[QUOTE=Mustang;456244]I’ve not experienced the CS:GO system but for me equally skilled teams are more imporant that equally skilled individuals.
So if every player has a “skill rating” and you are searching for a game on your own then it uses your rating and makes sure the sum of the ratings of everyone on one team are close to the sum of the ratings of everyone on the opposing team. That is to say one team might be 8 people all with 100 rating (total 800), where as the other team might be guys with 25, 50, 75, 100, 100, 125, 150 and 175 (total 800). Of course there could be a system that does try to match closely skilled players initially, then quickly expands the net if not enough are available, but always does a team “shuffle” before the match begins to keep things even. Of course even the above example isn’t the best because ideally each team would have a couple of the very high skill, a couple medium skill, a couple low skill, etc.
If you’re searching for a game as a group there is an appropriate multiplier applied depending on the number of people in the group. For example if everyone in the world had rating of 100 then two guys that aren’t in a group would have a combined rating of 200, whereas two guys in a group would be 220. And instead of 300 three guys in a group would be 340. Et cetera.
Of course groups should be kept together through any team “shuffles”, so the matchmaking system should account for this. Which is when it gets tricky trying to maintain even total team rating (800 in the example above).
Something along these lines would work well I think… on paper at least.
Then there is the whole debate on how to calculate a players skill, but that’s for another thread I feel. :D[/QUOTE]
In topic of skill distribution - let’s agree on some resonable mid-way conclusion: the teams shall best have corresponding skill distribution patterns.
As the matchmaking with friends and clans and skill and ping being found important for decisions of groups chosen, there might be a slight problem with laws of addition in that groups - one problem; there also must be(assuming ideal accuracy of choice in terms of balance = algorithm can find the best balanced set of groups) downfall of the accuracy due to the fact, the algorithm cannot match friends to different servers - they have to play together, so only in rare case when that was the case in the best balance set - it will not lose quality in terms of balance. Thus adding additional conditions will make things only worse or the same(in a best case scenario only).
That is why I oppose mixed matchmaking based on multiple factors. Simply when something is designed for everything - it’s often good for nothing. When it comes for the problem we spoke of - we can expect it to output worse balance, rarely it will output the same and never ever anything better. Unpinning the assumption of an ideal algorithm - we still can say, that if the algorithm makes sense at all, we still expect the results to be worse when the “facebook” factor is involved.
Then we have the addition laws in a group of players, considering skill. Is that the same to have 1 highly skilled and 4 unskilled against 5 semi-skilled, so the average is the same? What kind of average shall it be then? It’s hard to prove it, maybe statistically, but I’ve remember games when playing against higher skilled ppl we won due to some drastic change of tactics… There is a difference between two teams that avarages to the same arithmetically - in CS:GO universe 1 skilled can be in one place, so whole team’s skill is defending only one spot. It’s then enough to send two guys to check main routes to objectives and go behind the one that survived. It’s easier to defend some places than to gain the control of them. Therefore teams are equal when they have a corresponding distributions of skill. Trivially - when they are; that is the same in other words for our context here as we only care about skill level and number of ppl when comparing :D.
And finally about measuring the skill - I’ve read the thread you are alluding to, and I believe I don’t know which way is the best. In CS:GO it’s simply win-lose, with whom(ranks) and against whom. In the thread there were pretty many different propositions.