EPTExperienced that also playing against him several times, just forgot once again he's doing that, and got punished immediately
BGH auto balance -> http://bghmmr.com/ - Page 37
| Forum Index > BW General |
|
Deleted User 556753
126 Posts
Experienced that also playing against him several times, just forgot once again he's doing that, and got punished immediately | ||
|
D3AD-R3TARD
Poland288 Posts
On November 10 2021 02:05 IAMTHEONEO wrote: please ban Rastamasta, he just goes afk randomly without saying a word at some point of the game, and he does that quite frequently Experienced that also playing against him several times, just forgot once again he's doing that, and got punished immediately Can you list some games where he goes AFK? | ||
|
MeSaber
Sweden1234 Posts
| ||
|
vfsjhvfsjh
92 Posts
| ||
|
Deleted User 556753
126 Posts
On November 11 2021 07:36 D3AD-R3TARD wrote: Can you list some games where he goes AFK? Game I was talking about was this one: https://repmastered.app/game/cHFbKzdJBWYbCjT5es_4nimu4SAFwlqm8ZOam80JlHs as mentioned, I have seen him doing this before several times on both sides, so I just quit meanwhile, when he ends up in my team, didn't notice it in this one though. Will see, if I find some older replays. Anyone knows a way to filter replays by player from the folder? | ||
|
Deleted User 556753
126 Posts
On November 11 2021 19:56 vfsjhvfsjh wrote: yeah, why does it matter? I know he's a poor player, but did not break a rule. well yeah, not really breaking a rule, true. Still, I think it's just super annoying and killing anyones fun in the end, when ppl do shit like that | ||
|
kogeT
Poland2029 Posts
| ||
|
kurrrak
Poland105 Posts
On November 11 2021 21:08 kogeT wrote: Out of curiosity, how does "balancing" work? Is it a sum of MMR for both teams that is compared? Or is it more sophisticated.. It's so sophisticated I once went over 1500 MMR and fell to 1000 MMR in one week. | ||
|
bbz
Sweden416 Posts
Game: 3x3 Big Game Hunters match played on 2021-11-12 10:25 Player maximusxx Reason: Giving out information from team to opponent in chat. | ||
|
kogeT
Poland2029 Posts
On November 12 2021 03:10 kurrrak wrote: It's so sophisticated I once went over 1500 MMR and fell to 1000 MMR in one week. So me being in 3000+ I realize that when I get two very low players, and I get zerg, there is usually nothing I can do to beat the opponents.  And that is usually based on sum of mmr feels like, but reality is that lower level players are way worse than mmr indicates. I mean a difference between 500 mmr and 1000 mmr is like 1st one builds cannons around his nexus while 2nd one goes 4 gate zealot pressure. Even if controlled poorly that is basicially 2 players difference in power level lol. | ||
|
Jealous
10096 Posts
On November 12 2021 20:01 kogeT wrote: So me being in 3000+ I realize that when I get two very low players, and I get zerg, there is usually nothing I can do to beat the opponents. And that is usually based on sum of mmr feels like, but reality is that lower level players are way worse than mmr indicates. I mean a difference between 500 mmr and 1000 mmr is like 1st one builds cannons around his nexus while 2nd one goes 4 gate zealot pressure. Even if controlled poorly that is basicially 2 players difference in power level lol. Unfortunately, without enough information, it is hard to accurately judge a player. For example, it's difficult to tell apart a 30 apm noob from a person who just had a few bad games if they both play just three games. With the addition of APM and previous W/L outside of BGHMMR (this is still a factor, right?) as a starting point, you can get some more accurate results, but even then there will be exceptions. There are 300 APM D- players and 120 APM A rank players in regular ladder, and there are people who play 7v1 Comp Stomps to get the initial game count needed to play while others reach that figure organically and thus start out with a 50% WR as a basis for their evaluation for BGHMMR. The system relies on people playing enough games to be accurately placed, and if I'm not mistaken, it has variance factored into it - meaning that for the first few games, you are likely to gain or lose more points as the system tries to accurately place you where you belong. So, if you lose your first 2-3 games, you are more likely to lose more points than you are to gain back if you win the next 2-3 games. All of these factors contribute to the fact that someone who hasn't played enough games in BGHMMR cannot be placed accurately. Thus, you have two people at the same rank with drastically different skill levels. Frankly, I don't think anything can be done about that - no analysis can be accurate with insufficient data and a loose hypothesis based on the average player. | ||
2Pacalypse-
Croatia9475 Posts
