|
On July 07 2011 01:57 Excalibur_Z wrote:
I think I get it though. You only get promoted when you fit within a tier (and if you overshoot a league you just move temporarily to the highest tier of that league). So, that means it should have some degree of confidence that you belong in that tier because if it was less conservative in its estimate, you'd just move to a different league anyway. The point offsets work to define what is considered "high points" across all leagues and divisions, so usually you're not going to find someone who has like 1000 adjusted points in Gold and dwarfs all these other Gold players, it's going to be around a certain range. For "larger" divisions like Diamond and Bronze which cover a wider skill range but have the same percentage of the active population, you just add more tiers. Like right now the bonus pool is 1275 and the top points across almost all the leagues is around 1600-1700 so it seems to work out (Bronze has a lot of 2000ish-point people and so does Master).
I get it too, but it seems that not in the same sense as you do. Look, Ill try to explain it with temperature:
Lets say in the spring common temperatures vary from 10-25 °C, while in summer its more like 20-35 °C. Now we can normalize temperature adding modifier to summer temperature -10. Then in both summer and spring we can use rank 10-25 °C to describe temperature outside, however as side effect our modified temperature lost its objective value. So while it makes it "easier" in sense that we use smaller range for temperature (10-25 °C in both spring and summer), we also lost objective value of that.
In other words (w/o knowing all modifiers), if someone says what temperature is outside, it doesnt tell us anything about real temperature.
Perfect example of that is daylight saving time effect. Naturally, 0:00 is when sun is in lowest possition under horizon and 12:00 when sun is in highest possition (its also called nadir and zenit, but I dont remember which is which ^^). OFC because of time zones, this is reduced to only 24 times on Earth (while in reality there is OFC infinity of them). In summur to save daylight, time is modified which makes it that sun possition can be much more different from real nadir/zenit, but it normalize time.
So in winter we go to work from 8:00 and in summer the same. If daylight saving time didnt exist, it would be more like 7:00 in summer.
And here is difference between daylight saving time and ladder - while in life it is *important* that time is normalized (for work, TV, street lights etc.) while real possition of sun is irrelevant, on ladder quite the opposite is correct. You dont really need points to be normalized, cause it doesnt really help anything - but it should have that "skill" information behind it...
|
They should overhaul the matching and ratings system. I am top plat and in a 4v4 my team was 2 plat 1 diamond and one gold. Team we faced were 4 masters who randomed. I checked if they were an arranged team. Also I was #1 in my league with over 200 points more than #2. Then got promoted after they applied a patch only to be demoted when I was still #13 in diamond. Something be fishy.
Ps. Show me my losses dammit!!!
|
On July 07 2011 03:01 aksfjh wrote:Show nested quote +On July 07 2011 01:57 Excalibur_Z wrote:
I think I get it though. You only get promoted when you fit within a tier (and if you overshoot a league you just move temporarily to the highest tier of that league). So, that means it should have some degree of confidence that you belong in that tier because if it was less conservative in its estimate, you'd just move to a different league anyway. The point offsets work to define what is considered "high points" across all leagues and divisions, so usually you're not going to find someone who has like 1000 adjusted points in Gold and dwarfs all these other Gold players, it's going to be around a certain range. For "larger" divisions like Diamond and Bronze which cover a wider skill range but have the same percentage of the active population, you just add more tiers. Like right now the bonus pool is 1275 and the top points across almost all the leagues is around 1600-1700 so it seems to work out (Bronze has a lot of 2000ish-point people and so does Master). Yea, but those 1600-1700 players are most likely in the lowest tiered league since the promotion system doesn't make future predictions. The tier system seems to try to predict a player's "success" in a given league, but that prediction will almost always be wrong. For example, I could inch my way to diamond, get placed in the lowest tier, then inch my way to masters. By the time I'm almost masters, I'm rocking the highest points in the league due to that tiered offset. Meanwhile, somebody in a higher tier could come back from a break and do very poorly, almost dropping down to plat. These players make up the extremes of each league, and league tiers is not likely to change that. In the end, tiers don't even matter to the overall point totals, since people who are placed in the lowest tiers can easily rise to the top of the league anyways, and vice versa for high tiered players. In the end, they do the primary job of obfuscation, but fail at that secondary task of properly predicting end results of players. Why waste the resources in this futile prediction?
