Halo 3 Forum
This topic has moved here: Subject: In depth explanation of the Halo 3 skill ranking system.
  • Subject: In depth explanation of the Halo 3 skill ranking system.
Subject: In depth explanation of the Halo 3 skill ranking system.

Posted by: ReaperX6 BR
Sigma is influenced by SOCIAL matches?


No, sigma is only influenced by the games played in ranked games.

p.s Sigma isnt an overall thing, for each playlist your sigma is different

  • 12.03.2007 3:54 AM PDT

Halo/Bungie FTW!

Oh, i think i understand now =D

So the thing is not that bad at all =DDDD


I'll continue to kick ass through Team Slayer with my friend, let's see what happens



ps.: we got almost the same armor and stats

[Edited on 12.03.2007 5:11 AM PST]

  • 12.03.2007 5:08 AM PDT

etc etc/glaringly obvious/and so on, and such <=Not redundant!
Posted by: Cr4ne Style
Taxes do nothing to affect the share of wealth, since taxes are only applied to income.

So that's not even a part of the conversation at all, so it's pointless talking about it....

"for a "best" moral to exist, there must exist the "best" moral base. If the base of morality varies from location to location, culture to culture...then there can't be an absolute moral..

El Kafungus and TwoFaced1680, i have a question for you:

can you explain to us why you think that streakiness affects our ranks? i just looked over the trueskill pages, and i cannot find anything that leads me to believe that having streaks affects how our ranks change.

for example, consider two players, A and B, both of whom are new to h3 and are ranking up in a given playlist. A has the habit of winning 8 out of every 10 games played, and out of every 10 games played, there will be two losses mixed in. B also wins 8 out of every 10 games played, but B wins 8 consecutive games and then loses 2 games in a row.

as far as i can gather, both A and B would rank up the same way (supposing that their opponents and teammates have identical ranks and factors that go into ranks), and that player B is streaky is not an issue.

what do you guys think?

  • 12.03.2007 6:47 AM PDT

The reason it wouldn't be the same is because player B isn't as consistent as player A. While they win the same amount of games, player B may only win if he is in a team of 4 and lose almost every other time. Player A however is consistently winning 8/10 games (4 games win streaks followed by 1 loss).

Although when you take into account what other team levels are, that will also change rank progression.

  • 12.03.2007 7:10 AM PDT

etc etc/glaringly obvious/and so on, and such <=Not redundant!
Posted by: Cr4ne Style
Taxes do nothing to affect the share of wealth, since taxes are only applied to income.

So that's not even a part of the conversation at all, so it's pointless talking about it....

"for a "best" moral to exist, there must exist the "best" moral base. If the base of morality varies from location to location, culture to culture...then there can't be an absolute moral..

Posted by: SatansReverence
The reason it wouldn't be the same is because player B isn't as consistent as player A. While they win the same amount of games, player B may only win if he is in a team of 4 and lose almost every other time. Player A however is consistently winning 8/10 games (4 games win streaks followed by 1 loss).

Although when you take into account what other team levels are, that will also change rank progression.


no, they are both consistent to the same degree since they both win 8 out of every 10 games. for the sake of the example, suppose that all of their traits are the same aside from how they manage to win 8 out of every 10 games.

  • 12.03.2007 7:18 AM PDT

Posted by: sesquipadelian

no, they are both consistent to the same degree since they both win 8 out of every 10 games. for the sake of the example, suppose that all of their traits are the same aside from how they manage to win 8 out of every 10 games.


Well, overall they calculate to the same consistency, but B is less consistent due to not winning in a more predictable manner.

Although, now that I think harder about it, player B might rank up faster due to his Sigma value staying higher.

  • 12.03.2007 7:26 AM PDT

I said brown.

does anyone know if there is any allowing for extra-ordinary games in the system? For example, tonight i played a team slayer match, 3v3 on narrows. All players on the other team where ranked higher than me, and all were in foreign countries. We had a terrible connection and the game started with us going 10-0 down, then one of my team mates quit, and the other lagged out seconds later, and it ended up being me V 3, like i said they were all ranked higher... obviously i lost, 15-30 or something, but i still leveled down from this... so what im asking, is if there is a condition in the system to say 'dont count that game, its not a fair representation'

ps, sorry for the long post, but i didn't want to be a little -blam!- and start a whole topic for it, but this level down did seem pretty unfair and i wanted to ask about it..

