[quote][b]Posted by:[/b] VinoVeritas [quote]I've heard this argument many, many times, but I do not agree with your rationale. The stated goal of the ranking system has always been to match players up with other players of similar skill, to create more fun games. [/quote] Correct. In the case of TEAM games, this means matching you up with teams of similar skill. If you're continually performing "well" but your team is never winning, you're going to have a low skill so that you get matched up with other teams that also lose very often.[/quote] This is the basis of your whole argument, but it's wrong. Even Microsoft and Truskill's designers admit that the objective is to "infer your individual skill level from team game outcomes". The system IS supposed to match individual players of similar skill, not just teams. If what you're saying were true then Bungie would only allow full teams to enter ranked games. Do you really think that 8-player ranked games are never comprised of 8 random players? It happens all the time. The system is trying to create two evenly matched teams, but how could it possibly do that if it doesn't know the INDIVIDUAL SKILL of all the players involved? Microsoft also admits that the system cannot accurately assess your INDIVIDUAL SKILL if you always play with the same team. You are expected to switch who you play with, and your rank goes with you. It is not a team rank. To suggest that your rank is not an individual assessment of skill is ridiculous. Also, I am not playing with "my team". If I go in randomly, it is pairing me with a different team every time. [quote]It makes perfect sense... regardless of your individual skill if you cannot figure out how to work well with a team and come out victorious, then it makes no sense to rank you up based on your individual performance… you will just end up being matched with better and better teams, meanwhile never actually figuring out what it takes for your team to win. This defeats the purpose of the ranking system.[/quote] This is so stupid. Do you ever play Halo? Sometimes I go, literally, +25 in a game of CTF. The other night I stood at the wall on Highground and killed the entire other team twice over, with the flag already on the beach side of the wall. But even then, none of my teammates could stay alive long enough to finish running the flag down to the beach. I kept asking them, I tried to be nice, but there is nothing I can do to compensate for that kind of incompetence. Finally, I tried to grab the flag myself before it returned (I was the one who had already gotten it that far), but I was killed by the other team, who had just literally respawned for the SECOND time after I killed them all. We tied, despite all my efforts, and I reinforced the system's "certainty" that I am truly a level 8. I am clearly not a level 8. If you want to assume that I'm some kind of noob and that I really belong at level 8, you are free to do so, but I doubt you really understand the game that well. Like you said, it's a TEAM game, and I can't singlehandedly compensate for a team consisting of me and three noobs consistently enough to win games, especially in objectives. Despite what you say, the system IS supposed to match me with players of a similar skill. It is NOT doing that in my (real) example. Also, I don't object to the fact that winning is a prerequisite to level up. I think that it should be a major consideration. But there are ways to require a win to level up that would still use individual performance as a mitigating factor. [quote] [quote]I can only influence the outcome of a game so much, no matter how well I play. It's ridiculous that the system couldn't recognize that I deserved to be playing at a higher rank. The games were no fun for me at all because no one else in most of the games besides me knew what they were doing. And I couldn't even find anyone decent to party up with because ALL the players in every game sucked.[/quote] You don't deserve to be playing at a higher rank. Your team is losing and that's not a successful round regardless of your individual performance. You seem to be missing the point. You weren't "dominating" at all because part of "dominating" a team game is having a good team where all members do well and work together properly to achieve victory. If you don't have a reliable team that can win consistently, you "deserve" to continue to play with other teams who cannot win consistently (i.e. are low ranked). [/quote] I already said that I am not playing with the same team every game. Again, if this were really MY account I would get some people from my H2 friend's list and recruit some of the good players I've met on these forums and I would quickly level up. But playing my friend's account made me realize what a disadvantage you are at if you don't already have that friends list. You should still be able to enjoy a reasonably competitive ranked game even if you don't go in with a party. Again, I say this not as someone who is obsessed with rank as status, but as someone who just wants to have competitive, evenly matched games for the sake of greater enjoyment and challenge. This system largely fails at accomplishing that goal. Your whole argument rests on the false assumption that rank is not an individual assessment. It clearly is. [Edited on 11.02.2007 8:16 PM PDT]