I think Blizzard's main intent is to give the people easier rewards. Most casuals will look at the position in their division, so it makes sense to at least try and group them with people who have a similar skill level. If this grouping was successful however, the tier system would be directly visible for everybody, since divisions with better players would consistently have higher points. So they introduce an obfuscation factor to mask this effect. It somewhat makes sense to me.
|
On July 14 2011 21:34 oZe wrote:
only to be demoted when I was still #13 in diamond. Something be fishy.
Rank within a division has more to do with how many games played than skill, due to the bonus pool. So it's not at all fishy that you could be demoted when you were #13.
|
Hello,
I understand the reason that Blizzard had for division tiers. But I don't understand how the system works right at the start of season.
For example: Players A and B play their placement match. A is master, B is placed in tier 7 diamond division (Tier 7: +528 from Master).
For this to work Player A should display 0 points and Player B 528 points. But don't both start at 0?
This really puzzles me.
I understood that after some games (say in the middle of the season) The one in Masters would have 1000 points but the one in the bottom of diamond needs to have his points "inflated" by 528 in order for him also to have around 1000 points. This works well at middle/end season, but at beginning I don't see how they manage it unless Player B started with 528 points, which is not the case.
Can anyone help me explain how this works and where is my error?
|
United States12235 Posts
On July 16 2011 01:11 Kakaru2 wrote:
Hello,
I understand the reason that Blizzard had for division tiers. But I don't understand how the system works right at the start of season.
For example: Players A and B play their placement match. A is master, B is placed in tier 7 diamond division (Tier 7: +528 from Master).
For this to work Player A should display 0 points and Player B 528 points. But don't both start at 0?
This really puzzles me.
I understood that after some games (say in the middle of the season) The one in Masters would have 1000 points but the one in the bottom of diamond needs to have his points "inflated" by 528 in order for him also to have around 1000 points. This works well at middle/end season, but at beginning I don't see how they manage it unless Player B started with 528 points, which is not the case.
Can anyone help me explain how this works and where is my error?
If skill levels coincide with points for a Master and a Tier 7 Diamond player and the skill levels of those two players are equal, then the Diamond player will have 528 points more than the Master player. At the start of the season, though, points don't coincide with MMR, therefore players will earn more points than they lose until they approach their MMR. In this instance, the Diamond player will be earning more points per game than the Master player despite playing the same skill level of opponents.
However, this is an unlikely scenario to happen immediately at the start of a season because the start-of-season Placement Match will put each player into a tier that most accurately represents their skill level. That is, if those same two players from the example started the next season at the same skill level, they would join the same tier.
|
On July 16 2011 01:11 Kakaru2 wrote:
Hello,
I understand the reason that Blizzard had for division tiers. But I don't understand how the system works right at the start of season.
For example: Players A and B play their placement match. A is master, B is placed in tier 7 diamond division (Tier 7: +528 from Master).
For this to work Player A should display 0 points and Player B 528 points. But don't both start at 0?
This really puzzles me.
I understood that after some games (say in the middle of the season) The one in Masters would have 1000 points but the one in the bottom of diamond needs to have his points "inflated" by 528 in order for him also to have around 1000 points. This works well at middle/end season, but at beginning I don't see how they manage it unless Player B started with 528 points, which is not the case.
Can anyone help me explain how this works and where is my error?
People end up staying at 0 points for up to a week in some cases after a reset.
|
TY for the excellent effort on the OP, just won me an argument + $5 (I owe you royalties, Excalibur_Z!)
|
Thanks a lot Excal, I see the light now. It never crossed my mind that the Diamond player could earn more points each game so in the END it will have 528 points more than the Master, being same skill.
|
Team Matchmaking
It is believed that when a team is first formed, the MMRs of all members are averaged to form the team's initial MMR during the first placement match. After that first placement match, the team's MMR is separate per normal. For Random Teams, it is likely that the MMRs of all members are averaged first, then an opposing team with a similar MMR (based upon a degree of uncertainty) is located.
Do we know, or have a strong theory, as to which MMR's are used for new or random teams?