  • 12.03.2007 7:28 AM PDT
  • gamertag:
  • user homepage:
  • last post: 01.01.0001 12:00 AM PDT

i thank you for finally clearing this up for me and everyone else who was trying to figure it out themselves. BUT, I'm kind of pissed to find out after winning so many games in a row and raising my Mu and consistently doing so I'm really screwing myself and having to wait every 20 wins to level once. Just the other night me and 3 other friends WON 43 team slayer games in a row, at first i was on my lower tag(sic hunter) helping him out but i realized we were doing so good i might as well get on this one. So we continued to not let one team get above 35 kills on us and my friend that wanted help (hes a 38) did not level ONCE! While another guy guy jumped from 31 to 45 by the end of all the games. So from what you just said he had high Mu from us winning but a lower Sigma? I think...idk. But anyways thanx again for the thorough explanation, and i just thought i should share my own story of bull leveling to everyone.

  • 12.03.2007 8:54 AM PDT

Halo/Bungie FTW!

I'm starting to get irritated

I won 19 CONSECUTIVE matches without loosing and leved up just once while my friend got about 10 levels!!

  • 12.03.2007 10:28 AM PDT

Posted by: sesquipadelian
Posted by: SatansReverence
The reason it wouldn't be the same is because player B isn't as consistent as player A. While they win the same amount of games, player B may only win if he is in a team of 4 and lose almost every other time. Player A however is consistently winning 8/10 games (4 games win streaks followed by 1 loss).

Although when you take into account what other team levels are, that will also change rank progression.


no, they are both consistent to the same degree since they both win 8 out of every 10 games. for the sake of the example, suppose that all of their traits are the same aside from how they manage to win 8 out of every 10 games.


They have the same win% but they are not the same consistency assuming A goes 4-1 & 4-1 while B goes 8-2. First off, their sigma is adjusted more for a streak of 2 or more games. So, Player A would have his sigma stalled or lowered twice when he broke up his streaks while B would only have his stalled or lowered once after his first loss then RAISED again after his second loss.

A couple things to keep in mind about TrueSkill that you may or may not care about:

Sigma is adjusted up slightly after each game to keep it from getting to 0 and is adjusted up more if you are on a streak of 2 or more games.
It takes in to account if you were supposed to win or lose and bases it's adjustments off of that so if you were supposed to win or lose, your sigma lowered.

Here are a couple quotes from the MS pages that give a few clues.

Before starting to determine the new skill beliefs of all participating players for a new game outcome, the TrueSkill ranking system assumes that the skill of each player may have changed slightly between the current and the last game played by each player. The mathematical consequence of making such an assumption is that the skill uncertainty σ will be slightly increased, the amount of which is, in principle, a configurable parameter of the TrueSkill ranking system. It is this parameter that both allows the TrueSkill system to track skill improvements of gamers over time and ensures that the skill uncertainty σ never decreases to zero ("maintaining momentum").

The TrueSkill ranking is assuming a small skill change between any two consecutive games in a game mode so it is able to identify your new, higher skill. But, if your skill has completely changed (you became the best player in the world from previously being the worst player in the world), then you would need to play a large number of games. We designed the system such that it would need between 50 - 100 games before the system would be able to track a substantial skill increase/decrease.

  • 12.03.2007 11:33 AM PDT

Posted by: ReaperX6 BR
I'm starting to get irritated

I won 19 CONSECUTIVE matches without loosing and leved up just once while my friend got about 10 levels!!
Your friend has a different background then you. It only took him 45 games to get to his 34 (which is about where he was when your streak started). Out of those 45 games he only lost 7. His situatoin puts him in a prime position to level fast.

You on the other hand took much longer to get to your 34. Sorry to say but at this point your not in a position to level as fast as your friend. It may be frustrating that you are being treated differently but you have different histories so you are not going to level the same.