On November 12 2021 20:01 kogeT wrote: So me being in 3000+ I realize that when I get two very low players, and I get zerg, there is usually nothing I can do to beat the opponents. And that is usually based on sum of mmr feels like, but reality is that lower level players are way worse than mmr indicates. I mean a difference between 500 mmr and 1000 mmr is like 1st one builds cannons around his nexus while 2nd one goes 4 gate zealot pressure. Even if controlled poorly that is basicially 2 players difference in power level lol. Those are the best games to experiment and try something new :d. Just resign to the fact that you'll probably lose the game, and try to have fun. And who knows, sometimes it will work and you'll actually win some points! | ||
|
Deleted User 556753
126 Posts
On November 11 2021 20:32 IAMTHEONEO wrote: Game I was talking about was this one: https://repmastered.app/game/cHFbKzdJBWYbCjT5es_4nimu4SAFwlqm8ZOam80JlHs as mentioned, I have seen him doing this before several times on both sides, so I just quit meanwhile, when he ends up in my team, didn't notice it in this one though. Will see, if I find some older replays. Anyone knows a way to filter replays by player from the folder? yet another one which just occured (he was on the other team this time though, as mentioned, I dont play with him anymore) https://repmastered.app/game/WnVv-ADySPEFctfIRFavIEONBUBfHSj6DaUaMJA-Lps Edit: and just another one the very next game: https://repmastered.app/game/5XVzyeHny57B9pWxVl8JFQYrPJ--VGyxNRMVzCe-1l0 | ||
|
PsiAmp
Ukraine4 Posts
https://repmastered.app/game/5XVzyeHny57B9pWxVl8JFQYrPJ--VGyxNRMVzCe-1l0 | ||
|
Deleted User 556753
126 Posts
| ||
|
D3AD-R3TARD
Poland288 Posts
On November 11 2021 21:08 kogeT wrote: Out of curiosity, how does "balancing" work? Is it a sum of MMR for both teams that is compared? Or is it more sophisticated.. Team MMR is sum of all members MMRs + 30% of leader ( highest MMR player ) MMR divided by 3.3 . Balancing is finding a partition of the set of players into two teams so that difference between team MMR of both sets is minimal. On November 12 2021 19:01 bbz wrote: Ban request Game: 3x3 Big Game Hunters match played on 2021-11-12 10:25 Player maximusxx Reason: Giving out information from team to opponent in chat. Banned. I also banned Rastamasta as he repeatedly goes AFK even when game is winnable. On November 12 2021 20:01 kogeT wrote: So me being in 3000+ I realize that when I get two very low players, and I get zerg, there is usually nothing I can do to beat the opponents. And that is usually based on sum of mmr feels like, but reality is that lower level players are way worse than mmr indicates. I mean a difference between 500 mmr and 1000 mmr is like 1st one builds cannons around his nexus while 2nd one goes 4 gate zealot pressure. Even if controlled poorly that is basicially 2 players difference in power level lol. I am aware of limitations of this system. The basic assumption is that players have certain "strength" and you can combine their "strength" to calculate team "strength". The problem I see is that players have different early game, mid game and late game skill. So as you mention a 500MMR players can be comepletely useless in early game making it impossible to win. But at the same time in mid game this player can easily make half the units of a 1000MMR player. So he can be stronger than 500MMR in mid game but weaker than 500MMR in early game. It is impossible to contain all this information in one MMR number yet it is crucial to determine real chances of a team to win the game. From what I've seen with better players early game skill, mid game skill, late game skill are more balanced. But thats not the case with weaker players. Thats why you see games where worst player by MMR kills most units - once past early game his mid\late game skills were better. There are also players who can rush very well but are less powerfull in mid\late game. Thats also reflected in many games. All that shows that this assumption of player "strength" that is linear over game time is not correct. One possible solution would be to describe MMR as two or more numbers but that would complicate things and I don't really have an idea how to compute early game skill and mid/late game skil separately. Another solution would be to calculate team skill in more complicated - nonlinear way. With good ML algorithm and enought data it might be possible to get closer to real game winning chances. Players position and race composition also add a lot of noise to those calculations. On November 12 2021 22:34 Jealous wrote: Unfortunately, without enough information, it is hard to accurately judge a player. For example, it's difficult to tell apart a 30 apm noob from a person who just had a few bad games if they both play just