On August 19 2011 00:51 IronDoc wrote:
But if his points reflected his mmr and yours did, then the points would zero sum too. You can only look at the points (and infer a likelihood of winning) on the side of a player whose mmr and point value are close enough to be assumed identical. Essentially, the thing is that if the points won/lost don't zero sum, at least one players' points are not reflective of mmr.
Show nested quote +
On August 19 2011 00:44 JackDragon wrote:
But thats talking about promotions, not points. You could also see that as number of points don't matter when it comes to promotions. Which is true, points don't matter. So they don't actually say that time have nothing to do with points gained.
Also IronDoc, what I mean is doesn't his argument base on that likelihood in winning does sum to zero? I would argue that for the formula to be correct our likelihood of winning needs to sum to zero or it won't add up.
Example from my last game: I won 13 points he lost 12 points (before bonus pool) and so I had 45.83% chance to win right? And he would have 50% chance of winning. And this would mean that there is 4.17% chance for it to be tie or what? well I guess that could be but...
Point being. If the persons likelihood of winning don't add up you can not say that your formula appropriately reflects the truth.
On August 18 2011 23:44 BlasiuS wrote:
yes you missed it, it's right here:
On August 18 2011 23:38 JackDragon wrote:
I found the thread and post else I wouldn't be able to qoute from it right? I just can't find within that post were it says that it is pure mmr. I am dyslectic so it is possible that I missed it but please show me.
On August 18 2011 23:10 Not_That wrote:
It states specifically that points won/lost are only based off of who won in another thread by excalibur_z. In the sticky it can be seen from the part discussing the favoured system. If you won more points for longer games, score, or anything else it would be entirely abusable.
It's not true that both players' points have to fit their MMR for my formula to be true. It is sufficient that your points match your MMR for you to be able to calculate what was your chance to win. It would only be required for your opponent's MMR to match his points for the formula to be also applicable to your opponent.
The part you quote is referring to the 'favored' / 'slightly favored' / 'even' caption at the beginning of matches, and how both sides can see each other as 'favored', etc.
Small note although I hate to make this any more complicated than it already is, but I can not resist:
If for some reason you are confident that your opponent's current points better reflect his MMR than your current points reflect your MMR, then it is entirely possible to calculate your winning chances based off of his points won / lost - simply calculate HIS winning chances, and yours are equal to 100 minus his chances. Just be sure you aren't counting his bonus pool points.
On August 18 2011 22:57 JackDragon wrote:
I can't find were it says that the points aren't reflecting game length, or that it only reflects mmr gap.
I mean even so it is a bit to simplified to actually work. Saying that the skill is constant and that it only works when you have reached your true mmr makes it not very usefull.
What I mean is this:
"The important thing to remember is that "favored" does not always mean "better" unless both players' points have approximately reached their MMRs. Until that time, the "favored" indicator only serves to determine how many points a match is worth, and is not an indicator of skill."
which mens that both have to fit their mmr for your formula to even be true.
On August 18 2011 22:07 Not_That wrote:
It's not official, it's something I thought of and I explain how I came to the numbers I did in the OP.
Match duration definitely does not affect points. You can read about it in the ladder sticky.
On August 18 2011 22:02 JackDragon wrote:
Is this official or were did you get the numbers from? I was certain that time of the match also affected points but whatever. I do think that there is a relation betwen skill gap and points but...
I'm not sure you are correct
Is this official or were did you get the numbers from? I was certain that time of the match also affected points but whatever. I do think that there is a relation betwen skill gap and points but...
I'm not sure you are correct
It's not official, it's something I thought of and I explain how I came to the numbers I did in the OP.
Match duration definitely does not affect points. You can read about it in the ladder sticky.
I can't find were it says that the points aren't reflecting game length, or that it only reflects mmr gap.
I mean even so it is a bit to simplified to actually work. Saying that the skill is constant and that it only works when you have reached your true mmr makes it not very usefull.
What I mean is this:
"The important thing to remember is that "favored" does not always mean "better" unless both players' points have approximately reached their MMRs. Until that time, the "favored" indicator only serves to determine how many points a match is worth, and is not an indicator of skill."
which mens that both have to fit their mmr for your formula to even be true.
It states specifically that points won/lost are only based off of who won in another thread by excalibur_z. In the sticky it can be seen from the part discussing the favoured system. If you won more points for longer games, score, or anything else it would be entirely abusable.
It's not true that both players' points have to fit their MMR for my formula to be true. It is sufficient that your points match your MMR for you to be able to calculate what was your chance to win. It would only be required for your opponent's MMR to match his points for the formula to be also applicable to your opponent.
