Can someone PLEASE explain this to me a little more simply? I just don't get it. Microsoft TrueSkill System
They made it so the skill placements would be bell curve like. Meaning most people in the middle, and less going out in each direction. No f-ing clue how they get it to do that though.
So it keeps track of matches and games you play on XBL and assumes your skill level based on the outcomes of those matches/games? After reading this page 5-6 times, I gather that the system ranks you on all games and places you in a match that will put you against similar ranked players or an average of similar ranked players. Is that correct or did i miss something?
I heard about it somewhere and went to read about it to see if I could figure out why my teammates are so bad sometimes. I'm not sure if it's utilized in MNC though. Hence off-topic.
Sounds too technical for me but if it did match you with people the who are around the same skill level as yourself it would be pretty cool, do you know long this has been going for?
Okay, this is easy for anyone at uni. I'm surprised that this hasn't been implemented before. Now, take the first image; The dotted line is where the computer thinks the global average skill level is. Before I get ahead of myself, I should explain the bell curve. In statistics, and probability, bell curves are everything. Everything random in real life fits the bell curve; intelligence, height, shoe size, the size of something else ;p It's safe to say that gaming skills follow this as well (it's really safe, people have written mathematical proofs on it). It's quite safe to say that the majority of gamers have some skill, and that those with more skill are less common, and those with lots of skill are even rarer. At the same time, people with less skill are also less common (just ignore the nubs online you've seen that would make you question that). Now, back to Mu. Since everyone has a skill, everyone fits somewhere on the global bell curve. Most people will be around the middle somewhere. Where you are on the curve is *called Mu (it's the backwards u shaped character). That's calculated on all your games played, if you've played one game it could have been an exceedingly good game, or an exceedingly bad one, or it could have been average. The problem is, the computer can't tell how good you are from one game. It takes dozens, or hundreds of games to get a good idea of that. This is where Sigma (the character that looks like a p) comes in. Sigma is how the computer measures how reliable the data it has is. It calculates Sigma by grabbing Mu (your average) then finding the difference between that and every individual match you play, adding that all up. Then dividing that by the number of matches minus one. Say you play three games, you score 1, 5, 9 kills. You average is Code: 1 + 5 + 9 / 3 It's pretty clear that the average is five. I chose those numbers to be easy to understand. Now the Sigma takes a little more effort. And I've conveniently made several small errors here (Churro to the first person to spot them all) to simplify things. Code: 5 - 1 = 4 5 - 5 = 0 9 - 5 = 4 (4 + 0 + 4) / (3 -1) = 4 So if you play three games, like above, you'd have an average of 5 with a variance of 4 (yeah, that makes sense). What happens when you play 50 games where you have 10 games scoring nine points, 30 games scoring ten points, and 9 games scoring eleven points, and one game where the phone rang and went AFK, scoring one point. Doesn't take a maths genius to see that the average is close to ten. But what is the variance? I'll do the maths for you, and say the variance is 0.59. That's a much smaller number. In fact, the smaller the number, the more reliable the average is. *I think Mu is actually mislabelled on the graph. There's two Mu's that they could be talking about, but they rest of the article makes me think the graph is slightly wrong.
BulletMagnet, I really appreciate you taking the time to explain that. That's exactly what I wanted to know. Thank you.
TL;DR: basically there's two numbers, your rating, and the uncertainty about your rating. The more you play, the tighter the uncertainty is.
It's not clear from this but trueskill uses matches from the game you are playing, not from all games you have played. I.e. a ranked halo match will use your halo based trueskill.
The biggest problem with this is that there is absolutely no video game with a scoring system that can reliably show the 'true' skill of a player. The exception to this is straight deathmatches, but nobody cares about straight deathmatches because they're awful in every almost* every game ever. *every shooter, but not, say, Bomberman