[quote]I've heard this argument many, many times, but I do not agree with your rationale. The stated goal of the ranking system has always been to match players up with other players of similar skill, to create more fun games. [/quote] Correct. In the case of TEAM games, this means matching you up with teams of similar skill. If you're continually performing "well" but your team is never winning, you're going to have a low skill so that you get matched up with other teams that also lose very often. It makes perfect sense... regardless of your individual skill if you cannot figure out how to work well with a team and come out victorious, then it makes no sense to rank you up based on your individual performance… you will just end up being matched with better and better teams, meanwhile never actually figuring out what it takes for your team to win. This defeats the purpose of the ranking system. [quote] I can only influence the outcome of a game so much, no matter how well I play. It's ridiculous that the system couldn't recognize that I deserved to be playing at a higher rank. The games were no fun for me at all because no one else in most of the games besides me knew what they were doing. And I couldn't even find anyone decent to party up with because ALL the players in every game sucked.[/quote] You don't deserve to be playing at a higher rank. Your team is losing and that's not a successful round regardless of your individual performance. You seem to be missing the point. You weren't "dominating" at all because part of "dominating" a team game is having a good team where all members do well and work together properly to achieve victory. If you don't have a reliable team that can win consistently, you "deserve" to continue to play with other teams who cannot win consistently (i.e. are low ranked).

[quote][b]Posted by:[/b] VinoVeritas [quote]In other words, the only real assessment the system makes of your performance in the game is whether your TEAM won or lost. Which is really a totally inaccurate, random way to assess the individual player's performance[/quote] Uhh, right. Your skill in a TEAM-based game should NOT be based on individual performance. It's a TEAM game. The primary determining factor of success or failure of your strategy is whether or not your TEAM is victorious, not how many headshots you got or what your K/D was on the map. If you want to be ranked based on INDIVIDUAL performance, play an INDIVIDUAL play list.[/quote] I've heard this argument many, many times, but I do not agree with your rationale. The stated goal of the ranking system has always been to match players up with other players of similar skill, to create more fun games. The current system works in the broadest sense, in that as you increase in rank the GENERAL level of competition does increase. But the system really does an extremely poor job of accurately assessing the real skill level of individual players, because it never really looks at how you play. Your logic makes sense to me if I only ever play in a full party. But I feel that we should be able to play ranked games even if we don't have friends online to party up with (because I really just want competitive matches, I am not obsessed with my rank). And the ranking system does not work if that's what you want to do. Example: I don't even own H3 yet, but I've been playing for the last two days on my neighbor's borrowed 360 and on his account. He never played ranked Team Objective so I was playing it on his account with a fresh start. I only ever got to a level 8 in that playlist, but I was absolutely dominating every single game I played. (In case you think I just suck, I got him up to a 31 in Team Slayer in a matter of hours, and that was without a party as well). Literally going at least +10 in every game, and frequently +20 or more. I was the only player in almost every game who had experience playing objective games (from Halo 2), and so besides dominating in the kill department, I clearly had the best understanding of tactics to control the objective. But I couldn't get a single other player on my team who knew what they were doing, and so, despite my influence, almost every game of Team Objective I played ended in a tie. Either 0-0, 1-1, or 2-2. I can only influence the outcome of a game so much, no matter how well I play. It's ridiculous that the system couldn't recognize that I deserved to be playing at a higher rank. The games were no fun for me at all because no one else in most of the games besides me knew what they were doing. And I couldn't even find anyone decent to party up with because ALL the players in every game sucked. So you can say "well just find a team to party up with" and if it was my own account, I would have done just that. But should that really be necessary to get a good game? I feel that if the system really worked well you could still get reasonably competitive games without playing with a party. But as it stands right now, the ranking system is only a very crude assessment of a player's REAL skill, as I've illustrated with my example above. The players it matches you up with in a given game will always vary widely in terms of their real skill for this reason. This is my experience with the system so far. Play a randomly matched game of Team Slayer, and you will almost always have some players going way positive and some going WAY negative. If everyone were really at the same skill level that wouldn't happen. [Edited on 11.02.2007 2:18 PM PDT]

