|
Do you like the current points system?
The point system definitely does its job well. That is to stimulate players to advance through leagues and generally increase their skill.
I don't think Blizzard envisioned its purpose to be measuring skill levels at any point. And we shouldn't try to measure skill with it at all.
Example:
I'm about 750 diamond player. When I play autoMM games, I get matched against decent opponents - our skill levels are pretty close therefore my win percentage is about 50.
When I try custom games, sometimes I get other diamond players. Some of them have Elo of about 1200 or higher. The skill disparity however is quite often in my favor. And what I mean is, I obliterate them. I really had no idea how bad some players in diamond were before playing a few custom 1v1 games.
Conclusion:
Blizzard does a great job determining skill levels. However not through the system of leagues and Elo numbers that are exposed to us.
What we should do:
We get weekly top 200 players of n region. These rankings are not determined by players Elo number (I think) and maybe are a result of Blizzard's hidden logic of determining skill levels. Therefore someone with vast mathematical knowledge should try to reverse engineer this hidden logic by looking at consequentive top 200 lists and the games players in those lists played during the time between the lists being published.
Once we know the algorithm, we could use it to measure skill levels between players the same way the AutoMM system does.
Until Blizzard decides to change their logic, of course 
|
On September 20 2010 16:17 Zamkis wrote:
The main problem isn't the point inflation really, it's the fact that we have no other means to compare skill levels. While points weren't used much in ICCUP, ranks at least meant something and could be used to judge someone's skill, thus they could have implemented the point inflation feature and we wouldn't have cared much. In other words, don't fix the points, fix the leagues!
What's wrong with the leagues?
|
On September 20 2010 15:12 cHaNg-sTa wrote:
iCCup's system is great. Too bad Blizzard won't like the idea of having a casual player constantly have a D or D- rating by their name.Maybe have an A to D rating within each league? I'm a Diamond C player, or Gold B player, etc.
I love the ICCup/PGtour system as well. Hell, even WGtour was better than what Blizzard has implemented.
Anyone who is even remotely serious about the game is in Diamond, and because Diamond is as far as you can go, a gosu will be in the same leagues as would be D players. The point system then gets these casuals even closer to the range of the gosu players, because even on days when I lose more than I win, I am still up 15 or so points.
I really wish the ICCup system would be implemented, and the random matchmaking would match you up per letter grade. The way I see it is because Diamond is such a broad range and the hidden skill number factor likes to sometimes place you against gosus, and sometimes against platinums, it could just be throwing you into a match with another player your current rank.
|
I just want to see my AMM rank. Blizzard, if you're listening, PLEASE give us the option of looking at our AMM rank!
Edit: I also would REALLY like to see a ranking subsystem in Diamond to distinguish the best from the rest, a-la iccup.
|
On September 20 2010 16:19 Schplyok wrote:Do you like the current points system?The point system definitely does its job well. That is to stimulate players to advance through leagues and generally increase their skill. I don't think Blizzard envisioned its purpose to be measuring skill levels at any point. And we shouldn't try to measure skill with it at all. Example: I'm about 750 diamond player. When I play autoMM games, I get matched against decent opponents - our skill levels are pretty close therefore my win percentage is about 50. When I try custom games, sometimes I get other diamond players. Some of them have Elo of about 1200 or higher. The skill disparity however is quite often in my favor. And what I mean is, I obliterate them. I really had no idea how bad some players in diamond were before playing a few custom 1v1 games. Conclusion: Blizzard does a great job determining skill levels. However not through the system of leagues and Elo numbers that are exposed to us. What we should do: We get weekly top 200 players of n region. These rankings are not determined by players Elo number (I think) and maybe are a result of Blizzard's hidden logic of determining skill levels. Therefore someone with vast mathematical knowledge should try to reverse engineer this hidden logic by looking at consequentive top 200 lists and the games players in those lists played during the time between the lists being published. Once we know the algorithm, we could use it to measure skill levels between players the same way the AutoMM system does. Until Blizzard decides to change their logic, of course
This is complete nonsense.