three games. With the addition of APM and previous W/L outside of BGHMMR (this is still a factor, right?) as a starting point, you can get some more accurate results, but even then there will be exceptions. There are 300 APM D- players and 120 APM A rank players in regular ladder, and there are people who play 7v1 Comp Stomps to get the initial game count needed to play while others reach that figure organically and thus start out with a 50% WR as a basis for their evaluation for BGHMMR. The system relies on people playing enough games to be accurately placed, and if I'm not mistaken, it has variance factored into it - meaning that for the first few games, you are likely to gain or lose more points as the system tries to accurately place you where you belong. So, if you lose your first 2-3 games, you are more likely to lose more points than you are to gain back if you win the next 2-3 games. All of these factors contribute to the fact that someone who hasn't played enough games in BGHMMR cannot be placed accurately. Thus, you have two people at the same rank with drastically different skill levels. Frankly, I don't think anything can be done about that - no analysis can be accurate with insufficient data and a loose hypothesis based on the average player. APM and games other than Bot games are only a factor below first 4 MMR games. That's true that players with little to no games are rearly correctly ranked. Nevertheless as I mentioned beore, even with more games there are still problems with correct prediction of winning chances. I am aware that this hits best players the most as they are forced to play with DEF players most often and those players are least predictable. I plan on impoving this system but since that requires some serious work I can't say when I'll be able to release new version. | ||
|
MeSaber
Sweden1234 Posts
So a noob can be manipulated (ie told what to do) early game but late game they dont have the apm or knowledge to keep up with higher mmr players. At some point like 1500 mmr, players are good enough to be useful in your team, while a lesser mmr player (if let alone) will just defend himself and make team get killed. My recommendation is to ALWAYS split the lesser mmr players in separate teams and then balance the higher mmr players so each team get one super gosu and one super noob. Edit: As koget mentions, he always get the two super noobs because of his super high mmr and that literally makes it a 1v3 unless he can care enough to tell his team what to do early on. At later stages as mentioned they dont have the apm/knowledge to compete, then its koget vs rest. | ||
|
kogeT
Poland2029 Posts
| ||
|
psyCrowe
Scotland195 Posts
Sometimes having an unranked player in your team is a positive as the bot has estimated them a low MMR when they are actually pretty good. The initial MMR is also very ropey which is understandable - players getting carried to high MMR in a few games skew the balance of future games. I'm not sure how it can be fixed tbh and really is just a product of varying levels of competency of an already complex game. | ||
|
Jealous
10096 Posts
On November 14 2021 00:08 psyCrowe wrote: The MMR really is a phenomenon. I have played in teams with players rated 2k+ and they are somewhat clueless as to the general flow of the game, make very bad decisions and have no real late game knowledge (effective use of spellcasters & unit compositions), then there are some ranked lower who are much more stable/reliable partners. I don't know if it's because they play random and are possibly playing a race they aren't as competent with, but when I see a 2k player who can't micro, makes mass DT or does some dumb one base carrier build, I get incredibly frustrated :D . Sometimes having an unranked player in your team is a positive as the bot has estimated them a low MMR when they are actually pretty good. The initial MMR is also very ropey which is understandable - players getting carried to high MMR in a few games skew the balance of future games. I'm not sure how it can be fixed tbh and really is just a product of varying levels of competency of an already complex game. Should be noted that these off-beat mass DT and 1 base Carrier builds *do* sometimes win games outright. There is just a lot of variety in lower level play so sometimes things like this auto win and sometimes they auto lose and because of the number of variables/complexity as you said (positions, 3 players per team, races of respective teams, skill mix, etc.) I imagine lower level players don't have a feel when to go for them or not go for them. Some players might do them exclusively and if they win 60% of the time their rating goes up until they hit a roadblock. Also, people play drunk and do fun builds sometimes >_> Not naming names <_< | ||
| ||