A few things you need to remember: - Data for Trueskill calculus is submitted only at the end of a game - Trueskill works on teams only, so if you're in a solo game with 16 players, there are actually 16 teams. - If the game does not find people of your relative skill , it WILL eventually find people for you to play with. You can get a private versus a general in theory, but because the population of this game, it's highly unlikely. - Apart form the skill and uncertainty values the developers have a lot more options to submit to the system. The most notable ones are 1) [b]an estimate match quality factor[/b] - which goes can alter the amount of skill gained or lost. This is used when matchups are really screwed. So, if you get the private vs general matchup you can get a game where the maximum amount of skill gained is 0 because of the wide skill spread. 2) [b]a percentual time participation factor[/b] - this means, if you disconnect halfway during the game, you might get only 50% of the skill gain (or loss). On the other hand, your skill will count only as 50% towards the team total. These 2 factors are optional for developers when submitting games, and I don't know if Halo actually uses them. However, since H3 wants to be a very competitive and fair, it most likely does.

[quote]In other words, the only real assessment the system makes of your performance in the game is whether your TEAM won or lost. Which is really a totally inaccurate, random way to assess the individual player's performance[/quote] Uhh, right. Your skill in a TEAM-based game should NOT be based on individual performance. It's a TEAM game. The primary determining factor of success or failure of your strategy is whether or not your TEAM is victorious, not how many headshots you got or what your K/D was on the map. If you want to be ranked based on INDIVIDUAL performance, play an INDIVIDUAL play list. [Edited on 11.02.2007 12:19 PM PDT]

Copy and paste from a post by Creepy Al in one of the groups on here, where he points to another thread here on this forum. [quote] Easy explaination by Achronos (cut/paste from [url=http://www.bungie.net/Forums/posts.aspx?postID=13992631&postRepeater1-p=1#13996166] THIS POST:[/url] "Look, this is getting silly. Things are working correctly, they just don't work like they did in Halo 2. Here's a few things to remember. 1. Your individual performance during a team game doesn't matter - if you lost, you lost, it doesn't matter how well you did personally. 2. Rank progression is NOT linear. 3. Rank progression is based on consistency and the rank of your opponents. That is, if you always lose to higher level opponents, and win most but still lose some to equal level opponents, you won't go up that match. 4. If you lose to someone lower level than you, your rank is going to suffer more. Winning against lower level players is expected, and will not assist you in going up. 5. I cannot say this enough - just winning more than you lose isn't going to get you rank. It is but one requirement. As I said, if you're not beating higher leveled players consistently, you don't go up a level because you haven't earned it. 6. Rank will initially change often, but once you get to a certain point, the system becomes statistically certain about your level in a playlist, and just playing games and winning is no longer enough to go up - you have to consistently win against higher leveled players and not lose to lower level players. 7. Please, PLEASE stop assuming things aren't working because it doesn't work how it did in Halo 2. Playing and winning only guarentees your EXP goes up, it doesn't guarentee your skill level goes up. That depends on who you are playing."[/quote]

Some really good reading to be found in this thread and the provided links - thanks guys!

'''''' [Edited on 10.29.2007 10:46 AM PDT]

[quote]imagine trying to program this ranking system into a computer :|[/quote] It's possible. That's what we have done, in fact :) Ralf Herbrich [GT: SniperEye] Microsoft Research Ltd.

[quote]The IGN article is wrong. It is contradicted by all the Microsoft articles listed in this thread. The guy who wrote the IGN article misunderstood what Microsoft's Truskill designers are saying the same way you are misunderstanding it. [/quote] It's true that the IGN article is not as specific as all the other material we released. If in doubt, consider the technical material we released the official details. Ralf Herbrich [GT: SniperEye] Microsoft Research Ltd.

[quote]Believe me, if I had their e-mail address, I would write them immedieately.[/quote] Just send your questions to [email]trueskil@microsoft.com[/email] and you reach the "architects" of TrueSkill directly. Ralf Herbrich [GT: SniperEye] Microsoft Research Ltd.