The truth is the MMR or ELO or whatever you want to call it is very capable and successful in measuring skill, as evidenced by having nearly all players with a win record close to 50%.
|
On September 20 2010 16:25 GagnarTheUnruly wrote:
I just want to see my AMM rank. Blizzard, if you're listening, PLEASE give us the option of looking at our AMM rank!
Edit: I also would REALLY like to see a ranking subsystem in Diamond to distinguish the best from the rest, a-la iccup.
I would like a global ladder too.
But that's never going to happen. Because its more important not to hurt the feelings of bad players than to make a correct and accurate ladder ranking system.
Until then the best we've got is sc2ranks, which does break diamond league into percentiles.
|
If the system is doing its job, everyone but the very best and very worst would have a 50%
Untrue. If the system is doing his job, win% shows how fast you learn in comparison with other players and/or how much experience you had with SC(or RTS) before. And I think the matchmaking system is doing an A-OK job with that. Let me try to explain it a bit further.
Player A:
- Never played an RTS before.
- Bought SCII and immediately started playing Multiplayer
- Learns the game by playing Multiplayer over and over.
----> Will have around 48-50% after 200 games and is stuck in bronze because he will be matched up against other people who have no RTS background and don't use TL.net or youtube or w/e to progress.
Player B:
- played WC3 ROC and TFT for 5 years straight but took a 3 year break from RTS.
- Bought SCII and immediately starts playing Multiplayer.
- Starts in Bronze and learns faster than others how to play this game due to his background.
----> Will have around 55% wins after around 30 games.
- Promotes to Gold and he starts losing games. Stops playing multiplayer
- Studies TL.net for BO's and watches all dailies from Day9.
- Finds himself a practice partner in Platinum.
----> Will promote to diamond after +- 100-200 games with a 55-60 win%. Which both depends on how much he learns from external sources.
Player C:
- Small RTS background
- Got into Beta.
- Played 500 games in Beta.
- Bought SCII and immediately starts playing Multiplayer.
- Starts in Platinum and goes fast to Diamond. Keeps playing many games and is high on points.
------> Will have 50% after 200 games and faced top players in these 200 games.
We can all agree that player C is the most skilled player here. However, player B has the highest win%. These examples do not justify my statement that win% shows your learning progress towards other players because there are some other factors you need to take into account. A couple listen below:
- Amount of custom games. ( ex. If you play a lot of tournaments from external sites)
- Amount of games you play (ex. By playing one game a day as an experienced RTS-player, you will face less skilled but higher on point opponents, which boost your win%)
- Amount of time learning outside the game. ( ex. TL.net, youtube)
|
On September 20 2010 16:30 paralleluniverse wrote:Show nested quote +On September 20 2010 16:19 Schplyok wrote:Do you like the current points system?The point system definitely does its job well. That is to stimulate players to advance through leagues and generally increase their skill. I don't think Blizzard envisioned its purpose to be measuring skill levels at any point. And we shouldn't try to measure skill with it at all. Example: I'm about 750 diamond player. When I play autoMM games, I get matched against decent opponents - our skill levels are pretty close therefore my win percentage is about 50. When I try custom games, sometimes I get other diamond players. Some of them have Elo of about 1200 or higher. The skill disparity however is quite often in my favor. And what I mean is, I obliterate them. I really had no idea how bad some players in diamond were before playing a few custom 1v1 games. Conclusion: Blizzard does a great job determining skill levels. However not through the system of leagues and Elo numbers that are exposed to us. What we should do: We get weekly top 200 players of n region. These rankings are not determined by players Elo number (I think) and maybe are a result of Blizzard's hidden logic of determining skill levels. Therefore someone with vast mathematical knowledge should try to reverse engineer this hidden logic by looking at consequentive top 200 lists and the games players in those lists played during the time between the lists being published. Once we know the algorithm, we could use it to measure skill levels between players the same way the AutoMM system does. Until Blizzard decides to change their logic, of course This is complete nonsense. The truth is the MMR or ELO or whatever you want to call it is very capable and successful in measuring skill, as evidenced by having nearly all players with a win record close to 50%.
