EPTThe only way to do it 'properly' is the way that ICCup does it. That is to say have everything very open and transparent. Any and all abnormalities and disagreements were looked into case by case by the ICCup staff with replay support, etc. The problem is the size of SC2 and the amount of players playing it means that this is just logistically impossible. It's a shame, but that's the way it is.
Is A Real MMR Too Much to Ask For? - Page 4
| Forum Index > SC2 General |
|
fire_brand
Canada1123 Posts
The only way to do it 'properly' is the way that ICCup does it. That is to say have everything very open and transparent. Any and all abnormalities and disagreements were looked into case by case by the ICCup staff with replay support, etc. The problem is the size of SC2 and the amount of players playing it means that this is just logistically impossible. It's a shame, but that's the way it is. | ||
|
Warble
137 Posts
There are behavioural reasons to hide MMR. If we want to encourage more people to player more, then hiding MMR is a good idea. Blizzard demonstrated that they were willing to apply behavioural science when they hid win/loss ratios. For some players, it's annoying. For esports? Ｗａｎｔｉｎｇ ｔｏ ｓｅｅ ＭＭＲ ｈｕｒｔｓ ｅｓｐｏｒｔｓ！ | ||
|
sandyph
Indonesia1640 Posts
On June 15 2011 14:24 MuffinFTW wrote: But you can't deliberately choose your opponents on ladder, you might be able to snipe someone, but even then that still requires luck. you can leave a game on the loading screen if you know the other person have a lower ELO than you (which give minimal points on win and reduce maximal points on lose) | ||
|
FeyFey
Germany10114 Posts
So the ladder is searching equally powerfull opponents for us and it works as long as we play constantly ladder and nothing else. Sounds to me it does what it was supposed to. And elo is a way better measurement of your skill level then the mmr. Maybe remove the bonus pool, because we don't want to get rewarded for not playing, if it cuts into some rankings. I guess people like iccup because you could be proud of yourself if you reached c- tehehe and now because everyone talks the other down in sc2 ladder, everyone thinks master means you are like a d+ , so they feel like they get not enough pats on the head for being awesome. | ||
|
TheRabidDeer
United States3806 Posts
On June 15 2011 15:33 FeyFey wrote: wonder why people think mmr shows there true skill, in reality they just want to see the hidden stat. mmr is supposed to find you opponents at your skill level, so everytime something unusual happens that is outside the ladder (you make 100 training custom games), your mmr will go wild. So the ladder is searching equally powerfull opponents for us and it works as long as we play constantly ladder and nothing else. Sounds to me it does what it was supposed to. And elo is a way better measurement of your skill level then the mmr. Maybe remove the bonus pool, because we don't want to get rewarded for not playing, if it cuts into some rankings. I guess people like iccup because you could be proud of yourself if you reached c- tehehe and now because everyone talks the other down in sc2 ladder, everyone thinks master means you are like a d+ , so they feel like they get not enough pats on the head for being awesome. Your MMR isnt adjusted by non-ladder games. So if you play 100 training custom games and improve dramatically, you will have to play a lot of ladder games to allow your MMR to catch up. Also, the MMR is essentially a tweaked ELO... and the bonus pool has no effect on MMR. People see they are a 2000 point diamond league player, but that has little to no actual meaning. Different divisions in a league will have different "real" point values. If you are in a horrible division, then you actually have an artificially inflated point value by up to nearly 400 points when compared to a top class division. There is no way to know which class your division is either. I want to see my ELO/MMR because I want to know how well I rate. I want to more easily see that I am actually improving and not just going with the crowd thanks to all of the various adjustments that occur. Getting a higher rank doesnt mean I improved, it just means I played ladder games. | ||
|
[F_]aths
Germany3947 Posts
On June 15 2011 10:50 Chronald wrote: The display of the MMR would lead to a "maximize MMR" metagame. This is not what Blizzard wants.Thoughts? Comments? Flames? All are welcome. A MMR metagame also could lead to practice in custom games and to ladder only when you feel ready. The ladder game search would have a smaller playerpool because too many guys are worried about their MMR. Lets say you had a good streak and your MMR peaked. You could get afraid to lose it and stay away from laddering (either playing customs or not playing at all.) | ||
|
[F_]aths
Germany3947 Posts
On June 15 2011 16:10 TheRabidDeer wrote: Those are just points. How good are xxxx points? If you want to measure yourself, you could play in one of the many cups and see who you can beat. If you can constantly beat some names, then you know you are somewhat good. Otherwise you just have some numbers.I want to see my ELO/MMR because I want to know how well I rate. I want to more easily see that I am actually improving and not just going with the crowd thanks to all of the various adjustments that occur. Getting a higher rank doesnt mean I improved, it just means I played ladder games. | ||
|
TheRabidDeer
United States3806 Posts
On June 15 2011 16:21 [F_]aths wrote: Those are just points. How good are xxxx points? If you want to measure yourself, you could play in one of the many cups and see who you can beat. If you can constantly beat some names, then you know you are somewhat good. Otherwise you just have some numbers. I cant play in cups because I am pretty much always at work when they happen. In addition, the skill gap between us wouldnt tell me much other than "hey, that guy crushed me... i am not as good as him". Meanwhile, the MMR can directly tell me an estimate on how good I am. MMR is not points, points are just points and are visible and useless. | ||