Thanks for all the interest in the TrueSkill algorithm! To answer your question [quote]However, I still believe that my interpretation is correct, and individual performance IS NOT considered in your rank for team games.[/quote]This is correct, only the final standing of the teams matters. BTW, there is a new slide deck with videos and audio available from GameFest 2007 that might help answering some of the questions asked here (see [url=http://research.microsoft.com/mlp/apg/details.aspx]this link[/url]) Feel free to ask more questions. Ralf Herbrich [GT: SniperEye] Microsoft Research Ltd. (co-inventor of TrueSkill)

imagine trying to program this ranking system into a computer :|

the system is easy to understand, well not easy but just read over and really take in what the micrrosoft article is saying and you'll eventually understand it. i now understand completely why im stuck at level 47 in ts, mostly because ive played with the same 3 people EVERY single game. theres no way of knowing when your going to level up tho. because theres no way of knowing your opposing team members uncertainty of winning, so really understanding the system won't hepl you out at all, just know that it works better then any before it [Edited on 10.27.2007 5:49 AM PDT]

If they make Big Team Battle ranked I think I know how to exploit the ranking system without cheats or hacks.

Heh, CoD4 on 360 is using the TrueSkill ranking system (whch I'm sorry but it really is excellent), but the PS3 version doesn't have any sort of ranking system! Hah, unless you're an amazing player PS3 online is gonna be shiz.

I blame President Bush. Damn Republicans with their ranking systems........

w00t

i enjoyed that read alot more then i thought i would, i think this formula pretty much sums it up nicely. R' = R + K * (S - E) R' is the new rating R is the old rating K is a maximum value for increase or decrease of rating (16 or 32 for ELO) S is the score for a game E is the expected score for a game Much of the trick is in figuring out what the (E)xpected score of a game is. ELO uses the following formulas for players A and B: E(A) = 1 / [ 1 + 10 ^ ( [R(B) - R(A)] / 400 ) ] E(B) = 1 / [ 1 + 10 ^ ( [R(A) - R(B)] / 400 ) ] It's a good model because, using the two formulas, it means that a great player gains little from beating an average player, but an average player gains a lot from beating a great player. Take the following example: R(A) = 1900 R(B) = 1500 E(A) = 1 / [ 1 + 10 ^ ( [1500 - 1900] / 400 ) ] = 1 / [ 1 + 10 ^ ( -400 / 400) ] = 1 / [ 1 + 10 ^ -4 / 4 ] = 1 / [ 1 + 10 ^ -1 ] = 1 / 1 + .1 = .91 = 91% E(B) = 1 / [ 1 + 10 ^ ( [1900 - 1500] / 400) ] = 1 / [ 1 + 10 ^ ( 400 / 400 ) ] = 1 / [ 1 + 10 ^ 1 ] = 1 / 11 = .09 = 9% Player A is expected to score .91 in an average game, which is to say he should win 91% of the time, and will be punished accordingly if he loses to player B: R' = 1900 + 32 * (0 - .91) R' = 1900 - 29.12 R' = 1871 Conversely a win nets him very little: R' = 1900 + 32 * (1 - .91) R' = 1900 + 32 * .09 R' = 1900 + 2.88 R' = 1903

[quote][b]Posted by:[/b] pugzr1 The bottom line is who realy cares...play the game and have fun period...[/quote] i care because i enjoy using my brain. i want to know how everything works! because knowing how things work makes you respect the game more and for me at least makes the game more enjoyable. more over, i feel that knowing how it works might give me some in site into improving it! the [i]real[/i] bottom line is that without people like me and Clarkosaur here, and the many others who have posted information on this thread there would not be games like halo or websites like b-net. "The Greatest Advances of our Civilization Are from a little Curiosity in just the right places" Never Insult the curious! the drive to understand is the definition beauty.

this whole true skill thing is crazy...the other day in slayer one of my friends lagged at the start of the game. we still won and i was +12 in my kills and when we joined up with him in the lobby after the game he ranked up i stayed the same. The bottom line is who realy cares...play the game and have fun period...

personally i think at this point it would be nice to see a few numbers in a pre game lobby showing our estimated skill level as represented by the "σ" value (where the game assumes our rank lies such as between levels 15-20) in the Microsoft calculations. i know it would not be wise to show us the probability of beating the team we are matched up ageist because it would give undue confidence in the team expected to win but i would like to know my personal stats that the game is using to determine my rank.