So, you are saying if we factor in bonus pool and look at the numbers, we will have an adequate measure of skill?
|
I don't like it anymore because i got to 1300 earlier today, played like 10 more games, and ended at 1295.
QQ i think I've hit my limit.
|
On September 18 2010 20:16 Numy wrote:Show nested quote +On August 29 2010 22:40 Klockan3 wrote:On August 29 2010 19:13 Inori wrote:
Why this bothers me is because this system has no stable grounds - a way to know what level a player is based on his rating in any given time. For example in chess 1500 is mid, 1800 - good and 2200+ is pro level. This was true 30 years ago. This will be true 30 years from now. In SC2, however, 2-3 months from now everyone will have 5k+ points. And in another 2-3 months, what? 9999 points? And after that? Sure, resets may come, but they don't really solve the core of the problem.
You don't know how the rating works, not at all. None does but the designers working for Blizzard. It could be that the bonus pool artificially inflates the ratings but we don't know that and we wont know that until either Blizzard makes an official statement or in at least a few months. And look at these arguments: 1. It takes a long while for the system to stabilize when everyone starts out at x rating. chess haven't had a situation like this in ages. 2. If suddenly the whole world started to play competitive chess then everyones chess ranking would get a huge increase simply because the system puts all new players at the same ranking and most of the world are way worse than the worst parts of the competitive chess community. 3. The top players scores increases faster than the bonus pool adds points. You are completely missing the point. Bonus pool creates a large inflation. This is a FACT. No matter how match making works the points you see in your profile is increasing. This creates massive problems for the player. Now if I want to know if I have improved there is no baseline for me to compare myself to since 1000 points. today could be the same as 500 points last week. This inflation is stupid and serves only to boost the egos of those that don't understand. Bonus pool inflates POINTS - Rating is something we don't see and will most likely not see for a long time so it's meaningless as a visual indicator of improvement. EDIT: Show nested quote +On September 18 2010 20:11 TheRabidDeer wrote:
I hate the system, I dont get to play very much so I am constantly behind everybody because of the bonus points. I have about 650 points in diamond but my hidden rating has me consistently getting matched against people that are at 1100-1200 now.
A couple weeks ago it was me against 800 point players, now its against 1200 point players.
(I am 38-24 at this point) I get the same thing. I'm around 800 points and always get matched up against 1200-1500 points players. The actual matchmaking is fine, I really enjoy it but what I don't understand is why the system has to hide the fact that I am a 1200-1500 player because of this stupid inflation. I want to be able to look at my points and get a rough idea of if I have improved from last week or not.
Probably you are missing the points? Please read Excalibur_Z articles (mentioned later in this thread) and the comments about bonus pools. Stating that inflation is A FACT can only be a true if you really prove it (mathematically), which you can't as you don't know the details.
The article describes quite well what is happening and it is quite certain (and well accepted) that the bonus pool only increases your shown points not the hidden MMR.
So, there is some (!) inflation in the shown points, but definitely only a certain (not infinite!!!) amount. Why?
The points gained after a win are calculated comparing your shown points with the opponents hidden MMR. If your points are inflated (the MMR is not, because no bonus pool), they will be much higher than the opponents MMR and, thus, the points you gain by winning will be getting lower and lower. The bonus points are only a constant to be added determined by the 1 point per 2 hours rate.
To make it even more clear: if 2 top players play against each other, they should both (!) see that they are favored compared to each other. Thus, winning won't get them many points.