You also are playing in a very mixed party and I think that is effecting you more then your other friend. I believe because he accomplished his rank so fast he is in a better position to move quickly so the problem with the mixed party isn't obviously effecting him yet. There are known issues with mixed parties that can cause individuals to level slowly or even stop all together. Bungie is working on this issue. That could be part of your problem.

I don't know the best course of action but going in alone or dropping the lowest rank from your party might help you level faster. However, your friend who's now a 40 is still going to be in a better position to level do to his history. Your going to have to just live with the fact you won't level the same.

  • 12.03.2007 1:02 PM PDT

etc etc/glaringly obvious/and so on, and such <=Not redundant!
Posted by: Cr4ne Style
Taxes do nothing to affect the share of wealth, since taxes are only applied to income.

So that's not even a part of the conversation at all, so it's pointless talking about it....

"for a "best" moral to exist, there must exist the "best" moral base. If the base of morality varies from location to location, culture to culture...then there can't be an absolute moral..

Posted by: Jay120171
They have the same win% but they are not the same consistency assuming A goes 4-1 & 4-1 while B goes 8-2. First off, their sigma is adjusted more for a streak of 2 or more games. So, Player A would have his sigma stalled or lowered twice when he broke up his streaks while B would only have his stalled or lowered once after his first loss then RAISED again after his second loss.


they do have the same consistency if looked at over the life of their careers or from the perspective of 10 games. they do not have the same consistency if you look from the perspective of 5 games.

you mention that the level of sigma adjustment is contingent upon streaks (you talk about a streak of 2 games). where did you see anything about that in the trueskill descriptions? i have seen nothing along those lines. your account would be correct based on a lot of assumptions, and those assumptions are built upon arbitrary considerations. the rank system may very well take streaks into consideration, but i doubt it, and i also doubt that it does so as you have explained.

also, what makes you think that it would view streaks as you have described? what is more valuable, having a streak of 8 wins, a streak of 2 losses, and so on? or having a streak of 4 wins, a loss, and so on? my claim is that the streakiness is irrelevant in determining skill in these cases. and, as far as i can see, the trueskill explanations do not say anything about how streakiness affects sigma or ranks.

  • 12.03.2007 1:22 PM PDT
  • gamertag:
  • user homepage:
  • last post: 01.01.0001 12:00 AM PDT

finaly a good source.... the ranking system realing confused me

  • 12.03.2007 2:02 PM PDT
  • gamertag:
  • user homepage:
  • last post: 01.01.0001 12:00 AM PDT

Great post that explains why I rank up and down more frequently than my buddies. It certaintly answered many of our questions.

  • 12.03.2007 3:04 PM PDT
  • gamertag:
  • user homepage:
  • last post: 01.01.0001 12:00 AM PDT

as soon as i got to a highest skill of 39 it wouldn't let me get any higher even if i won

  • 12.03.2007 3:06 PM PDT

Posted by: sesquipadelian
Posted by: Jay120171
They have the same win% but they are not the same consistency assuming A goes 4-1 & 4-1 while B goes 8-2. First off, their sigma is adjusted more for a streak of 2 or more games. So, Player A would have his sigma stalled or lowered twice when he broke up his streaks while B would only have his stalled or lowered once after his first loss then RAISED again after his second loss.


they do have the same consistency if looked at over the life of their careers or from the perspective of 10 games. they do not have the same consistency if you look from the perspective of 5 games.

you mention that the level of sigma adjustment is contingent upon streaks (you talk about a streak of 2 games). where did you see anything about that in the trueskill descriptions? i have seen nothing along those lines. your account would be correct based on a lot of assumptions, and those assumptions are built upon arbitrary considerations. the rank system may very well take streaks into consideration, but i doubt it, and i also doubt that it does so as you have explained.

also, what makes you think that it would view streaks as you have described? what is more valuable, having a streak of 8 wins, a streak of 2 losses, and so on? or having a streak of 4 wins, a loss, and so on? my claim is that the streakiness is irrelevant in determining skill in these cases. and, as far as i can see, the trueskill explanations do not say anything about how streakiness affects sigma or ranks.


Assumptions:
-Both players already have a high sigma because they are new to H3.
-Both players had their wins and losses to equally skilled players even though they were at different points in their career.

Facts:
-Sigma is decreased for every game played depending on the outcome along with the probability of the outcome.
--"Every match provides the system with more information about each player’s skill belief, usually driving σ down."