The part you quote is referring to the 'favored' / 'slightly favored' / 'even' caption at the beginning of matches, and how both sides can see each other as 'favored', etc.
Small note although I hate to make this any more complicated than it already is, but I can not resist:
If for some reason you are confident that your opponent's current points better reflect his MMR than your current points reflect your MMR, then it is entirely possible to calculate your winning chances based off of his points won / lost - simply calculate HIS winning chances, and yours are equal to 100 minus his chances. Just be sure you aren't counting his bonus pool points.
I found the thread and post else I wouldn't be able to qoute from it right? I just can't find within that post were it says that it is pure mmr. I am dyslectic so it is possible that I missed it but please show me.
yes you missed it, it's right here:
On February 22 2011 15:13 Excalibur_Z wrote:
Promotion
By outperforming the rest of your league, it is possible to get promoted into a higher league. If you are in Bronze but playing against Gold players, you would expect to be promoted to Gold, but that doesn't always happen immediately.
This is because the system requires a certain degree of confidence before players can be moved to a new league, otherwise they would bounce around from league to league too frequently for leagues to be meaningful. That confidence is measured in two ways:
First, the player must prove that he is capable of maintaining a certain level of skill. This is done by measuring the moving average of a player's MMR. The below image should help demonstrate. When this moving average stabilizes within the confines of a league, a player can be promoted into that league.
![[image loading]](http://imgur.com/aYJDn.jpg)
The second factor is a "confidence buffer" that exists between leagues. That is, if a Bronze player is only slightly better than the lowest-rated Silver players according to his moving average, that is not reason alone to promote him, even though he has crossed into the Silver MMR region. If that player slumped, he would fall right back into Bronze within a couple of games, only to return to Silver a couple of games later, making his promotion far less meaningful. Leagues are sticky, therefore the moving average must cross beyond the destination league's confidence buffer, cementing that player's position within the higher league. In the picture above, the confidence buffer is represented by the yellow glow region.
Note that these are the only two factors required for promotion. Any in-game behaviors or statistics beyond winning, losing, and the opponent's MMR are not relevant to the system.
Promotion
By outperforming the rest of your league, it is possible to get promoted into a higher league. If you are in Bronze but playing against Gold players, you would expect to be promoted to Gold, but that doesn't always happen immediately.
This is because the system requires a certain degree of confidence before players can be moved to a new league, otherwise they would bounce around from league to league too frequently for leagues to be meaningful. That confidence is measured in two ways:
First, the player must prove that he is capable of maintaining a certain level of skill. This is done by measuring the moving average of a player's MMR. The below image should help demonstrate. When this moving average stabilizes within the confines of a league, a player can be promoted into that league.
![[image loading]](http://imgur.com/aYJDn.jpg)
The second factor is a "confidence buffer" that exists between leagues. That is, if a Bronze player is only slightly better than the lowest-rated Silver players according to his moving average, that is not reason alone to promote him, even though he has crossed into the Silver MMR region. If that player slumped, he would fall right back into Bronze within a couple of games, only to return to Silver a couple of games later, making his promotion far less meaningful. Leagues are sticky, therefore the moving average must cross beyond the destination league's confidence buffer, cementing that player's position within the higher league. In the picture above, the confidence buffer is represented by the yellow glow region.
Note that these are the only two factors required for promotion. Any in-game behaviors or statistics beyond winning, losing, and the opponent's MMR are not relevant to the system.
But thats talking about promotions, not points. You could also see that as number of points don't matter when it comes to promotions. Which is true, points don't matter. So they don't actually say that time have nothing to do with points gained.
Also IronDoc, what I mean is doesn't his argument base on that likelihood in winning does sum to zero? I would argue that for the formula to be correct our likelihood of winning needs to sum to zero or it won't add up.
Example from my last game: I won 13 points he lost 12 points (before bonus pool) and so I had 45.83% chance to win right? And he would have 50% chance of winning. And this would mean that there is 4.17% chance for it to be tie or what? well I guess that could be but...
Point being. If the persons likelihood of winning don't add up you can not say that your formula appropriately reflects the truth.
But if his points reflected his mmr and yours did, then the points would zero sum too. You can only look at the points (and infer a likelihood of winning) on the side of a player whose mmr and point value are close enough to be assumed identical. Essentially, the thing is that if the points won/lost don't zero sum, at least one players' points are not reflective of mmr.
Indeed that makes sense.
Not_That. I don't really know much about how mmr works to know if the numbers fitt so I don't know if how to compare mmr. But If i undestand it right IronDoc summed it up nicly.
I guess the thing missing then would be to know how close you are to your mmr and how close your opponent are to his. Before then I don't think we can comfirm or deny this theory.