Now you could discuss if this makes sense or not (both seeing themselves as favored if they are actually at the same level, meaning that MMR and shown points are always separated by the aforementioned constant, but it will stop inflation.)
|
On September 20 2010 16:34 Schplyok wrote:Show nested quote +On September 20 2010 16:30 paralleluniverse wrote:On September 20 2010 16:19 Schplyok wrote:Do you like the current points system?The point system definitely does its job well. That is to stimulate players to advance through leagues and generally increase their skill. I don't think Blizzard envisioned its purpose to be measuring skill levels at any point. And we shouldn't try to measure skill with it at all. Example: I'm about 750 diamond player. When I play autoMM games, I get matched against decent opponents - our skill levels are pretty close therefore my win percentage is about 50. When I try custom games, sometimes I get other diamond players. Some of them have Elo of about 1200 or higher. The skill disparity however is quite often in my favor. And what I mean is, I obliterate them. I really had no idea how bad some players in diamond were before playing a few custom 1v1 games. Conclusion: Blizzard does a great job determining skill levels. However not through the system of leagues and Elo numbers that are exposed to us. What we should do: We get weekly top 200 players of n region. These rankings are not determined by players Elo number (I think) and maybe are a result of Blizzard's hidden logic of determining skill levels. Therefore someone with vast mathematical knowledge should try to reverse engineer this hidden logic by looking at consequentive top 200 lists and the games players in those lists played during the time between the lists being published. Once we know the algorithm, we could use it to measure skill levels between players the same way the AutoMM system does. Until Blizzard decides to change their logic, of course This is complete nonsense. The truth is the MMR or ELO or whatever you want to call it is very capable and successful in measuring skill, as evidenced by having nearly all players with a win record close to 50%. So, you are saying if we factor in bonus pool and look at the numbers, we will have an adequate measure of skill?
Well, sort of. Yes, it would be a good measure of skill.
But I also did say this: http://www.teamliquid.net/forum/viewmessage.php?topic_id=143158
Which I stand by completely.
|
The only thing that bugs me about the current system is, that both scenarious have its disadvantages.
If you have bonuspool - you can loose 2 times in a row and still gain points with 1 win over a favoured or even opponent. This is ok in my opinion on diamond but if I go back to platinum I almost can loose almost 5-10 games and win 1 and still will gain some few points if the win is vs a favoured player. How is it in gold (I dont know) then?
Without bonuspool tough - you loose one game to cheese, loose 13 points. Then you win and gain 12 points - huh?
I think the system has some major flaws in it. To be honest, those couldnt be set on the beta since the size of the system is way higher than during beta. Tough they should adapt the numbers to the current inflation like a currency.
|
On September 20 2010 16:39 Drazzzt wrote:Show nested quote +On September 18 2010 20:16 Numy wrote:On August 29 2010 22:40 Klockan3 wrote:On August 29 2010 19:13 Inori wrote:
Why this bothers me is because this system has no stable grounds - a way to know what level a player is based on his rating in any given time. For example in chess 1500 is mid, 1800 - good and 2200+ is pro level. This was true 30 years ago. This will be true 30 years from now. In SC2, however, 2-3 months from now everyone will have 5k+ points. And in another 2-3 months, what? 9999 points? And after that? Sure, resets may come, but they don't really solve the core of the problem.
You don't know how the rating works, not at all. None does but the designers working for Blizzard. It could be that the bonus pool artificially inflates the ratings but we don't know that and we wont know that until either Blizzard makes an official statement or in at least a few months. And look at these arguments: 1. It takes a long while for the system to stabilize when everyone starts out at x rating. chess haven't had a situation like this in ages. 2. If suddenly the whole world started to play competitive chess then everyones chess ranking would get a huge increase simply because the system puts all new players at the same ranking and most of the world are way worse than the worst parts of the competitive chess community. 3. The top players scores increases faster than the bonus pool adds points. You are completely missing the point. Bonus pool creates a large inflation. This is a FACT. No matter how match making works the points you see in your profile is increasing. This creates massive problems for the player. Now if I want to know if I have improved there is no baseline for me to compare myself to since 1000 points. today could be the same as 500 points last week. This inflation is stupid and serves only to boost the egos of those that don't understand. Bonus pool inflates POINTS - Rating is something we don't see and will most likely not see for a long time so it's meaningless as a visual indicator of improvement. EDIT: On September 18 2010 20:11 TheRabidDeer wrote:
I hate the system, I dont get to play very much so I am constantly behind everybody because of the bonus points. I have about 650 points in diamond but my hidden rating has me consistently getting matched against people that are at 1100-1200 now.