-Sigma is increased for every game played.
--"Before starting to determine the new skill beliefs of all participating players for a new game outcome, the TrueSkill ranking system assumes that the skill of each player may have changed slightly between the current and the last game played by each player. The mathematical consequence of making such an assumption is that the skill uncertainty σ will be slightly increased, the amount of which is, in principle, a configurable parameter of the TrueSkill ranking system. It is this parameter that both allows the TrueSkill system to track skill improvements of gamers over time and ensures that the skill uncertainty σ never decreases to zero ("maintaining momentum")."

I did make an assumption about "consecutive games" meaning consecutive wins or losses and that is probably wrong. It doesn't spell out details about streaks directly affecting sigma. It does however, explain that your sigma starts off high and is decreased after every game and increased after every game. The amount of decrease is based on the systems assumption of you winning or not. The amount of increase is a constant set by the developer.

MS also states that when you win and lose predictably, it is sure about your rank and slows your progression down. I assume this means that when you win and lose predictably, you have a low sigma because they define it as "uncertainty". It then takes a large number of games to go up or down (10 for 2 Teams of 4 Players). I assumed this was cause by your sigma getting increased because you were now winning games you shouldn't but this may be incorrect as well. It may just be you slow down because you've played a lot of games therefor your mu increases are smaller and not necessarily your sigma is low.

Yes, I assumed a player winning 8 in a row would have a higher sigma because the system figures they should lose once in a while but I cannot find a direct quote to prove it and this is the assumption we were all making. You are saying that a person winning 8/2 should be the same skill as a person going 4/1 & 4/1but there is no direct evidence to prove that is correct or incorrect and I think you will settle for nothing less.

[Edited on 12.03.2007 4:20 PM PST]

  • 12.03.2007 4:19 PM PDT

Halo/Bungie FTW!

Ok i mean..WHAT THE -blam!- PIECE OF -blam!-?!?!!??!?!?!?

I won 23 matches IN A SINGLE STUPID ROW and leveled up O-N-C-E???!?!?!?!?!?!?!!?!??!?!??!?!?!??!?!


What's wrong with this stupid fakeskill ???!!?

  • 12.03.2007 5:15 PM PDT
  • gamertag:
  • user homepage:
  • last post: 01.01.0001 12:00 AM PDT

HALO 3 IS A -blam!- PEICE OF -blam!-! ITS SOOOO GLITCHY, SUCH AS WHEN YOU SHOOT SOMEONE IN THE FACE WITH A SHOTGUN AND THEY BEAT YOU DOWN, -blam!- HALO 3, HALO 2 IS WAYYY BETTER

  • 12.03.2007 6:04 PM PDT

is there any way to see my sigma or MU?

  • 12.03.2007 6:08 PM PDT

Posted by: sesquipadelian
El Kafungus and TwoFaced1680, i have a question for you:

can you explain to us why you think that streakiness affects our ranks? i just looked over the trueskill pages, and i cannot find anything that leads me to believe that having streaks affects how our ranks change.

for example, consider two players, A and B, both of whom are new to h3 and are ranking up in a given playlist. A has the habit of winning 8 out of every 10 games played, and out of every 10 games played, there will be two losses mixed in. B also wins 8 out of every 10 games played, but B wins 8 consecutive games and then loses 2 games in a row.

as far as i can gather, both A and B would rank up the same way (supposing that their opponents and teammates have identical ranks and factors that go into ranks), and that player B is streaky is not an issue.

what do you guys think?
I didn't help write this post in anyway and my knowledge of the system is more in knowing the principles rather then the technical details. I've never personally made a statement that streakiness effects your ranking (i don't think I have anyway). I don't know the OP's opinion on the subject and I certainly can't speak for him. However, I believe this statement "Play really consistently, and you have a low Sigma. Play streakily, and you have a higher Sigma." may be a bit misleading/confusing. The OP later states that it's his theory sigma can actually be increased if you continue to go on a winning streak. This was stated to be theory not actual fact. Obviously if your sigma could go back up that would allow you to rank faster again, but there are no statements I know of that prove this theory.