[quote][b]Posted by:[/b] Clarkosaur The IGN article is wrong. It is contradicted by all the Microsoft articles listed in this thread. The guy who wrote the IGN article misunderstood what Microsoft's Truskill designers are saying the same way you are misunderstanding it. [/quote] i agree. i think the IGN person either misunderstood the systems parameters or just plain out speculated about how the systems works. i believe there is a lot of speculation going on about how the system works but i really think "Clarkosaur" has gotten it right. at this point i don't really trust any source but Microsoft them selfs because this is a very complicated algorithm to understand and when people don't understand it completely they start speculating and inserting features they [i]Want[/i] even though the system may not actually have those features. thank you "Clarkosaur" for you work. eventually (maybe over the weekend when i have some free time) i would like to get together a very simple bullet point outline of how the system works and what numbers it uses to run the algorithm. i might be shooting you some questions to make sure i get everything dead on "Clarkosaur". hopefully we can get everyone on the same page soon enough :)

[quote][b]Posted by:[/b] Clarkosaur Richcricketz, you should really add that link to your list. [/quote] done :)

[quote]The IGN article also says that it is considered and they interviewed the actual architects of the TrueSkill system. So, while the quote above is not directly from them, guess who they got that information from. If the makers of this system say that it does, then it must be true.[/quote] The IGN article is wrong. It is contradicted by all the Microsoft articles listed in this thread. The guy who wrote the IGN article misunderstood what Microsoft's Truskill designers are saying the same way you are misunderstanding it. [quote]Ralf Herbrich, Tom Minka, Thore Graepel. TrueSkill(TM): A Bayesian Skill Rating System 2007 Advances in Neural Information Processing Systems 20 We present a new Bayesian skill rating system which can be viewed as a generalisation of the Elo system used in Chess. The new system tracks the uncertainty about player skills, explicitly models draws, can deal with any number of competing entities [b]and can infer individual skills from team results.[/b] Inference is performed by approximate message passing on a factor graph representation of the model. We present experimental evidence on the increased accuracy and convergence speed of the system compared to Elo and report on our experience with the new rating system running in a large-scale commercial online gaming service under the name of TrueSkill.[/quote] SuperScoutin, you don't understand what infer means. The system INFERS individual skills FROM TEAM RESULTS. Read the whole Q&A article you posted the link to. It explains how if you just play with the same people over and over again it CAN'T DISTINGUISH YOUR INDIVIDUAL SKILL LEVELS. Only if you play with and against a variety of players can it INFER you individual skill level from a variety of TEAM RESULT OUTCOMES. It says the same thing as the other Microsoft articles, but it's the simplest explanation. Read it slowly and carefully. I'm really trying to be nice about this, but you are in over your head, both in terms of understanding some mathematical concepts of statistical analysis, and in terms of reading comprehension. I'm sorry bro, but you are wrong. If you are a student take one of the technical microsoft articles to your math teacher and ask him to explain what it's saying. You are reading the words but you're not understanding what's really being said. Microsoft is making certain claims about their system's ability to INFER your individual skill from numerous game outcomes. They themselves admit that the only assessment they use is which games you won and which you lost, and who you beat or lost to. Please read the Q&A more carefully, it lays it out in pretty simple terms. I'm sorry if you think I'm being rude, but I think you should stop being insulted and stop letting your hurt feelings cloud your judgement. I'm just trying to help you understand this. -Clark

[quote]i think the old H2 system would really shine now that modders don't control the 30-50 ranks.[/quote] First of all the modders controlled no ranks. There are simple methods to avoid modders. Don't play new maps, don't go into games without boxes. Most of my friends were 32, seemed to be the average to me. I've been symbols before the stats reset, both in DT and slayer. They didn't own anything. If there's a way to cheat, there's a way to prevent cheating (with the exception of packet spamming) that everybody has the option to do. Second, modders couldn't level past the high 30 low 40 range since it wasn't a viable option to waste a 2 month on maybe 5 games. At those levels, if you were modding, you would search with a bridger, or on boxes and only new maps. The chances of finding a game is slim. Standbyers and packet spammers as well as lag out glitchers were plenty but still easy to beat. The bottom line is that it wasn't hard to rank in Halo 2 if you were good and careful, it just took more time at higher levels, and no 30 isn't a high level, those start at around 39.