? If you think I'm looking at this completely wrong that will be interesting to hear.

seeing as there are no algorithms for matchmaking at the moment, i have no criteria for judging different algorithms.

there is no matchmaking. i see people say that there is all the time, but there is not. whoever is available in the queue goes into a warzone; that is why it is so easy to game the system and get 2 4-man groups together into the same wz to form a full 8-man.

and your rocket analogy is sort of off. i state what the final objective of this is: PvP matchmaking that sorts by a skill-based algorithm, resulting in more competitive and fair PvP in warzones. im assuming the fuel is the actual algorithm?

i dont say i want to make a "better algorithm". again, there is no algorithm. do you really think if there was even the slightest kind of actual matchmaking in this game that bioware wouldnt have patted themselves on the back for it with a dev-blog feature???


as far as measuring potential other algorithms against each other, the criteria would be which algorithm results in more PvP matches where all participants are of relatively the same skill level. obviously the only way to test that is to either implement an algorithm to PTS, or to create a massively elaborate simulation that basically plays warzones.

as far as whether i think you are looking at it wrong, i dont. the end goal is not the algorithm itself; it is skill-based matchmaking for more fun/competitive/fair warzones. there are no algorithms to compare it to right now, but the best criteria for comparing potential algorithms is whichever algorithm is able to produce the most warzones where all participants are of the same relative skill level, while at the same time balancing time spent in the queue and the number of times warzones take place where there is a disparity of skill level between participants.

i think that is what you mean, you are still very cryptic.

and to be completely honest, this is not my job. if bioware wants to hire me ill design the whole freaking matchmaking system. but i dont think they do, so i am not going to try.

first off, thanks for the thoughtful reply!

If a match occurs that was not self scheduled then there is an automated match making algorithm. I think when you say "there is no algorithm right now" what you really mean is "there is no intelligence in the current algorithm".

Perhaps the current algorithm is "take the first 8 imps and the first 8 repubs and put them in a random WZ". That is an algorithm, albeit a simple one.

"more competitive and fair" sounds good. How do we turn this qualitative judgement into something measurable? I attempt to convert my idea of "competitive and fair" into a quantitative measure by saying:

a hutball match is competitive / fair if the final score is fairly close. A huttball match is uncompetitive if the final score is 6-0. (imagine similar statements for other WZs). More close matches is better, more one sided matches is worse

The above may be a bad measure, but little is gained by saying it is bad without proposing a better measure.

Of course one could say "trying to quantify competitve and fair is a hopeless task".

By your metric, if I put 4 strong players (e.g. premade from a top RWZ team) with 4 bads then I have done a "bad job match making" even if the match is a close one, right?

I place a higher priority on creating balanced matches (higher probability of a close score), even if the players are of widely disparate skills than on making teams more homogeneous (i.e. roughly same skill level).

I think that with the current player pool, trying to get 16 players who are roughly the same skill level is too difficult a problem. I recommend limiting the requirement to having more close matches and then providing the option of self schedule 8v8 matches (or even <n> v <m> matches where <n> and <m> are <= 8) as a safety value. I'm fine with the idea that self schedule matches give no rewards (experience or commendations). I expect that even with no rewards the feature would be used a lot. A top guild could challenge a weak new team to a 7v8 or 6v8. I suspect handicapped match would be more fun for both sides compared with the faceroll that frequently ensues with the current system or regular and rated WZ.

We definitely agree on the above! I too want more fun/competitive/fair matches. I also agree that having a team with a more homogeneous skill level is more fun that a widely disparate skill level.

In your original post you provide formula that you think will be good measures of individual player skill. Now how about providing a formula to measure whether the resulting matches were more or less fun / competitive and fair? After all, the parameters to your formula may need tuning. We can't tune your formula if we have no way to measure the results of each change, hence the desire for a metric of "fun/competitive/fair".

I apologize, I'm trying to be as clear as possible but, as you can see from my post above, we start out with very different assumptions and terminology.

With regards to requirements, I would break it down thus:

1. homogenous skill level of players in a match (you provided a metric for this)
2. more fun / competitive / fair matchs

Does #2 need a single metric or more than 1? If multiple (e.g. one for "fun" and another for "competitive/fair"), then how do we weight the importance of each? In my view, #1 is a proxy measure for "fun" that is not captured by the final score -- I agree a that 2-2 huttball where all 16 players are close in skill (be they good or bad) is more fun than a 2-2 huttball where the players skill ranges from 98th percentile to 5th percentile.

I claim that a 2-2 result with a wide range of skills is still more enjoyable than a 6-0 where 12 players are very good and 4 are terrible.

Thanks again for your patience and thoughtful replies.

ah, i see what you mean now. i definitely think that the algorithm i wrote here is just a first step toward better matchmaking. and yes, technically we have a "matchmaking algorithm" now, but imho it barely qualifies. when you say matchmaking, i think of how halo 2 started doing it; skill based.

i agree, quantifying "fair" and "fun" are tough to do. i also agree that close matches are probably more often than not considered more "fair" and "fun" than blow-outs.

i think that having several % increments for margin of victory, and then possibly trying to match that player against players with similar margins of victory. so players that routinely are on the winning side of blowouts would be matched with players also on the winning side of blowouts. likewise if you are on the losing side of blowouts, you are matched w/ other players that are often on the losing side of blowouts.

that would be the two ends of the spectrum, and then youd have everyone else somewhere in between. the matches should become more contested, and i would wager more "fair" and "fun"

correct me if im wrong, but i think that is sort-of what you are suggesting? if so, i like the idea.


lets say that we break the increments into >75%, >50%, >25% and >10%where the % represents the score differential. so it'd look like this:

Winning-----75----50----25----10----0----10----25----50----75----Losing

each increment would be weighted, with winning being positive weights and losing being negative. lets say >75% is 4, >50% is 3, >25% is 2 and >10% is 1. losing or winning determines +/-

every WZ outcome adds its weight to your overall weight. over time, blow-out victors would start to play each other, blow-out losers would start to play each other, and close-game players (the middle people) would keep playing each other.



there is a good chance that doesnt make sense as im kinda tired lol. and all the numbers are hypothetical, definitely open for changing.

my one concern with this kind of matchmaking is that it is still tied to winning and losing. in non-ranked warzones, it is very easy for a single player to make a mistake that the group just cant recover from. being punished/rewarded for the mistakes/accomplishments of others, especially people you did not overtly choose to play with, does not seem like a good system in my eyes.

perhaps a combination of the two, where you win/loss margin % is tabulated w/ your skill rating, and then from there the system tries to match up similar players? that sounds good in my head, but idk if this game has the population to sustain something like that so that it could produce the desired result on a regular basis

Honestly i think the only criteria that a matchmaking system needs to be measured against at this point is "does it make the current situation worse", if it doesn't, lets go for it. There is no reason, in my opinion, to come up with the perfect algorithm, we can very well start out with a faulty / imperfect one and improve it over time, and since it's all hidden from the user they don't even need to comunicate if / what they change. Most people will never realize that anything changed.

The trade off seems to be between quality of the matching and time spent in the queue, and right now the matching obviously prioritizes the later, probably for historic reasons when, back a year ago, many servers had hardly enough players to even get any warzones started. This showed clearly that there is only so much time people are willing to spend in a queue, if it takes too much people will give up and do something else. By the same token, the current abundancy of threads complaining about unbalanced matches, wether premades or gear is blamed, indicates that the quality of the matchmaking also can be "too" bad, so we need to find a balance between those too.

The proposed algo does allow for this balancing, so all bioware would have to do is make sure that the "time spent in queue" factor stays within an acceptable range and we're good to go, imo