As far as my answer to wheather or not streakiness effects how you rank...I don't know.

What I do know is according to the documentation your sigma is supposed to go down after every game, although there can be exceptions. Also before every calculation is made sigma is raised a bit (I believe the developer can set this as they like) so that the system can maintain momentum. Does Bungie handle this in such a way that sigma could increase again? No idea. Without knowing that, there is really know way for us to know wheather longer win streaks are better then shorter ones.

I decided to run an experiement using Microsofts TrueSkill calculator. If you want to check it out the TrueSkill Calculator can be found here.

I wanted to test both your scenerios. The rank calculator will tell you what the resulting MU and Sigma values will be based on the given inputs. The Red team here always wins and the Blue team always loses. Note the game mode was set to "2 large teams". So if you want the imaginary team to win they should be red, if you want them to lose they should be blue. I gave my team an initial MU=1 and a Sigma =8.33. The opponents were always given the same MU as our imaginary team but a sigma of 1. After each calculation I would take the results for the imaginary team (rounding MU down) and insert that data into the appropriate inputs for the next game as well as adjust the opponents MU to match. I noted the MU and Sigma values after each result. Here are the results.

Test #1
Here I tested what happens to a team that wins 8 in a row then loses 2

Game 1 (W): 3 (MU)/7.869 (Sigma)
Game 2 (W): 5/7.451
Game 3 (W): 7/7.072
Game 4 (W): 9/6.727
Game 5 (W): 11/6.413
Game 6 (W): 13/6.126
Game 7 (W): 14/5.864
Game 8 (W): 15/5.624
Game 9 (L): 13/5.404
Game 10 (L): 11/5.201

Test #2
Here I tested a team that won 4, lost 1, won 4, lost 1

Game 1 (W): 3/7.869
Game 2 (W): 5/7.451
Game 3 (W): 7/7.072
Game 4 (W): 9/6.727
Game 5 (L): 6/6.413
Game 6 (W): 8/6.126
Game 7 (W): 9/5.864
Game 8 (W): 10/5.624
Game 9 (W): 11/5.404
Game 10 (L): 9/5.201

Summary
Test 1 shows the team to have a final MU = 11 and a Sigma = 5.201
Test 2 shows the team to have a final MU = 9 and a Sigma = 5.201

Sigma is lowered the same regardless of win or loss. Thus the final Sigma values are identical regardless of how we got there. The notable difference is the fact that Test 1 resulted in a MU = 11 vs Test 2 which resulted in a MU = 9. Does this mean streaks help? Not necessarily. The reason behind the difference I believe has to do with the earlier loss but not really the streak. Sigma decreased the same after every game. Thus when we lost earlier the team had I higher Sigma and I think that hurt our MU much more. While the team that continued to win benefited from the higher sigma when it won. Also because it lost so much later Sigma was lower and the loss wasn't as penalizing.

Is this a definitive study? No. It's just something I did quickly and there might be something I'm missing. If anyone has a better test give it a shot. Until we know exactly how Bungie handles Sigma from game to game we won't really know if/how much streaks help. Assuming Sigma always decreases winning early seems to be beneficial because sigma is always decreasing. So to anyone reading this if you are planning on winning 8 out of 10 tonight try to get the wins in early and save the losses for last :) Truthfully though there are so many other variables that the order you win is insignificant and in the end probably has zero impact on leveling. Most of us can't control when we win either and those that can generally don't choose to lose. So really none of this matters.

[Edited on 12.03.2007 6:35 PM PST]

  • 12.03.2007 6:19 PM PDT

I've been here A WHILE. sneaking,creeping,hiding. Don't let levels ever fool you, I will put you in the ground.


The number on my gamertag is a 4 but it should now be a 6 =O

P.S- Be respectful and play nice!

Nice post. My skill is 40 LW and it usually raises every 1-2 games.

  • 12.03.2007 7:45 PM PDT

etc etc/glaringly obvious/and so on, and such <=Not redundant!
Posted by: Cr4ne Style
Taxes do nothing to affect the share of wealth, since taxes are only applied to income.

So that's not even a part of the conversation at all, so it's pointless talking about it....

"for a "best" moral to exist, there must exist the "best" moral base. If the base of morality varies from location to location, culture to culture...then there can't be an absolute moral..

