- last post: 01.01.0001 12:00 AM PDT
Posted by: TwoFaced1680
Posted by: squomp
Posted by: El KafungusIf you're worried that you're not getting your fair shake in the ranking system, look at the last 50 games that you've played in that hopper. (If you haven't played 50 games, play more. The system needs more data.)
Take your win/loss average. If it's 65% or above, and you've been in that playlist for a while, you've probably increased in level 2 or more times over the last 50 games (or you will in the next 20 or so). Otherwise you are going to be hovering around the same level or dropping in rank.
http://www.bungie.net/stats/PlayerStatsHalo3.aspx?player=vDo% 20Matt&ctl00_mainContent_bnetpgl_recentgamesChangePage=6
This team only plays together. They are something like 137-0. They hit level 14 towards the top of page 6. Please, explain why they have been stuck at level 14, for quite some time.I think I counted 66 straight wins at level 14 with no prior losses...Well that's just messed up :) Someone mentioned something a while back about the TrueSkill system having some difficulty with people who only play together. Not sure if that's true. Perhaps the OP can help explain that one. I'd send the OP a message just in case this gets lost in the shuffle. I'm curiose about the answer.
Sure, I can explain.
Here's the problem with these guys. They don't lose. In fact, as far back as I can see, they NEVER have lost a game in team slayer. Nor in a matchmaking hopper.
Without a loss, the system doesn't have anything to gauge your performance against. Therefore, the Sigma value in that hopper stays REALLY high. Now the problem with that is, that the game might think these guys are level 30-40. However, there's where the conservative ranking system kicks in.
See, the game is unsure of their rank, and therefore ranks them lower. It is SO unsure, that their rank is underrepresented to the tune of 20 levels.
Mathematically, I've seen hypothetical Sigma values in excess of 7 from Microsoft. So when fit in to the Rank = Mu - (K * Sigma) equation, where K is assumed to be three (again, information from microsoft), that means that, say, 14 = 38 - (3 * 8). See how their Mu is actually 38? If their Sigma was smaller, they'd be ranked in the high 30's using those numbers. But they haven't lost. So, their Sigma might actually be HIGHER than 8. Which leads us to another problem. That Mu value is still being put into the experience calculations. So is the Sigma. And you get less "experience" from winning against a team with a lower Mu rating, and still less when one or both of the teams involved have a high Sigma value. And they're going to lose a HELL of a lot of experience when they finally DO lose.
This is why these unbroken teams have trouble ranking up. I would say that winning 80% of the time will actually make you rank up FASTER than winning 100% of the time. And winning 100% of the time is usually why a team stays unbroken. Yes, it's a flaw in the system, but designing whole systems around unbeaten teams makes NO SENSE. MOST extremely talented players are going to have a ratio of around .75-.9 wins per games played. You CANNOT design a meaningful and efficient system that makes room for people who NEVER lose. Because sooner or later, they will lose. Yes, the system punishes people who don't ever lose, but everyone loses eventually. At least once.
However, this is a perfect example of what I wanted to do myself, research wise. This is a team with no outside variables. They have all played in the same team from the start, with NO other ranked games than what they have played together. And if K/D ratio really DID matter inside matches, this guy (3.35 K/D) would be ranked lower and this guy (3.91 K/D) would be ranked higher. So we finally have experimental data to show that in-game performance truly doesn't count in the Trueskill system. Just whether or not your team wins.
I'll have to do some research into what actual mechanism there is, if any, to "punish" a group that plays only in one unbroken team. I doubt it, but it should be tested one way or another. Things are busy right now, and my birthday is Sunday (Finally a Weekend Birthday!) it might take me a while. But the model that I researched covers 98% of all outcomes so far. And if anyone wants to help with the research, PM me on the forums. But I need GOOD players (preferably even better than I, so we can test the whole lossless thing) with a matchmaking playlist that they've NEVER played in before on that gamertag, and their hours of availability either have to match mine (generally 3-7 central on weekdays, weekends by appointment), and/or they need to be dedicated to a scientifically rigorous style of play. Which means logging outcomes, and other unfun stuff. And they have to realize that we'll have to play something like 100 or more games in each hopper we test before I'm satisfied with the data set.
Oh, if you have a set of data you'd like to have analyzed, and that you think I'd like to analyze, please PM me on the forums with a set of links to the information.
Oh, and please don't message me in-game, just do it on the forums. I might be receiving a lot of PMs, and the forums are faster.