BGH auto balance -> http://bghmmr.com/ - Page 97
Forum Index > BW General |
Deleted User 556753
126 Posts
| ||
vfsjhvfsjh
86 Posts
On September 05 2023 03:37 IAMTHEONEO wrote: SABC bots broken once again (one is offline, other one doesn't open new game) He's no longer frequently fixing the recurring issues... Looks like it's time to say goodbye to bghmmr. XD | ||
D3AD-R3TARD
Poland256 Posts
On September 03 2023 01:23 Choosy5245 wrote: 18:50 - 20:30. The guy brought his carriers to the zerg's base and was continuously killing drones. Thanks for pointing that out. YesItisCarrier was banned. | ||
patetique
97 Posts
all bots broken now.. no one working. pls restart it. best regards. | ||
leonardorm
19 Posts
User KUWA https://bghmmr.com/player.php?player=Kuwa please check Game ID: 20230916/151721 played on 2023-09-16 14:55 can Provide with Replay and screenshots if required | ||
Choosy5245
10 Posts
Thank you once again for your work! Could you please kindly explain how MMR changes are being calculated? I feel like too many times the distribution is far from being fair. Consider the following game: ![]() And here are the MMR changes: name race initial MMR MMR change SdAleks P 1598 +32.34 ChatGPT5 Z 783 +32.34 dogsowner T 771 +75.63 dogsowner scored the least in the game and gets the biggest increase. ChatGPT exactly the opposite. I believe I have lots of games like this, where I score high, but get the least MMR increase | ||
Jealous
10057 Posts
On September 22 2023 05:30 Choosy5245 wrote: Hi host, Thank you once again for your work! Could you please kindly explain how MMR changes are being calculated? I feel like too many times the distribution is far from being fair. Consider the following game: ![]() And here are the MMR changes: name race initial MMR MMR change SdAleks P 1598 +32.34 ChatGPT5 Z 783 +32.34 dogsowner T 771 +75.63 dogsowner scored the least in the game and gets the biggest increase. ChatGPT exactly the opposite. I believe I have lots of games like this, where I score high, but get the least MMR increase Score is not factored into MMR. It would be horrible if it was, for many reasons. Some examples: 1. Zerg often score less than other races due to the race and their role in team games. 2. Score is not a good determination of what a person contributed to the game; 5 pool player can make another person quit when they are rushed and have the lowest score on the team. 3. Score is affected by fairly useless things like killing buildings of a person who left. And there are many, many others. Score should never be counted in MMR. AFAIK, MMR is affected by planned variance. Meaning, if you have very few ranked games, you gain and lose more points for win/loss just to get to your rank faster. As you can tell, two people got the same MMR increase despite having very different MMRs at the start. That is because it is calculated based on team average MMR, I am guessing. It is also how the bot balances games, which again is a reason why people gain or lose more points when they are new to the ladder, so that they get placed more accurately more quickly more often. | ||
patetique
97 Posts
On September 22 2023 05:30 Choosy5245 wrote: Hi host, Thank you once again for your work! Could you please kindly explain how MMR changes are being calculated? I feel like too many times the distribution is far from being fair. Consider the following game: And here are the MMR changes: name race initial MMR MMR change SdAleks P 1598 +32.34 ChatGPT5 Z 783 +32.34 dogsowner T 771 +75.63 dogsowner scored the least in the game and gets the biggest increase. ChatGPT exactly the opposite. I believe I have lots of games like this, where I score high, but get the least MMR increase he had just 75 games total. befor 90 games he will recive big + and big - every game.. u can check stats here - http://bghmmr.com/player.php?player=dogsowner i think he will E..) PS. Last time, to much new players. system get them D-C-B.. when whey get E.. and befor, strongly unbalancing game.. | ||
Choosy5245
10 Posts
I agree that accounting for scores is not a good option, but the current scheme seems wrong as well. Basically, once you reach a certain number of games, it's really hard to progress afterwards (and to decline as well). And I see lots of A, B ranked players with lots of prior games that have the same win/ratio as E ranked players. There should be a better way. Is the MMR algorithm the same as chess platforms use? For instance, I play chess.com, it seems that ratings there reflect the actual game skills much better than in bghmmr. | ||
patetique
97 Posts
read befor pls. | ||
D3AD-R3TARD
Poland256 Posts