A couple weeks ago it was me against 800 point players, now its against 1200 point players.
(I am 38-24 at this point) I get the same thing. I'm around 800 points and always get matched up against 1200-1500 points players. The actual matchmaking is fine, I really enjoy it but what I don't understand is why the system has to hide the fact that I am a 1200-1500 player because of this stupid inflation. I want to be able to look at my points and get a rough idea of if I have improved from last week or not. Probably you are missing the points? Please read Excalibur_Z articles (mentioned later in this thread) and the comments about bonus pools. Stating that inflation is A FACT can only be a true if you really prove it (mathematically), which you can't as you don't know the details. The article describes quite well what is happening and it is quite certain (and well accepted) that the bonus pool only increases your shown points not the hidden MMR. So, there is some (!) inflation in the shown points, but definitely only a certain (not infinite!!!) amount. Why? The points gained after a win are calculated comparing your shown points with the opponents hidden MMR. If your points are inflated (the MMR is not, because no bonus pool), they will be much higher than the opponents MMR and, thus, the points you gain by winning will be getting lower and lower. The bonus points are only a constant to be added determined by the 1 point per 2 hours rate. To make it even more clear: if 2 top players play against each other, they should both (!) see that they are favored compared to each other. Thus, winning won't get them many points. Now you could discuss if this makes sense or not (both seeing themselves as favored if they are actually at the same level, meaning that MMR and shown points are always separated by the aforementioned constant, but it will stop inflation.)
That is not correct.
There is no evidence to suggest that: "the points you gain by winning will be getting lower and lower".
Points will most likely increase indefinitely until a ladder reset, and converge not to MMR, but rather, (MMR + 0.5t), where t is the number of hours from the beginning of the season to right now.
|
On September 20 2010 08:33 TheOGBlitzKrieg wrote:Show nested quote +On September 20 2010 06:57 TheRabidDeer wrote:On September 20 2010 06:53 TheOGBlitzKrieg wrote:
i like the bonus points because some people don't get a lot of time to play but still are really good players, and i'm serious about this it is true although not very often... but for those who don't get the time to play a lot the bonus pool system act's sort of like rested xp in wow it allows you to atleast somewhat keep up with the ppl who play 500 games a week Except the people that play 500 games a week are where they would be anyway. The system is designed such that you would normally stay around a certain amount of points. You would win some games, and lose some... you would find your own cap. Then they add in bonus points, which inflates that cap and makes it so people like me (the people that dont get a lot of time to play) actually fall farther behind. i was actually referring to myself too the bonus pool helps players who don't play as often stay caught up with the competition not fall behind by giving you bonus points for not playing as much... how does this make you fall farther behind than if they didn't give you any bonus points?
It is not a system that promotes "catching up". Everybody gets bonus points, which means that any distance that you might be away from somebody that is at your skill level is actually just there BECAUSE of the bonus points
|
On September 20 2010 15:17 Sentient wrote:Show nested quote +On September 20 2010 14:57 Rabiator wrote:
Personally I think that "higher win-rate = better player", but the points system allows a player to get to the top by simply playing more games than a competitor with a higher win rate. A win rate would also be easily useable to put someone in a higher or lower league. If the system is doing its job, everyone but the very best and very worst would have a 50% win rate.
