Gagis, on 19 November 2022 - 04:09 PM, said:
You fundamentally misunderstand how statistics works.
What team you get is the major random component of MWO matches. The random part is the same for all players and equally likely to put you in the weak team as to the strong team. This means the random component cancels out when you average the statistics. Games with large luck based components can be statistically analyzed just fine.
There are games with even larger luck based components than MWO, such as Magic the Gathering, where w/l ratios of players are extremely consistent and predict their tournament performance accurately.
i do know statistics and probability, hell i did one of my highschool science fair projects on Luck (mind you that was many many moons ago). keep in mind my sample size was small (100 people). i had each person role a pair of dice (standard 6 siders) 300 times breaking each person's rolls into 100, 200 and 300 data blocks (i had concluded before the experiment that after about 300 rolls everyone would have the same spread, i was wrong). in any statistic there are outliers. my experiment was geared as a game where the higher the roll the better (i think the person who got the highest average roll overall got a candy bar or something stupid like that, it doesn't matter). i found that after 300 rolls each very few of those 100 people had the same exact array of numbers rolled. there were indeed those that rolled low more consistently and those that rolled high. (to take any kind of rolling method out of the equation i used a hand made dice tower that you dropped the dice into and everyone dropped the dice in at the same orientation, this was before portable computers were as common as they are now so i didn't have access to something like say a Random number generator). the one thing that turned out as i expected was the more rolls one made the closer the differences in the arrays were but it never became exactly the same, there were always those that deviated considerably from the expected average.
now to take this into the perspective of the discussion you can conclude that there will be people who get the luck of the draw on the MM more often than others. also factor in groups that throw a wrench into any such equation and you have have my point. now for those people that play constantly (say playing every day) the numbers might even out more but it will never be the same. now for those people who play say a dozen or so games a day but only on say the weekend it is far more probably that you will get more outliers. (hell i have had entire days where all i get are teams that nascar or scatter with the inevitable constant losses no matter how well i play individually. but rarely do i get win streaks (though honestly this could be the mental bias of one remembering negative results more readily than positive))
now the current system the MM uses if i remember right to determine what tier you are in is based at its core on a scale that is based on how well you did in the match (using MS) compared to the other players on the team (mind you this in itself is flawed as in the current system damage and kills far outweigh more tactical contributions like scouting, TAG, objective capture, AMS and other non damage factors). the MM then takes where you are on this scale (going up if your performance is better than average on your team going down if you are worse) and then trying to place you in a group of as closely ranked players as possible. now what the game does not factor in is your mech selection (hence matches where one team has 5 assaults and the other has zero) and it doesn't seem to know how to properly deal with groups in QP. now with the low population the MM has to reach outside your Tier to fill a match sometimes considerably. now if we didn't have groups it would try to put an equal number from each tier on each team with as close a distribution of mech classes as possible but with groups this gets tossed out the window as the MM doesn't to calculate groups properly.
so the end goal of a good MM is to put people up against equally skilled players with equal team composition (as best it can since there is no way to factor builds into the mix). the end goal being as many "enjoyable" matches as possible. the trouble with using W/L as your main factor in determining a "good" match is in itself flawed as you can have a loss that was still a fun match and a win that is boring as hell. a match that ends in a 0-12 stomp isn't all that fun for either team really (unless you are one of those that enjoys seal clubbing) where an 11-12 match even if you are on the losing side is more often than not a fun match. ok its gonna suck for those first couple of people killed no matter what but someone has to die first.
EDIT: i forgot to even mention your own build selection and how it works on the map that ends up selected (say picking your LRM mech only to get Solaris, or a brawler and getting say Polar (though Polar is much better in that regard than it once was). there are certain maps that come up more than others but you can never predict what maps you can pick from when you select your mech (this is one of the reason i prefer mid-range builds and back-up weapons on LRM boats but thats for another conversation)
Edited by VeeOt Dragon, 20 November 2022 - 07:04 AM.
2