For example, for a new 3v3 team, is it the MMR of all 3 players 1v1 that's averaged? Or the average MMR of all the 3v3 teams of all 3 players? Or the average MMR of all the teams of all 3 players?
For a random 3v3 team is it each of the 3 players 3v3 Random MMR that is averaged, or would it be the average of all 3v3 teams each player has or even the average of all teams that each player has?
|
First of all a big thanks to Excalibur_Z for explaining the system so detailed.
I think i understand the system pretty well now but i have a question/theory/strat on the ladder reset. Maybe someone can find a hole/error in my reasoning.
I think that you have a better chance with a low master mmr than a high master mmr to place well in the ladder after the reset in the first couple of days/matches.
Let's take me as an example. I am currently a mid master player.
If i loose a lot of games (on purpose) before the ladder reset so that i am just on the edge in masters my mmr will be at low masters.
When the new season starts i will face people with aprox. the same mmr. Due to the fact i only artificially lowered my mmr i will likely win a lot more games or get more points for my wins.
Even if i am correct this would not be a huge deal because it balance quite soon but would be a nice trick for people who want to take an excursion to the top of the division.
(btw i really like the ladder system. I think a lot of people don't realize how frustrating it would be if they would be able to see their real mmr remain constant above month or even drop. I think a lot less players would be motivated to play. That would be bad for everybody. )
|
On July 16 2011 05:29 OrbitalPlane wrote:
First of all a big thanks to Excalibur_Z for explaining the system so detailed.
I think i understand the system pretty well now but i have a question/theory/strat on the ladder reset. Maybe someone can find a hole/error in my reasoning.
I think that you have a better chance with a low master mmr than a high master mmr to place well in the ladder after the reset in the first couple of days/matches.
Let's take me as an example. I am currently a mid master player.
If i loose a lot of games (on purpose) before the ladder reset so that i am just on the edge in masters my mmr will be at low masters.
When the new season starts i will face people with aprox. the same mmr. Due to the fact i only artificially lowered my mmr i will likely win a lot more games or get more points for my wins.
Even if i am correct this would not be a huge deal because it balance quite soon but would be a nice trick for people who want to take an excursion to the top of the division.
I think that people that have a high master mmr their opponents will be favored all the time at the reset, thus they will have more points, but if you lower your mmr you'll be even or favored.
|
On July 16 2011 05:29 OrbitalPlane wrote:
First of all a big thanks to Excalibur_Z for explaining the system so detailed.
I think i understand the system pretty well now but i have a question/theory/strat on the ladder reset. Maybe someone can find a hole/error in my reasoning.
I think that you have a better chance with a low master mmr than a high master mmr to place well in the ladder after the reset in the first couple of days/matches.
Let's take me as an example. I am currently a mid master player.
If i loose a lot of games (on purpose) before the ladder reset so that i am just on the edge in masters my mmr will be at low masters.
When the new season starts i will face people with aprox. the same mmr. Due to the fact i only artificially lowered my mmr i will likely win a lot more games or get more points for my wins.
Even if i am correct this would not be a huge deal because it balance quite soon but would be a nice trick for people who want to take an excursion to the top of the division.
You are wrong, the lower your opponent, the less points you will get. So stay at high master to get the more points.
|
On July 16 2011 05:41 SDream wrote:
You are wrong, the lower your opponent, the less points you will get. So stay at high master to get the more points i get.
No that was what i wanted to point out. There are 2 possibilities:
(keep in mind my mmr and his points will be compared for the points i gain)
1. i match the same person
i will get more points because the difference between my mmr and his points is bigger
2. i will meet people who are worse, as a result i have a better winning chance and get the same amount of points as i would winning against a mid master with high mmr.
|
United States12235 Posts
On July 16 2011 05:48 OrbitalPlane wrote:Show nested quote +On July 16 2011 05:41 SDream wrote:
You are wrong, the lower your opponent, the less points you will get. So stay at high master to get the more points i get. No that was what i wanted to point out. There are 2 possibilities: (keep in mind my mmr and his points will be compared for the points i gain) 1. i match the same person i will get more points because the difference between my mmr and his points is bigger 2. i will meet people who are worse, as a result i have a better winning chance and get the same amount of points as i would winning against a mid master with high mmr.