The current system isnt doing its job then. Just from the worldwide ladder system we can see
#1 Huk, 2010 points, 75.66% win rate
#2 Fenix, 2003 points, 70.27% win rate
#3 dayvie, 1995 points, 57.81% win rate
Huk and Fenix being "at the top" is ok, dayvie shouldnt be there with such a low win rate. Better players win have a higher chance of winning a game, but currently the points system puts too many people at the top of the rankings who are simply massing their games instead (and with a win rate of less than 60%) of improving their quality as a player. That is wrong and needs to be changed.
http://www.sc2ranks.com/#ratio:0
That shows the top 100 players and there is a win rate fluctuation from 81% (DeMusliM) to 52%. That is hardly good.
|
On September 20 2010 16:40 paralleluniverse wrote:Show nested quote +On September 20 2010 16:34 Schplyok wrote:On September 20 2010 16:30 paralleluniverse wrote:On September 20 2010 16:19 Schplyok wrote:Do you like the current points system?The point system definitely does its job well. That is to stimulate players to advance through leagues and generally increase their skill. I don't think Blizzard envisioned its purpose to be measuring skill levels at any point. And we shouldn't try to measure skill with it at all. Example: I'm about 750 diamond player. When I play autoMM games, I get matched against decent opponents - our skill levels are pretty close therefore my win percentage is about 50. When I try custom games, sometimes I get other diamond players. Some of them have Elo of about 1200 or higher. The skill disparity however is quite often in my favor. And what I mean is, I obliterate them. I really had no idea how bad some players in diamond were before playing a few custom 1v1 games. Conclusion: Blizzard does a great job determining skill levels. However not through the system of leagues and Elo numbers that are exposed to us. What we should do: We get weekly top 200 players of n region. These rankings are not determined by players Elo number (I think) and maybe are a result of Blizzard's hidden logic of determining skill levels. Therefore someone with vast mathematical knowledge should try to reverse engineer this hidden logic by looking at consequentive top 200 lists and the games players in those lists played during the time between the lists being published. Once we know the algorithm, we could use it to measure skill levels between players the same way the AutoMM system does. Until Blizzard decides to change their logic, of course This is complete nonsense. The truth is the MMR or ELO or whatever you want to call it is very capable and successful in measuring skill, as evidenced by having nearly all players with a win record close to 50%. So, you are saying if we factor in bonus pool and look at the numbers, we will have an adequate measure of skill? Well, sort of. Yes, it would be a good measure of skill. ...
Then why do people get matched against players with various amount of points when laddering? Example: you can be favored vs someone who has higher Elo rank that you (and lets say you both have an empty bonus pool).
I also think the Elo system is quite capable of measuring skill. The thing is Blizzard has made modifications to it so that you gain points faster and seeing your points increase when laddering is not a slow and tedious process. The most obvious of these modifications is the bonus pool but maybe its not the only one.
On the other hand they can rank players pretty well somehow (as evidenced by everyone but the best getting ~50% win rate). We don't know that they use for that, maybe another Elo system with some other modifications that we don't know.
So, what I'm saying is that we have 2 systems - one designed to make people have a sense of accomplishment when laddering, and another that ensures everyone is facing an opponent of similar skill.
If we as a community want to have a system that measures skill, we should either somehow use Blizzards hidden system or create our own.
|
On September 20 2010 16:53 Rabiator wrote:Show nested quote +On September 20 2010 15:17 Sentient wrote:On September 20 2010 14:57 Rabiator wrote:
Personally I think that "higher win-rate = better player", but the points system allows a player to get to the top by simply playing more games than a competitor with a higher win rate. A win rate would also be easily useable to put someone in a higher or lower league. If the system is doing its job, everyone but the very best and very worst would have a 50% win rate. The current system isnt doing its job then. Just from the worldwide ladder system we can see #1 Huk, 2010 points, 75.66% win rate #2 Fenix, 2003 points, 70.27% win rate #3 dayvie, 1995 points, 57.81% win rate Huk and Fenix being "at the top" is ok, dayvie shouldnt be there with such a low win rate. Better players win have a higher chance of winning a game, but currently the points system puts too many people at the top of the rankings who are simply massing their games instead (and with a win rate of less than 60%) of improving their quality as a player. That is wrong and needs to be changed. http://www.sc2ranks.com/#ratio:0That shows the top 100 players and there is a win rate fluctuation from 81% (DeMusliM) to 52%. That is hardly good.