On September 22 2023 05:30 Choosy5245 wrote: Hi host, Thank you once again for your work! Could you please kindly explain how MMR changes are being calculated? I feel like too many times the distribution is far from being fair. Consider the following game: ![]() And here are the MMR changes: name race initial MMR MMR change SdAleks P 1598 +32.34 ChatGPT5 Z 783 +32.34 dogsowner T 771 +75.63 dogsowner scored the least in the game and gets the biggest increase. ChatGPT exactly the opposite. I believe I have lots of games like this, where I score high, but get the least MMR increase `dogsowner` got the most points because is the least known player. He played only 80 games while others played over 300. As players play more games their MMR changes less as it is less likely that the system doesn't know their real skill. On September 23 2023 01:15 Choosy5245 wrote: Ok, I got the idea. I agree that accounting for scores is not a good option, but the current scheme seems wrong as well. Basically, once you reach a certain number of games, it's really hard to progress afterwards (and to decline as well). And I see lots of A, B ranked players with lots of prior games that have the same win/ratio as E ranked players. There should be a better way. Is the MMR algorithm the same as chess platforms use? For instance, I play chess.com, it seems that ratings there reflect the actual game skills much better than in bghmmr. Again and again: winrate is not a proxy to skill. If Cristiano Ronaldo and Messi played 10 games against each other and had 50% win rate and I would win 10/10 vs blind midget with one leg, am I a better player than Ronaldo? E players can have good winrate in E class games which is not the same as having good winrate in SABC games. And yes, this algorithm is related to ELO but slightly tweaked to work with 3x3 (classic elo only works for 1x1). I do agree that this algorithm is not perfect. But I think that formula is more or less ok. The weakness is that I don't have lots of in game info so it's difficult to know what is the contribution of each player. | ||
patetique
97 Posts
On September 23 2023 04:36 D3AD-R3TARD wrote: The weakness is that I don't have lots of in game info so it's difficult to know what is the contribution of each player. no.. u have it in winscreen. who kill most units, who destroy most buildings. who lost most of all units.. etc. if u add this parameters in your formula, will realy better. just becouse to much players, selecting JUST WIN GAMES (100500% win team) and get runk up. but still total noobs.. and again. i think we have large and stable community, and time come to donate our host. no. its will NOT your new job) but we can help u with your not to easy admin work .. | ||
vfsjhvfsjh
86 Posts
| ||
Jealous
10057 Posts
On September 25 2023 01:42 patetique wrote: no.. u have it in winscreen. who kill most units, who destroy most buildings. who lost most of all units.. etc. if u add this parameters in your formula, will realy better. just becouse to much players, selecting JUST WIN GAMES (100500% win team) and get runk up. but still total noobs.. I think you're exposing your own lack of understanding with this shit take. Points, kills, buildings, etc. all are not an accurate measure of how much a person has contributed to his team. | ||
vfsjhvfsjh
86 Posts
On September 25 2023 14:12 Jealous wrote: I think you're exposing your own lack of understanding with this shit take. Points, kills, buildings, etc. all are not an accurate measure of how much a person has contributed to his team. +1. I wrote 4 or 5 lines in the post above but just deleted and you see it's blank. lol it's just wasting of my time to explain for him. | ||
patetique
97 Posts
what is the exact measure? for understanding - who will win SEE or CCD? E must get same counts like S? lets check winscreens for this type of game.. who get most counts? ofcorse its teamwork. but F get E, becouse always in "right" team.. but he still campbuilder. its right? On August 11 2023 01:31 lllllllllllllll wrote: Today I looked through the past posts in this thread, and now I am very much sure that TT_TT(BW ID) is using multiple IDs in TL forum for advoating his opinion, manipulating public opinions. I say both Jealous and vfsjhvfsjh are his TL forum ID - Both of them are Koreans, but appears around similar time, same thoughts/writing patterns with stupid logics but advocate each other all the time, both flatters to administrator so hard. Also, Jealous used to type his BW ID symbols(TT_TT) in several TL posts. barcode was right... Are you just talking to yourself, or what else?..)))) On September 23 2023 04:36 D3AD-R3TARD wrote:E players can have good winrate in E class games which is not the same as having good winrate in SABC games. allright. but according to the timetable here - http://bghmmr.com/about.php SABC have this games just 25% of all time (just 6h per 24h).. and all the rest, its SABCDEF mix games.. +50wins(16h), 4ranked (2h) SABC (6h 16:00 till 22:00 EUT) and all 75% of time, E get for win SAME bghmmr count like SAB... | ||
vfsjhvfsjh
86 Posts
On September 26 2023 06:10 patetique wrote: ok)) what is the exact measure? for understanding - who will win SEE or CCD? E must get same counts like S? lets check winscreens for this type of game.. who get most counts? ofcorse its teamwork. but F get E, becouse always in "right" team.. but he still campbuilder. its right? barcode was right... Are you just talking to yourself, or what else?..)))) allright. but according to the timetable here - http://bghmmr.com/about.php SABC have this games just 25% of all time (just 6h per 24h).. and all the rest, its SABCDEF mix games.. +50wins(16h), 4ranked (2h) SABC (6h 16:00 till 22:00 EUT) and all 75% of time, E get for win SAME bghmmr count like SAB... Live your life. I am not jealous you no brainer. actually, I don't care even though you call me Jealous. For the same, do I call you barcode as you always thumbs up whatever he said? | ||
Dio boia
36 Posts
![]() It's simple: - Who says right gets thumb-ups ![]() - Who says fool plays multiple roles, like ninza ![]() | ||
shrutipathak
India1 Post
| ||
DogsAtCards
1 Post
At least run some 50+ win games on friday and saturday evenings EU time | ||
| ||