It's the other way around. Your points versus his MMR. Therefore, indirectly you're receiving points according to the gap between your own MMR and your points. The larger the gap, the larger the point change in a particular direction.
|
United States12235 Posts
On July 16 2011 05:12 Wonger wrote:Show nested quote +Team Matchmaking
It is believed that when a team is first formed, the MMRs of all members are averaged to form the team's initial MMR during the first placement match. After that first placement match, the team's MMR is separate per normal. For Random Teams, it is likely that the MMRs of all members are averaged first, then an opposing team with a similar MMR (based upon a degree of uncertainty) is located. Do we know, or have a strong theory, as to which MMR's are used for new or random teams? For example, for a new 3v3 team, is it the MMR of all 3 players 1v1 that's averaged? Or the average MMR of all the 3v3 teams of all 3 players? Or the average MMR of all the teams of all 3 players? For a random 3v3 team is it each of the 3 players 3v3 Random MMR that is averaged, or would it be the average of all 3v3 teams each player has or even the average of all teams that each player has?
That's the real question, isn't it? =) We're not really sure. If I had to speculate, it's probably something like "if you've played with other people in this team in other formats, use that as a starting point, otherwise use other team data, otherwise use 1v1 data."
|
On July 16 2011 06:27 Excalibur_Z wrote:Show nested quote +On July 16 2011 05:12 Wonger wrote:Team Matchmaking
It is believed that when a team is first formed, the MMRs of all members are averaged to form the team's initial MMR during the first placement match. After that first placement match, the team's MMR is separate per normal. For Random Teams, it is likely that the MMRs of all members are averaged first, then an opposing team with a similar MMR (based upon a degree of uncertainty) is located. Do we know, or have a strong theory, as to which MMR's are used for new or random teams? For example, for a new 3v3 team, is it the MMR of all 3 players 1v1 that's averaged? Or the average MMR of all the 3v3 teams of all 3 players? Or the average MMR of all the teams of all 3 players? For a random 3v3 team is it each of the 3 players 3v3 Random MMR that is averaged, or would it be the average of all 3v3 teams each player has or even the average of all teams that each player has? That's the real question, isn't it? =) We're not really sure. If I had to speculate, it's probably something like "if you've played with other people in this team in other formats, use that as a starting point, otherwise use other team data, otherwise use 1v1 data."
I think every team/player starts at "1500 MMR" (or whatever is the "zero"), regardless of having a master 1v1 etc.
What happens is that the system is more complex than just MMR. Everyone starts with a gold level MMR, but the system will still put you against a bronze first (generally), at least on placement matches. I think the same logic applies if you are master then try a random team. You will still start with 1500 MMR, but the system will already predict you might be able to beat another master team and your placements will be with these masters and your MMR will jump fast enough to actually be able to be placed at master in this random team, even with your MMR starting at 1500.
I think that is what happens, but I have no idea if it is possible to prove it somehow =(
Will a master player in 1v1, that loses the 5 placements matches in random team still get a master spot? That would prove that his MMR started that high, which I doubt.
|
United States12235 Posts
On July 16 2011 07:24 SDream wrote:Show nested quote +On July 16 2011 06:27 Excalibur_Z wrote:On July 16 2011 05:12 Wonger wrote:Team Matchmaking
It is believed that when a team is first formed, the MMRs of all members are averaged to form the team's initial MMR during the first placement match. After that first placement match, the team's MMR is separate per normal. For Random Teams, it is likely that the MMRs of all members are averaged first, then an opposing team with a similar MMR (based upon a degree of uncertainty) is located. Do we know, or have a strong theory, as to which MMR's are used for new or random teams? For example, for a new 3v3 team, is it the MMR of all 3 players 1v1 that's averaged? Or the average MMR of all the 3v3 teams of all 3 players? Or the average MMR of all the teams of all 3 players? For a random 3v3 team is it each of the 3 players 3v3 Random MMR that is averaged, or would it be the average of all 3v3 teams each player has or even the average of all teams that each player has? That's the real question, isn't it? =) We're not really sure. If I had to speculate, it's probably something like "if you've played with other people in this team in other formats, use that as a starting point, otherwise use other team data, otherwise use 1v1 data." I think every team/player starts at "1500 MMR" (or whatever is the "zero"), regardless of having a master 1v1 etc. What happens is that the system is more complex than just MMR. Everyone starts with a gold level MMR, but the system will still put you against a bronze first (generally), at least on placement matches. I think the same logic applies if you are master then try a random team. You will still start with 1500 MMR, but the system will already predict you might be able to beat another master team and your placements will be with these masters and your MMR will jump fast enough to actually be able to be placed at master in this random team, even with your MMR starting at 1500. I think that is what happens, but I have no idea if it is possible to prove it somehow =( Will a master player in 1v1, that loses the 5 placements matches in random team still get a master spot? That would prove that his MMR started that high, which I doubt.