If you look at the top of the ladder, there is a negative correlation between games played and high ranks.
|
Ive given up on the ladder as a method of rating skill. Instead ive been using sc2ranks.com to track my US server rank. At least this way i can tell if im really improving or simply rising in points due to inflation. I really wish blizzard would implement a true elo ladder. Although with so much of the focus in sc2 on casual gamers I doubt it ever will.
|
On September 20 2010 16:57 Schplyok wrote:Show nested quote +On September 20 2010 16:40 paralleluniverse wrote:On September 20 2010 16:34 Schplyok wrote:On September 20 2010 16:30 paralleluniverse wrote:On September 20 2010 16:19 Schplyok wrote:Do you like the current points system?The point system definitely does its job well. That is to stimulate players to advance through leagues and generally increase their skill. I don't think Blizzard envisioned its purpose to be measuring skill levels at any point. And we shouldn't try to measure skill with it at all. Example: I'm about 750 diamond player. When I play autoMM games, I get matched against decent opponents - our skill levels are pretty close therefore my win percentage is about 50. When I try custom games, sometimes I get other diamond players. Some of them have Elo of about 1200 or higher. The skill disparity however is quite often in my favor. And what I mean is, I obliterate them. I really had no idea how bad some players in diamond were before playing a few custom 1v1 games. Conclusion: Blizzard does a great job determining skill levels. However not through the system of leagues and Elo numbers that are exposed to us. What we should do: We get weekly top 200 players of n region. These rankings are not determined by players Elo number (I think) and maybe are a result of Blizzard's hidden logic of determining skill levels. Therefore someone with vast mathematical knowledge should try to reverse engineer this hidden logic by looking at consequentive top 200 lists and the games players in those lists played during the time between the lists being published. Once we know the algorithm, we could use it to measure skill levels between players the same way the AutoMM system does. Until Blizzard decides to change their logic, of course This is complete nonsense. The truth is the MMR or ELO or whatever you want to call it is very capable and successful in measuring skill, as evidenced by having nearly all players with a win record close to 50%. So, you are saying if we factor in bonus pool and look at the numbers, we will have an adequate measure of skill? Well, sort of. Yes, it would be a good measure of skill. ... Then why do people get matched against players with various amount of points when laddering? Example: you can be favored vs someone who has higher Elo rank that you (and lets say you both have an empty bonus pool).
Because of "expanding search".
I also think the Elo system is quite capable of measuring skill. The thing is Blizzard has made modifications to it so that you gain points faster and seeing your points increase when laddering is not a slow and tedious process. The most obvious of these modifications is the bonus pool but maybe its not the only one.
Yes.
On the other hand they can rank players pretty well somehow (as evidenced by everyone but the best getting ~50% win rate). We don't know that they use for that, maybe another Elo system with some other modifications that we don't know.
So, what I'm saying is that we have 2 systems - one designed to make people have a sense of accomplishment when laddering, and another that ensures everyone is facing an opponent of similar skill.
If we as a community want to have a system that measures skill, we should either somehow use Blizzards hidden system or create our own.
We have 2 systems: 1) MMR to measure skill and for matchmaking, 2) points and bonus pool (which can be thought of as a combination measuring skill and activity) used for ranking.
|
On September 20 2010 16:44 paralleluniverse wrote:Show nested quote +On September 20 2010 16:39 Drazzzt wrote:On September 18 2010 20:16 Numy wrote:On August 29 2010 22:40 Klockan3 wrote:On August 29 2010 19:13 Inori wrote:
Why this bothers me is because this system has no stable grounds - a way to know what level a player is based on his rating in any given time. For example in chess 1500 is mid, 1800 - good and 2200+ is pro level. This was true 30 years ago. This will be true 30 years from now. In SC2, however, 2-3 months from now everyone will have 5k+ points. And in another 2-3 months, what? 9999 points? And after that? Sure, resets may come, but they don't really solve the core of the problem.