|
Teael
United States724 Posts
When your MMR on WoW is 1800, it means you'll hit 1800 in real ratings once you play out the required amount of games. The reason why your real rating is often not your MMR is because the system requires you to play a certain amount of games before you can attain the real rating (because on WoW, get higher real rating=get better gear) Whereas On starcraft, you're placed directly into a league based on your MMR, so your league placement and your MMR won't be that different - at least not as different as it might be in WoW (there, you could have 3000 MMR, but your team rating will always start on 0. So if you had 3000 MMR but started a new team, you would have NO IDEA how good you were without them showing the MMR) Plus if they gave you a raw number (your MMR is 1200), they would also have to provide some kind of conversion table (Your MMR of 1200 is equivalent to a 1800 point platinum, after all bonus pool has been played out), which is maybe kind of a hassle, and doesn't really tell you as much as MMR on WoW might tell you (ok then how much more rating do I need to get to diamond?) Just my take on the situation - I would like to see the MMR, but I wouldn't be too fussed if I couldn't. | ||
|
[F_]aths
Germany3947 Posts
On June 15 2011 16:28 TheRabidDeer wrote: If you have no time to participate in cups, you likely have no competitive level. In this case, you could just measure the scrubness. (No offense intended, I play in Silver league.)I cant play in cups because I am pretty much always at work when they happen. In addition, the skill gap between us wouldnt tell me much other than "hey, that guy crushed me... i am not as good as him". Meanwhile, the MMR can directly tell me an estimate on how good I am. MMR is not points, points are just points and are visible and useless. Lets say you are "1000 MMR" good. What does it mean? What makes the difference between 1000 and 1015 MMR? The number itself means nothing as MMR is measured versus other players. If anyone but you gets better, your MMR gets lower even if you keep 100% of your skill level. You can roughly guess how good you are. Look at the level of your opponents (regardless if you win or lose.) I myself play almost any time versus Silver, so I am a silverplayer with no hope of promotion to gold, but no fear of demotion either. Silver means the bottom 20-40% of active players. Since I very seldom play bronze or gold, I estimate myself in the middle of silver, roughly at the skill level of the bottom 30% of active players. With random player assignments in the ladder I should lose about 70% and win about 30% of the games. You too can see if you make progress, meaning if you get better than the average skill increase – when you play versus higher league players more often. | ||
|
lololol
5198 Posts
On June 15 2011 15:04 TheRabidDeer wrote: Your win/loss is not necessarily 50%. The system tries to get you close to 50%, but it doesnt always do this. I am in diamond and I usually am closer to 60%... I just dont play enough for it to rank me up. At the end of Season 1 I was at a roughly 65% win/loss rate after like 80 games or something. The system does not try to bring you down to 50%, it tries to give you 50/50 chance of winning your next game. For example: if a player with 30 wins/10 losses plays 200 games with perfect matchmaking, he will end up with 130/110, i.e. if you had a win rate above 50% win rate before the system approximated your skill level properly, you will remain with a win rate above 50%. | ||
|
TzTz
Germany511 Posts
| ||
|
[F_]aths
Germany3947 Posts
On June 15 2011 16:50 lololol wrote: This is technically true, but it is very likely that you still will get random noise. In the end, the ratio will be close to 50%. The absolute difference will likely get higher, but the ratio is relative and will likely get closer to 50%. After a certain amount of games there is no way to tell if the deviation is caused by an initial bias during MMR stabilization or caused by noise. So any initial bias gets irrelevant over time. No one needs to worry about a positive or negative bias during MMR stabilization.The system does not try to bring you down to 50%, it tries to give you 50/50 chance of winning your next game. For example: if a player with 30 wins/10 losses plays 200 games with perfect matchmaking, he will end up with 130/110, i.e. if you had a win rate above 50% win rate before the system approximated your skill level properly, you will remain with a win rate above 50%. | ||
|
TheRabidDeer
United States3806 Posts
On June 15 2011 16:45 [F_]aths wrote: If you have no time to participate in cups, you likely have no competitive level. In this case, you could just measure the scrubness. (No offense intended, I play in Silver league.) Lets say you are "1000 MMR" good. What does it mean? What makes the difference between 1000 and 1015 MMR? It means nothing as MMR is measures versus other players. If anyone but you gets better, your MMR gets lower even if you keep 100% of your skill level. You can roughly guess how good you are. Look at the level of your opponents (regardless if you win or lose.) I myself play almost any time versus Silver, so I am a silverplayer with no hope of promotion to gold, but no fear of demotion either. Silver means the bottom 20-40% of active players. Since I very seldom play bronze or gold, I estimate myself in the middle of silver, roughly at the skill level of the bottom 30% of active players. With random player assignments in the ladder I should lose about 70% and win about 30% of the games. You too can see if you make progress, meaning if you