I have doubts about that. We know there's initial bootstrapping because they've said as such (it was applied about a month or so after launch).
Personally, I played 4v4 with Jathin, Vanick, and Cube. Jathin and Cube were already Master 4v4, Vanick and I were Plat 4v4. We immediately started facing Master-level 4v4 Random teams and a Master 4v4 arranged team, went 3-2, and got placed in Master. If we started with a Gold-level MMR then there's no way we would have faced Master-quality players in our first game.
|
On July 16 2011 07:37 Excalibur_Z wrote:Show nested quote +On July 16 2011 07:24 SDream wrote:On July 16 2011 06:27 Excalibur_Z wrote:On July 16 2011 05:12 Wonger wrote:Team Matchmaking
It is believed that when a team is first formed, the MMRs of all members are averaged to form the team's initial MMR during the first placement match. After that first placement match, the team's MMR is separate per normal. For Random Teams, it is likely that the MMRs of all members are averaged first, then an opposing team with a similar MMR (based upon a degree of uncertainty) is located. Do we know, or have a strong theory, as to which MMR's are used for new or random teams? For example, for a new 3v3 team, is it the MMR of all 3 players 1v1 that's averaged? Or the average MMR of all the 3v3 teams of all 3 players? Or the average MMR of all the teams of all 3 players? For a random 3v3 team is it each of the 3 players 3v3 Random MMR that is averaged, or would it be the average of all 3v3 teams each player has or even the average of all teams that each player has? That's the real question, isn't it? =) We're not really sure. If I had to speculate, it's probably something like "if you've played with other people in this team in other formats, use that as a starting point, otherwise use other team data, otherwise use 1v1 data." I think every team/player starts at "1500 MMR" (or whatever is the "zero"), regardless of having a master 1v1 etc. What happens is that the system is more complex than just MMR. Everyone starts with a gold level MMR, but the system will still put you against a bronze first (generally), at least on placement matches. I think the same logic applies if you are master then try a random team. You will still start with 1500 MMR, but the system will already predict you might be able to beat another master team and your placements will be with these masters and your MMR will jump fast enough to actually be able to be placed at master in this random team, even with your MMR starting at 1500. I think that is what happens, but I have no idea if it is possible to prove it somehow =( Will a master player in 1v1, that loses the 5 placements matches in random team still get a master spot? That would prove that his MMR started that high, which I doubt. I have doubts about that. We know there's initial bootstrapping because they've said as such (it was applied about a month or so after launch). Personally, I played 4v4 with Jathin, Vanick, and Cube. Jathin and Cube were already Master 4v4, Vanick and I were Plat 4v4. We immediately started facing Master-level 4v4 Random teams and a Master 4v4 arranged team, went 3-2, and got placed in Master. If we started with a Gold-level MMR then there's no way we would have faced Master-quality players in our first game.
rofl 3-2 to master with initial 2plat 2master MMR that can only happen on NA ^^
But appart from region difference in requirements for being placed to master, hes right. Long time ago placements started from very low level no matter what, but these days its more accurate, though still very very far from ideal state.
I have no experiences on NA realm, but in Eu even 4 master frorming AT can meet gold/plat teams in first 1-2 placement.
Still the worst part about this is that if new AT of 4 high master players plays first placement vs whatever bellow their level (gold, plat, diamond..) that RT team against them will lose more points than master team AT plays against in 5th placement. Its really sad considering that players in AT already played much more than enough games to make system sure about their skill level.
|
|
|
|
|
|
|
|
EPT