Posted by: Jay120171
Yes, I assumed a player winning 8 in a row would have a higher sigma because the system figures they should lose once in a while but I cannot find a direct quote to prove it and this is the assumption we were all making. You are saying that a person winning 8/2 should be the same skill as a person going 4/1 & 4/1but there is no direct evidence to prove that is correct or incorrect and I think you will settle for nothing less.


i cannot come up with any reason to suppose that streakiness is a factor in the ranking system. i cannot find any passages from microsoft to explain this element being present in trueskill, and i cannot imagine how such a trait could be used (but, my statistical intuitions and creativity are highly limited, so i am not claiming that it cannot be done). given what i know about bayesianism, i cannot imagine how streaks of any kind could affect how confidence levels could be influenced by looking for patterns in the order of wins and losses relative to one another. match results of each and every match will influence how the rank system assigns degrees of belief concerning player skill levels, and i literally cannot imagine how they would implement measures to evaluate streaks, let alone assign different weights to the many different possible kinds of streaks.

i am not demanding or expecting anything, but i am curious. my guess (and this is mere conjecture) is that streakiness is not a factor in the ranking system. you and the thread creator seem to understand the system completely and can grasp how different elements of the trueskill system affect ranks in different ways, so i am eager to know if either of you have any concrete reasons to suppose that streakiness is a factor.

  • 12.03.2007 9:31 PM PDT

etc etc/glaringly obvious/and so on, and such <=Not redundant!
Posted by: Cr4ne Style
Taxes do nothing to affect the share of wealth, since taxes are only applied to income.

So that's not even a part of the conversation at all, so it's pointless talking about it....

"for a "best" moral to exist, there must exist the "best" moral base. If the base of morality varies from location to location, culture to culture...then there can't be an absolute moral..

Posted by: TwoFaced1680
:) Truthfully though there are so many other variables that the order you win is insignificant and in the end probably has zero impact on leveling. Most of us can't control when we win either and those that can generally don't choose to lose. So really none of this matters.


i just read your last post. thanks for taking the time to run the numbers and to explain them in a way that makes the results accessible to us all.

while i agree that all any of us can do is make slightly informed stabs at the dark, it is fun (for me) to try to figure this sort of thing out, and it is fun to get to read posts by people who have a background in statistics. it is a rare treat in the forums.

  • 12.03.2007 9:37 PM PDT

Posted by: The Tim
does anyone know if there is any allowing for extra-ordinary games in the system? For example, tonight i played a team slayer match, 3v3 on narrows. All players on the other team where ranked higher than me, and all were in foreign countries. We had a terrible connection and the game started with us going 10-0 down, then one of my team mates quit, and the other lagged out seconds later, and it ended up being me V 3, like i said they were all ranked higher... obviously i lost, 15-30 or something, but i still leveled down from this... so what im asking, is if there is a condition in the system to say 'dont count that game, its not a fair representation'

ps, sorry for the long post, but i didn't want to be a little -blam!- and start a whole topic for it, but this level down did seem pretty unfair and i wanted to ask about it..
There is no contingency I am aware of to not count a game. Some games get marked unreliable (I think when fewer then half the people finish) but my guess is they still count those games. It's unlikely they would ever make exceptions as people are always looking for something to abuse. The level down might have felt unfair but in the long run you really won't even notice it.

  • 12.03.2007 10:47 PM PDT

"Our Arrows will Blot out the sun." "Then we will fight in the shade"
This is my favorite line from 300 THAT MOVIE RULEZ ALL

Thank you for tis explanation ive been wondering exactly how your level is decided....at first i thought it was exactly like H2 where you wn 2 or 3 gamesin a row against people n the same level as you and you go up or if you win a ew games against people higher then you you go up faster.....Obvously i figured out when i got to a 48 that this was NOT correct....at least now i have an explanation as to why i gain and lose levels at such an odd pace....I just wanted to thank you for the great explanation of the entire ordeal....even if it wont help 90% of the bungie community considering that most of them have no clue what True Skill is let alone your explanation using the Sigma Mu variables. Anyways once again Thank You.

  • 12.03.2007 10:57 PM PDT