You don't know how the rating works, not at all. None does but the designers working for Blizzard. It could be that the bonus pool artificially inflates the ratings but we don't know that and we wont know that until either Blizzard makes an official statement or in at least a few months. And look at these arguments: 1. It takes a long while for the system to stabilize when everyone starts out at x rating. chess haven't had a situation like this in ages. 2. If suddenly the whole world started to play competitive chess then everyones chess ranking would get a huge increase simply because the system puts all new players at the same ranking and most of the world are way worse than the worst parts of the competitive chess community. 3. The top players scores increases faster than the bonus pool adds points. You are completely missing the point. Bonus pool creates a large inflation. This is a FACT. No matter how match making works the points you see in your profile is increasing. This creates massive problems for the player. Now if I want to know if I have improved there is no baseline for me to compare myself to since 1000 points. today could be the same as 500 points last week. This inflation is stupid and serves only to boost the egos of those that don't understand. Bonus pool inflates POINTS - Rating is something we don't see and will most likely not see for a long time so it's meaningless as a visual indicator of improvement. EDIT: On September 18 2010 20:11 TheRabidDeer wrote:
I hate the system, I dont get to play very much so I am constantly behind everybody because of the bonus points. I have about 650 points in diamond but my hidden rating has me consistently getting matched against people that are at 1100-1200 now.
A couple weeks ago it was me against 800 point players, now its against 1200 point players.
(I am 38-24 at this point) I get the same thing. I'm around 800 points and always get matched up against 1200-1500 points players. The actual matchmaking is fine, I really enjoy it but what I don't understand is why the system has to hide the fact that I am a 1200-1500 player because of this stupid inflation. I want to be able to look at my points and get a rough idea of if I have improved from last week or not. Probably you are missing the points? Please read Excalibur_Z articles (mentioned later in this thread) and the comments about bonus pools. Stating that inflation is A FACT can only be a true if you really prove it (mathematically), which you can't as you don't know the details. The article describes quite well what is happening and it is quite certain (and well accepted) that the bonus pool only increases your shown points not the hidden MMR. So, there is some (!) inflation in the shown points, but definitely only a certain (not infinite!!!) amount. Why? The points gained after a win are calculated comparing your shown points with the opponents hidden MMR. If your points are inflated (the MMR is not, because no bonus pool), they will be much higher than the opponents MMR and, thus, the points you gain by winning will be getting lower and lower. The bonus points are only a constant to be added determined by the 1 point per 2 hours rate. To make it even more clear: if 2 top players play against each other, they should both (!) see that they are favored compared to each other. Thus, winning won't get them many points. Now you could discuss if this makes sense or not (both seeing themselves as favored if they are actually at the same level, meaning that MMR and shown points are always separated by the aforementioned constant, but it will stop inflation.) That is not correct. There is no evidence to suggest that: "the points you gain by winning will be getting lower and lower". Points will most likely increase indefinitely until a ladder reset, and converge not to MMR, but rather, (MMR + 0.5t), where t is the number of hours from the beginning of the season to right now.
When stating something like this you should at least try to explain the reasoning behind it....
What you are missing here is that you won't win all of your games.....you will lets say lose 50%
Lets say you play 1 game an hour, constantly. you lose one, you win one, you lose one, you win one and so on.
Your points are much higher than your MMR (as there is no bonus pool). And your equally skilled opponent has the same MMR and points. So, your points are compared to his MMR, his points are compared to your MMR.
So, if you win you get 1-3 normal points (as you are favored, your points >> his MMR) plus one bonus pool point (you win every two hours). If you lose you will lose 18-24 points as you are favored.
So, you win 2-4 and lose 18-24 points, win 2-4, lose 18-24.....How will this increase indefinitely?
|
|
|
|
|
|
EPT