get better than the average skill increase – when you play versus higher league players more often. MMR (or ELO) peaks for every player. It doesnt continuously rise. If the top in the world are a 2k MMR and I am at 1450, it gives me a rough idea, and it is also a much better standard of comparison. Like I said earlier, I usually have a 60-65% winrate, and I dont play very often... so I dont actually know what my real skill level is. The levels of my opponents also varies greatly, some are extremely high diamond, some are masters and others still are lower diamond. Some with lots of bonus points some with none. Most are equal favor, but some are them favored. Then you throw in the whole divisions thing where certain divisions are better than others by a certain margin and it becomes impossible to figure out your real rating. Also, just because I dont have time to participate in cups doesnt mean I have no competitive level. It just means I dont have time to participate in cups. I always play against people higher than me, but that doesnt indicate that I am improving... it just indicates that I am catching up in points and games played. Your MMR is a constant number with gradual increases/decreases. Blizzard uses the MMR because it is a constant number that accurately reflects skill. If there were other ways using the points/league to assign you an opponent they would not have even bothered to come up with a MMR. | ||
|
TheRabidDeer
United States3806 Posts
On June 15 2011 16:50 lololol wrote: The system does not try to bring you down to 50%, it tries to give you 50/50 chance of winning your next game. For example: if a player with 30 wins/10 losses plays 200 games with perfect matchmaking, he will end up with 130/110, i.e. if you had a win rate above 50% win rate before the system approximated your skill level properly, you will remain with a win rate above 50%. What you describe is a system trying to bring you down to 50%. Once you play enough games, you should be at 50%. Problem being, that may not happen for a very long time, which means you dont know your win/loss. If I go on a 50 game winning streak then go on a big losing streak with a few wins scattered in there is no way to know your win/loss without using a program to tell you. Yes, this is an exaggerated example, but it can happen... especially if you train outside of ladder games. If I start out as bronze, go 15-40 then train and win a bunch until the MMR figures out what I should be again (lets say it ends up being 80-60 and in silver now) then I train again and improve dramatically and end up being 170-100 in platinum. The system is trying to keep up, but it cant because of outside influences. So you never end up knowing your win/loss. | ||
|
[F_]aths
Germany3947 Posts
On June 15 2011 16:58 TheRabidDeer wrote: In this case, MMR will not help since it has not yet stabilized to get you about 50% win ratio. When the MMR stabilized, you can estimate your level with looking at the level of your opponents. In the current state, MMR (or Elo, btw it is Elo, not ELO) does not reflect your skill, so there is no need to see this number. The solution to your issue is not to display the (in this state very inaccurate) MMR but to play more often.MMR (or ELO) peaks for every player. It doesnt continuously rise. If the top in the world are a 2k MMR and I am at 1450, it gives me a rough idea, and it is also a much better standard of comparison. Like I said earlier, I usually have a 60-65% winrate, and I dont play very often... so I dont actually know what my real skill level is. | ||
|
oBlade
United States5821 Posts
On June 15 2011 10:55 ClysmiC wrote: I never really liked iCCup's system, as 90% of players stayed in D or D+ so it was really hard to gauge how good they were. I think Blizzard's current system isn't too bad, but I guess every ladder system is going to have flaws. This might be crass, but the truth is that most people suck. The fact that SC2 has a bunch of different leagues hides this a little bit more, but it's something that most people on any BW private server come to terms with. Perhaps a fallacy would be to think that all those people who aren't good want false goals presented at them all the time. After all, people played BW games all the time not because of where they were on any ladder, just because the game was sick fun. And as this guy says, he can't gauge so well how good he is. I don't know. | ||
|
seupac
Canada148 Posts
i didnt like it then and i dont really like it now but i suppose some people dont like the idea of facing someone much better than them, and the hidden ghost rating alleviates that i personally always improved from getting completely outplayed by someone way beyond my level so it bugs me | ||
|
TheRabidDeer
United States3806 Posts
On June 15 2011 17:05 [F_]aths wrote: In this case, MMR will not help since it has not yet stabilized to get you about 50% win ratio. When the MMR stabilized, you can estimate your level with looking at the level of your opponents. In the current state, MMR (or Elo, btw it is Elo, not ELO) does not reflect your skill, so there is no need to see this number. The solution to your issue is not to display the (in this state very inaccurate) MMR but to play more often. My MMR should be relatively accurate by now as I have played a lot of games, its obviously not perfect but it should tell me what I am close to. I get nothing from looking at who I am playing against because of reasons I have already stated in other posts (division differences, player bonus points, their MMR still tweaking, matchup performance [ie pvz or pvt or pvp] and various other things). Meanwhile I can look at my MMR and see where I actually am right at this moment and can see it move up/down as I play. | ||
|
L3g3nd_
New Zealand10461 Posts
But blizzard is too proud to implement someone elses system | ||
| ||
